EdTech & AI

When not to use AI in language learning is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Allow the tool to be unnecessary

A method or tool should earn its place by improving a defined opportunity to learn. If a simpler route protects attention, access or dialogue, that is evidence for choosing the simpler route.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Teaching a digital skill in stages is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Break the skill at a natural boundary

Stages should reflect real decisions within a skill: noticing, planning, selecting, producing, checking and revising. Avoid stages that exist only because a worksheet has sections.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Designing a second attempt with AI is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Give the second attempt a new reason

Changing the audience, time limit or evidence makes a second attempt more than a corrected copy. It asks learners to reuse the same learning in a slightly different situation.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Reflective questions for AI use is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Ask questions that produce evidence

A reflective question should lead to an example, a comparison or a plan. “What did you learn?” is broad; “Which phrase did you choose differently the second time?” invites evidence.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Connecting tools across a learning programme is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Connect routines across time

A small routine becomes part of a programme when it returns with a clear purpose. Signal the connection so learners can recognise that they are practising a durable strategy, not completing another isolated task.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Language support in AI-assisted tasks is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Support the language around the task

A language bank, rehearsal frame or example can lower an irrelevant barrier. The support should make participation possible without pre-writing the decision learners need to make.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Peer work with digital accountability is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Design accountability without surveillance

Clear roles, a shared product and a brief report-back can make collaborative work purposeful. Constant monitoring is rarely necessary when the task itself requires reciprocal contribution.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Helping AI-supported learning transfer is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Test use in a changed context

A later task with a new topic, audience or medium is a stronger indication of transfer than repeating the same item. Keep the change manageable enough that the underlying decision remains recognizable.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Reviewing learner evidence after using AI is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Review patterns rather than isolated moments

One response may be affected by timing, mood, unfamiliar content or a technical problem. Look across examples before changing a course, label or expectation.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Making AI-supported learning visible is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Define participation more carefully

Speaking first is not the only sign of engagement. A useful design makes preparation, listening, drafting, questioning and revision visible alongside public contribution.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Adapting a digital routine is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Keep the purpose while changing the route

Adaptation works best when the core learning decision remains stable. Change the format, support or timing as needed, but be clear about what learners are still expected to work out.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

What to measure when using edtech is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Measure what the decision requires

Do not let an easy-to-count measure replace the thing you actually value. If the goal involves communicating, include evidence of communication; if it involves judgement, include a choice with reasons.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Protecting attention in digital learning is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Protect attention as a learning resource

Every new instruction, tab, tool or criterion asks for attention. Remove anything that does not make the target decision clearer or more practicable.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Structure and agency in AI-supported learning is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Use structure to create room for choice

A clear sequence can free learners to make better decisions. The structure should handle the predictable parts—purpose, time, materials and turn-taking—so attention can go to the language or idea that matters.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Using AI feedback without outsourcing judgement is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Make revision part of the task

Feedback has more value when the task includes time to use it. Plan the second attempt when writing the first task, rather than treating revision as optional work if time remains.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Models, examples and AI-generated text is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Use models as lenses, not scripts

A model can show what success may look like. Invite learners to notice choices in it, then change the topic, audience or constraint so their own response has to do new work.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Independent and collaborative work with AI is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Choose the order of interaction deliberately

Independent thought, paired rehearsal and shared discussion each produce different evidence. The sequence should reflect the learning purpose rather than a habit of always starting in the same format.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

A closing question for digital learning is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

End while the evidence is still visible

A short closing question gives learners a chance to name what changed and gives teachers a clue about what to revisit. It is more useful than asking only whether they enjoyed the task.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

How to choose AI for English learning is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Start from the outcome

Describe the performance before collecting resources. This makes the article’s central recommendation testable: readers can ask whether each part of the design helps someone move towards that performance.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Evidence before adopting an edtech tool is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Decide what evidence would count

A score can be one form of evidence, but a sample of language, a recorded explanation, a peer response or a later transfer task may reveal a different part of the picture.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Confidence and risk with AI is not a technique that works independently of the people and conditions around it. In educational technology and AI, the central design question is usually simple: what should learners be able to do differently after this sequence, and what would count as convincing evidence? That question prevents a familiar trap—mistaking activity for learning because the room, screen or worksheet looks busy.

Define the decision before choosing the resource

A resource can be excellent and still be the wrong first move. Before selecting one, write a single sentence that names the decision learners will make. It might be how to organise a response, how to notice a recurring language choice, or how to choose between two options for a real audience. That sentence creates a boundary: anything that does not help learners make the decision can be shortened, delayed or removed.

This does not require a rigid lesson script. It gives teachers and learners a shared purpose. A group may arrive at different language, examples or strategies, but they should know what the task is asking them to work out. When the target is visible, feedback can address the work rather than the person.

Build a sequence that permits a second attempt

The first attempt should reveal something worth responding to. Use a brief model, a worked example, an information gap or a short rehearsal to help people enter the task. Then make the first attempt small: a thirty-second explanation, a paragraph outline, a recorded answer, or a decision with reasons. The smaller the first attempt, the easier it is to notice one useful thing and try again.

The second attempt is where a routine earns its place. It can change the audience, constraint, evidence or language support while preserving the core performance. That makes improvement visible without pretending that one revised answer proves mastery. It also gives learners a reason to use feedback rather than merely receive it.

Separate confidence from capability

Confidence affects willingness to participate, but it is not a direct measure of what someone understands or can do. Build in lower-pressure ways to demonstrate learning before drawing conclusions.

Do not correct everything at once

A first attempt can reveal several issues, but a second attempt needs a manageable focus. Decide whether the next version is primarily about message, organisation, interaction, accuracy or pronunciation. When feedback tries to address every issue, learners often see a marked-up product rather than a usable next move. A narrow focus can be more demanding because it asks for deliberate control.

Read evidence with appropriate caution

Research can sharpen a question, but it cannot select a lesson for a particular class. Studies differ in participants, duration, setting, outcome measures and the amount of teacher support involved. A review may show a pattern across many studies; it does not guarantee that the same result will appear with every learner, topic or platform. Treat an evidence source as a reason to test a design choice thoughtfully, not as a slogan.

The sources below have different jobs. Some describe learning or assessment principles; others report research syntheses or policy guidance. None should be used to claim more than it supports. In a local setting, learner work, attendance patterns, short reflections and teacher observation remain essential evidence alongside published research.

Separate evidence from interpretation

A useful editorial habit is to label claims carefully. “This review found” is different from “this will work for every learner.” “The framework describes” is different from “the framework requires.” Readers deserve to know when an article is presenting established evidence, an emerging research area, an expert recommendation or a practical teaching choice. That distinction also makes articles more useful when circumstances change.

Plan for access, confidence and prior knowledge

A task can be demanding without being opaque. Make the purpose, time limit and success criteria visible. Give learners a way to ask for clarification privately when possible. Offer a model or language bank as optional support, not as a substitute for thinking. For online work, check whether the task depends on a device, bandwidth, account, quiet space or prior digital confidence that not every learner has.

Differentiation is not simply making one version easier. It can mean preserving the communicative goal while varying preparation time, information load, collaboration or the form of response. A learner who records an answer after planning is still working on the same purpose as a learner who speaks live, but the evidence should be interpreted in context.

Before the task, consider what can be made visible: an example, a glossary, a planning frame, a countdown, a transcript, or a clear turn-taking rule. During the task, consider what can be optional: a camera, a spoken response, a shared document, or public reporting. These are design choices, not signs that expectations have disappeared.

Use professional judgement openly

Expert recommendations and common teaching practice are useful when their status is clear. A familiar activity may be a good choice because it fits a group, creates trust or saves time—not because a single study proves it is universally superior. Name the reason for the choice. This helps colleagues and learners understand what is being tested and makes later revision more intelligent.

A review protocol for the next cycle

Key takeaways

• Choose a decision before choosing an activity.
• Design a first and second attempt around the same meaningful performance.
• Use published research as a guide to questions, not a replacement for context.
• Collect learner evidence before making the next change.
• Keep access and learner agency inside the design, not at the end.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗
3. Qiao (2025), Artificial Intelligence for Language Learning: A Systematic Review ↗
4. Lin & Lin (2019), Mobile-assisted ESL/EFL vocabulary learning: a systematic review and meta-analysis ↗

Feedback with human judgement is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Keep feedback narrow enough to use

Choose a single priority and connect it to another attempt. Learners are more likely to act on a focused observation than on a catalogue of everything that could improve.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

Planning a technology-supported lesson is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Plan the return, not only the first use

A learning decision becomes more durable when it returns in a slightly changed context. Leave a note in the plan about where the same choice will appear again.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

Making edtech more accessible is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Remove barriers that do not serve the goal

Accessibility is often a matter of clarity, format and time. Ask whether a barrier is part of the intended learning challenge or simply an accidental obstacle.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

A realistic AI-supported task is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Make practice resemble the eventual use

Practice need not copy real life exactly, but it should preserve a meaningful decision: choosing a response, clarifying a point, noticing a pattern or adapting to an audience.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

Writing a safer AI prompt is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Make the prompt do useful work

A useful prompt names the audience, the action and the limit. It does not need to supply every sentence learners will use; it should make clear what they are trying to achieve.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

A quick check of AI output is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Collect evidence before giving advice

A short sample, a question, a draft or a rehearsal is usually more useful than guessing where the difficulty lies. Let the evidence determine the next piece of support.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

Pair work with a digital tool is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Give each person a reason to contribute

Pair work becomes purposeful when each learner has information, a role or a decision that the other person needs. A shared answer alone does not guarantee shared thinking.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

A reflection after using AI is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Turn reflection into a next move

Reflection helps when it ends with a specific choice for the next attempt. “I need to be better” is too vague; “I will prepare one example before I speak” can be tested.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

Choosing a useful learning tool is most useful when it solves a recognizable teaching or learning problem. Start with a real moment: learners hesitate before speaking, a task produces thin answers, or the same homework disappears into a folder. The point is not to add another routine. It is to make the next attempt easier to begin and more informative to review.

Start with one observable change

For educational technology and AI, choose one behaviour that a learner or teacher could notice within a week. “Improve confidence” is too broad to guide a decision. “Prepare two follow-up questions before a discussion” is observable. Keep the first version modest enough that people can do it without a long explanation or a new platform.

Use the routine in a real task

A routine becomes meaningful when it is attached to a genuine task. In an English lesson, that might mean planning language for a short recommendation, comparing two options, or recording a clearer second response. Show what a useful outcome looks like, but leave room for learners to make the language their own.

A useful test

If learners could complete exactly the same exercise without making a language, study or teaching decision, the routine probably needs more purpose. Add an audience, an information gap, a choice, or a short explanation of why the answer matters. This is not about making every activity elaborate. It is about giving attention somewhere useful to go.

For a learner working alone, the same principle applies. Replacing “review vocabulary” with “write three messages that need these phrases” gives practice a direction. For a teacher, replacing “check understanding” with “ask learners to choose and justify the better option” produces clearer evidence.

Begin with the smallest useful version

Keep the first use deliberately narrow. A routine that takes two minutes and produces one visible decision is easier to repeat, explain and improve than a large activity with several competing aims.

Check the cost as well as the benefit

Ask what the routine asks of attention, time and access. A productive choice for one group can become noise for another if the instructions are dense, the tool is unfamiliar, or the pace leaves no time to think. Adjust the support before deciding that learners are not engaged.

It is also worth checking the emotional cost. A public correction, an open microphone or a timed response can provide useful evidence for some learners and unnecessary pressure for others. Make the purpose explicit, offer preparation where it helps, and interpret a quiet response carefully rather than treating it as an absence of learning.

A small next step

• Name the performance you want to see.
• Run the routine once in a meaningful task.
• Collect one small piece of evidence.
• Decide what to keep, change or remove.

Review without overreacting

One lesson, conversation or study session rarely gives a complete answer. Look for repeated signals: the same instruction being misunderstood, the same language support being used, or the same point at which learners lose momentum. A short note after several attempts is more reliable than a strong impression from one unusually easy or difficult day.

If the routine is helping, make it simpler rather than larger. If it is not helping, identify whether the problem is the task, the support, the timing or the expectation. Removing an unnecessary step is a valid improvement. Good practice becomes sustainable when it respects the limited attention of teachers and learners.

Related reading

Sources

Sources

1. UNESCO (2023), Guidance for generative AI in education and research ↗
2. Kohnke, Zou & Zhang (2024), A systematic review of the first year of publications on ChatGPT and language education ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

A guide to useful learner evidence is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

Designing the next attempt is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

Making progress visible is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

How to use evidence without losing judgement is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

A stronger routine for everyday practice is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

What to keep, change and measure is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

Common pitfalls and better alternatives is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

A practical audit is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

From intention to learner action is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

EdTech and AI in English learning is most useful when it changes what a learner can do, not only what a lesson looks like. This guide offers a practical way to decide where attention belongs, what evidence to collect and how to plan a meaningful next attempt.

Define the decision before the activity

Start with one decision learners need to make. It might involve choosing language for an audience, understanding a detail, explaining an idea or revising a response. Naming the decision prevents a familiar resource or activity from becoming the purpose of the lesson.

Use a small cycle of evidence

Invite an initial attempt, observe what happens, give focused information and create a second attempt. The second attempt matters because it shows whether learners can act on the information rather than merely hear it.

The focus for this guide

The decisions that matter most is the specific lens here. Use it to decide what to notice first: the learner decision, the support provided, the response produced, or the next revision. A clear lens keeps reflection practical and prevents a useful routine from becoming a vague checklist.

Keep the support proportional

Too little support can make a task opaque; too much can complete the thinking for the learner. Use a model, a prompt, a word bank or a worked example only where it removes an irrelevant barrier. Reduce the support once the learner can make the central decision independently.

Check transfer, not only completion

A correct answer in one familiar exercise is encouraging but incomplete evidence. Change the topic, audience, medium or time pressure slightly and see what learners can still do. That small variation helps distinguish temporary rehearsal from flexible use.

A limitation worth keeping in view

No source can replace professional judgement about learners, context and access. Research identifies patterns and limitations; teachers and learners still need to decide what is appropriate for the next task. Treat a recommendation as a tested starting point, not a universal rule.

Key takeaways

• Name the learning decision before selecting the resource.
• Plan an observable first and second attempt.
• Use support that learners can gradually leave behind.
• Look for performance in a changed context.

Related reading

Sources

Sources

1. UNESCO, Guidance for generative AI in education and research ↗
2. Guo, Kim & Rubin (2014), Video production and learner engagement ↗

Assessment needs evidence of what the learner can do, not only what a system can generate. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Separate AI-supported drafting from independent performance. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Make the allowed support and evidence requirements explicit. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Banning or allowing AI is the only decision. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Policies, privacy and equity need local review. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Analytics can reveal patterns but cannot explain every learner decision. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Use data as a conversation starter, not a verdict. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Combine usage data with learner voice and performance evidence. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

More data means better teaching. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Privacy, consent and interpretation are essential. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Accessible design helps more learners participate from the outset. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Offer more than one route to instructions and response where appropriate. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Test the task with realistic devices and bandwidth. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Accessibility is a special accommodation added later. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Accessibility needs vary and should not lower the learning aim. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Online collaboration needs a shared purpose and visible responsibility. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Make each learner’s contribution necessary to the outcome. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Review how the group made decisions, not only the final product. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Putting learners in a breakout room creates collaboration. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Roles and platform familiarity shape participation. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Sustainable digital learning includes attention, boundaries and recovery. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Choose a realistic rhythm of focused work and offline practice. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Notice which format supports concentration and retention. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

More screen time means more learning. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Individual needs and access vary. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

AI suggestions can prompt noticing but need human evaluation. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Keep the learner’s original attempt and compare suggestions critically. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Use a teacher or peer check for important work. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

AI feedback is a final judgement. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Accuracy, bias and privacy require ongoing scrutiny. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

AI can offer prompts and simulated audiences for rehearsal. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Use it to prepare for human interaction rather than replace it. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Reflect on what the tool could and could not understand. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

A simulation is the same as real conversation. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Learners should know what data is shared. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Flashcards work best when they support retrieval and spaced return. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Include a meaningful cue and require an answer before revealing it. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Review difficult cards in a later communicative task. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Adding more cards is always progress. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Apps differ in scheduling and data practices. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Video tools can support modelling, replay and learner reflection. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Give viewing or recording a precise purpose. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Ask what changed between the two attempts. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Video automatically creates engagement. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Camera access and confidence affect participation. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗

Technology is useful when it improves a specific learning opportunity. is often discussed as if one technique will solve the problem for every learner and classroom. A more useful starting point is to identify the performance learners need, the conditions in which they need it, and the evidence that would show progress. This article turns that question into a sequence of practical decisions.

Start with the learning problem

Before choosing an activity, describe the problem in observable terms. Teachers and learners should be able to say what learners are trying to understand or do, what currently makes that difficult, and what a better attempt would look like. That prevents a familiar activity from becoming the goal in itself.

A strong working question is: “What will learners be able to do differently after this sequence?” It is more informative than “Have we covered the topic?” because it directs attention towards use, evidence and follow-up.

The central decision

Define the learning gain before choosing the tool. The choice should remain visible to learners: when they understand why an activity matters, they can allocate attention more purposefully and evaluate their own progress with more accuracy.

A sequence that makes practice count

First, give learners just enough orientation to begin. Next, let them attempt the work with an appropriate level of support. Then provide focused information about what happened and a second opportunity to use that information. This cycle is more valuable than adding activity after activity without a chance to revisit an important choice.

Review whether learners used the tool for the intended purpose. Keep the demand authentic enough that learners must make a decision, rather than simply recognise a previously supplied answer.

Prepare the conditions for success

Preparation matters because learners need enough language, time and clarity to engage with the intended challenge. Make success criteria visible, model only the move that is genuinely new, and let learners ask a question before the task begins. The aim is not to remove productive difficulty; it is to avoid spending attention on uncertainty that does not contribute to learning.

During the activity

Observe the choices learners make rather than only whether they finish. Listen for the language they reach for, notice which support they use, and identify one pattern worth returning to. Learners can do the same by pausing briefly to name what helped and what remained difficult.

After the first attempt

A first attempt supplies valuable information. Compare it with a model, a criterion or an earlier performance, then decide on one priority for improvement. A short second attempt is often more informative than a long explanation because it shows whether the learner can act on the information.

Build a return path

Return to the same underlying choice later in a changed context. This might mean shifting from a spoken conversation to a message, changing the audience, or asking learners to explain a decision. Reuse helps distinguish temporary familiarity from knowledge that can travel.

What evidence can and cannot tell you

A completed task, a score or a confident answer is useful evidence, but it is not the whole story. Look across several attempts, including a changed context where possible. This reduces the risk of mistaking rehearsal of one example for flexible learning.

A common misconception

Newer technology is automatically better teaching. The alternative is not to avoid the approach; it is to use it with a clear purpose, sensible limits and evidence from learners’ actual work.

Limitations and professional judgement

Access and privacy are part of the decision. Research can inform a decision, but it does not remove the need to consider age, proficiency, access, time, language background and learner goals. Adapt the route while keeping the intended learning visible.

What this means for Language Ticket learners and teachers

Language Ticket’s approach is to use evidence as a guide for purposeful teaching, not as a script. Teachers can observe a learner’s response, adjust the support and give feedback that leads into another meaningful attempt. Learners can use the same ideas to choose a specific next step between lessons.

Key takeaways

• Name the performance, not only the topic.
• Make the central learning decision visible.
• Plan a second attempt after feedback or reflection.
• Use evidence from several attempts before drawing a conclusion.

Related reading

Sources

Sources

1. U ↗
2. h ↗