Adapting Learning Materials with AI

AI can draft a clearer explanation, a description of a diagram, captions, or a translation. The educator must check that the adaptation preserves the meaning and the learning students are expected to do. This guide extends the production workflow in Creating course materials.

CAST’s Universal Design for Learning guidelines address multiple means of engagement, representation, and action and expression, including language and access to assistive technology.¹ They provide a design framework, not evidence that an AI-generated adaptation is accessible or educationally effective.

Define what must stay the same

State the learning outcome, disciplinary terms, relationships, uncertainty, numbers, and task demands that the adaptation must retain. Decide whether a difficult feature is part of the outcome or an avoidable access barrier. Shortening a problem can remove the reasoning it was intended to elicit.

Worked adaptation: preserve a statistical distinction

This short source is educator-authored for practice:

In this observational sample, participants who attended the optional workshop had higher average scores. Because participation was self-selected, the comparison does not establish that attendance caused the difference.

Flawed AI-style simplification: “The workshop helped students get higher scores.” It removes both the observational design and the limitation on a causal conclusion.

Reviewed adaptation: “People who chose to attend the workshop had higher average scores in this sample. They were not assigned to attend. The comparison therefore does not show that the workshop caused the higher scores.”

Ask learners to identify the relationship and its limitation in both the source and the adaptation. If interpreting observational evidence is the outcome, do not supply the conclusion as an answer to an assessment question. This example shows how to explain a distinction, not how to remove it from the task.

Worked description: a chart without an invented trend

Suppose a teaching chart contains two labelled bars: Group A, 40%; Group B, 60%. Its vertical axis runs from 0% to 100%. No group sizes, time points, or uncertainty intervals are supplied. Provide the data table as an alternative representation.

Flawed description: “Performance improved by 20% after the intervention.” The chart supplies neither a before/after comparison nor an intervention.

Reviewed description: “Bar chart showing 40% for Group A and 60% for Group B on a 0–100% axis. B is 20 percentage points higher. Group sizes and uncertainty are not given.”

Match the description to the activity. If the task assesses whether students can calculate the difference, supply the values and structure without giving the calculation away. Discuss an equivalent accessible representation with the accessibility team. A fuller description can accompany an explanatory resource.

Review captions and translations

Do not distribute a generated transcript without checking it against the recording. For example, a missing “not” in “this result is not significant” reverses the claim. Check spoken symbols, subscripts, and units against the original visual or notes. Describe meaningful non-speech information and give students a way to report errors.

For translation, maintain a short glossary of approved terms and ask the model to flag ambiguous phrases rather than silently selecting a meaning. A second AI translation back to the original language may expose issues, but is not independent validation. If a competent reviewer is unavailable, use approved language support or clearly limit the material’s use; do not make an unverified translation the sole route to essential information.

Copyable adaptation prompt

Draft an alternative version of this permitted teaching resource.
Audience and purpose: [context without personal student data]
Barrier to address: [language, visual access, format, or another barrier]
Learning outcome and challenge to preserve: [outcome]
Terms, numbers, qualifications, and relationships to retain: [list]
Source resource: [permitted material]

Return the adapted version and a comparison table locating each substantive
change in the original. Flag ambiguity, lost information, and anything that
requires specialist review. Do not add facts or certify accessibility. Do not
answer a question that students are meant to solve themselves.

Verify the final resource

Review meaning against the source, then inspect the final document, learning platform, slide deck, or media player. Text accuracy alone does not establish correct reading order, keyboard access, caption timing, or screen-reader support. Use institutional accessibility checks and involve relevant users without requiring disclosure of personal circumstances.

Retain the editable original, a change record, and the reviewed adaptation. Provide a correction route. Use the pilot record to check whether learners can access the material and still practise the intended capability. For multilingual writing support and its assessment boundaries, see AI-assisted writing.

References & Footnotes