AI-supported learner support can widen access, improve practice, and help some learners participate more fully in learning and assessment. The central assessment question is whether the support removes a barrier or changes what the learner is being asked to demonstrate. The closer AI gets to the final evidence, the more important disclosure, comparability, and moderation become. The current source set supports the inclusion case in principle, but stronger evidence is still needed on outcomes, subgroup effects, privacy, and fairness.

AI in learner support covers tools used, or proposed, for formative assessment, accessibility, SEND support, staff development, dialogue-based practice, personalised learning, simulation, virtual classrooms, cognitive apprenticeship, and early identification of learners who may be struggling. The underlying assessment issue is whether AI support improves participation without blurring what the learner can do independently. That boundary matters for fairness, comparability, and validity.

Support tools can widen access and reduce barriers, but they can also create unequal advantage, hidden dependence, or cognitive offloading if they are not designed and governed carefully. For assessment teams, the practical question is whether a tool is part of reasonable adjustment, part of teaching, or part of the assessed construct itself. That distinction changes what can be accepted, what must be controlled, and what evidence is needed.

- **Access support**: removes barriers so the learner can show the intended capability. - **Teaching and rehearsal support**: helps the learner develop before assessment. - **Performance support**: affects the evidence collected during assessment. - **Reasonable adjustment**: may support access, but needs clear rules where assessment evidence must remain comparable. - **Cognitive offloading**: where the tool starts to do work the learner should arguably be demonstrating independently.

There is broad convergence that AI can help learners practise, receive feedback, rehearse interaction, access personalised pathways, or get support that would otherwise be unavailable. The stronger source set also points to a simple assessment principle: the closer AI support gets to the final evidence, the clearer the rules, disclosure, and moderation need to be. AI support may be useful in learning and still inappropriate inside the assessed task if it changes what is being measured.

The evidence is much weaker on outcomes and comparability than on the general promise of support. The open question is not whether AI can assist, but when that assistance remains legitimate inclusion and when it becomes a change to the construct or standard being assessed. Vendor material and practitioner commentary often frame these tools as personalised, adaptive, or inclusive, but that is mainly a market signal unless independently validated. Stronger evidence is still needed on learner outcomes, subgroup effects, privacy, and whether support changes comparability between learners.

- Unequal advantage if some learners have access to better AI support than others. - Hidden dependence if learners appear capable only when the tool is present. - Reduced validity if performance support becomes part of the evidence without being recognised. - Privacy concerns for learners, especially where tools collect sensitive data. - Weak procurement decisions if inclusion claims are treated as proof rather than hypotheses to test.

1. Define the learner barrier the AI support is meant to reduce. 2. Decide whether the tool belongs to learning, reasonable adjustment, or assessed performance. 3. Check whether two learners with different levels of AI support would still be comparable. 4. Test what data the tool collects, especially where vulnerable learners are involved. 5. Put in place a way for staff to tell whether AI support is improving participation or masking dependence. The most useful evidence will show when AI support reduces barriers without replacing the learner judgement or capability the assessment is meant to evidence. Distinguishing support during learning from support during assessment remains the practical control point.

| Option | What it means | Main advantage | Main risk | |---|---|---|---| | Prohibit AI support | No AI in the learning or assessment pathway | Clear comparability | Can exclude learners who could benefit from support | | Permit AI in learning only | AI may help practice, rehearsal, and feedback, but not the final evidence | Supports inclusion without changing the assessed construct | Boundaries can be hard to explain and enforce | | Permit AI with disclosure in assessment | AI use is allowed if declared and governed | More flexible; may reflect real-world practice | Comparability and moderation become more complex |

A learner uses an AI-based prompt helper during revision to practise structuring answers and checking terminology. That use may improve participation without affecting the assessed standard. If the same tool is allowed to generate phrasing during the live assessment, the team now needs to decide whether the task is still measuring the learner’s own capability or a mixed human-plus-tool performance.

- Next Generation Assessment Conference 2026 panel on AI support for all learners. - Inside Higher Ed opinion piece on AI ethics and human effort. - Ametros Learning vendor source. - ChatEDU discussion of AI social simulation. - Pooks.ai source on personalised learning. - Pooks.ai source on personalised learning. - Pooks.ai audio description on personalised learning and engagement. - Pooks.ai audio description on visual study mind maps and personalised learning. - Pooks.ai audio description on education outcomes beyond tech and testing.

The vendor footprint is centred on personalised learning, engagement, simulation, and support tooling. These products are best read as market signals about how the sector is positioning AI for inclusion, not as validation that learning outcomes or assessment standards are protected.

### What is AI learner support in assessment? It is the use of AI to help learners access, practise, rehearse, or complete learning-related activities, with the key assessment question being whether that help changes what the learner is being asked to demonstrate. ### Why does it matter for inclusion? Because the same tool can reduce barriers for one learner and create unfair advantage or hidden dependence for another if it is not governed carefully. ### Can AI support be used safely in assessment? Sometimes, but only when the assessment rules, disclosure, and moderation make clear what support is allowed and what evidence is still expected to be independent. ### What evidence is still missing? Better evidence on learning outcomes, subgroup effects, privacy, cognitive offloading, and comparability would help assessment teams decide when AI support is a legitimate inclusion tool and when it changes the assessment standard.

Tim Burnett (Admin)

Test Community Network. "Learner support and inclusion." TCN AI & Assessment Wiki. Last reviewed 2026-04-24. https://www.testcommunity.network/wiki/learner-support-and-inclusion.html

- Next Generation Assessment Conference 2026 panel on AI support for all learners. - ChatEDU discussion of AI social simulation. - Inside Higher Ed opinion piece on AI ethics and human effort. - Ametros Learning vendor source. - Pooks.ai source on personalised learning. - Pooks.ai source on personalised learning. - Pooks.ai audio description on personalised learning and engagement. - Pooks.ai audio description on visual study mind maps and personalised learning. - Pooks.ai audio description on education outcomes beyond tech and testing.