Tailoring AI-Driven Reading Scaffolds to the Distinct Needs of Neurodiverse Learners

Why Read This from an HCI Perspective

This article provides empirical evidence that reading support does not always help—sometimes it can also add burden. For HCI/UX practitioners and researchers, it suggests that what matters may be the ‘combination’ and ‘adjustability’ of support elements more than the sheer amount of support, especially when considering both attentional resource demands and working-memory load. It also indicates that, in special-education contexts, experience assessments and performance outcomes may diverge, prompting a re-examination of accessibility design and evaluation frameworks.

CIT's Commentary

An interesting finding is that structural and semantic supports do not necessarily work in the same direction. Sentence segmentation may reduce burden, but the study also shows that some learners experience increased coordination cost when additional visual symbols or keyword labels are introduced. This challenges the intuition that ‘more support always leads to better comprehension.’ In particular, in supervised inclusive reading settings, the key is designing support that can be dynamically adjusted to each learner’s cognitive profile and the task context. Extending this to human–AI co-regulation implies that AI should act not merely as an information augmenter, but as an adaptive regulator that decides when and what kind of support to provide.

Questions to Consider While Reading

• Q.What adaptive interface would be appropriate for adjusting the effectiveness of support in real time to match individual differences and task complexity?
• Q.Among combinations of visual symbols, keyword labels, and sentence segmentation, which pairing creates the greatest coordination cost for which learner characteristics?
• Q.When applying human–AI co-regulation to reading support, how should the timing and intensity of AI intervention be designed?