Aalto University Builds AI That Mimics 3 Human Reading Decisions for Adaptive Texts
Updated
Updated · Futura · Aug 23
Aalto University Builds AI That Mimics 3 Human Reading Decisions for Adaptive Texts
2 articles · Updated · Futura · Aug 23
Summary
Aalto University researchers developed an AI that models how people read, deciding what to fixate on, what to skip and when to backtrack so digital text can adapt to each reader.
Using reinforcement learning, the system optimizes for comprehension while minimizing time and mental effort, rather than simply copying eye-movement patterns from large datasets.
The model makes 3 layers of choices—word focus, sentence-level prioritization and whether to continue, skip or reread—and researchers said its behavior closely matched human readers.
Potential uses include legal documents, smart glasses and augmented-reality displays that adjust wording, size, pace or layout for drivers, dyslexic readers or second-language users.
That promise comes with privacy risks: making the system useful would require tracking sensitive signals about reading difficulty, habits and context, raising surveillance and ethics concerns.