Using machine learning to choose the right font for everyone

Why Read This from an HCI Perspective

This article is especially meaningful for HCI/UX practitioners because it reframes the digital reading experience not as ‘content,’ but as a ‘personalized perceptual environment.’ Examples such as font personalization, the readability research consortium, and Liquid Mode show how interfaces should adapt based on users’ visual characteristics, reading proficiency, and context. It also invites us to think together about accessibility, learning-experience design, and the design principles behind AI-driven personalization.

CIT's Commentary

From a CIT perspective, the core of this article is not ‘making the screen prettier,’ but the potential of adaptive interfaces that systematically reduce reading burden. In particular, adjusting fonts individually is not merely aesthetic optimization; it can be seen as an HCI intervention that jointly tunes eye-movement patterns, character recognition, and cognitive load. However, for personalization to be effective, preferences and performance must be measured separately, and it’s important to balance transparency—where users can understand and control what’s happening—with the convenience of automatic optimization. In educational settings, it’s also necessary to validate not only short-term improvements in readability, but long-term reading ability and learning transfer. CIT interprets these kinds of cases as a starting point for ‘personalized accessibility.’

Questions to Consider While Reading

• Q.Which user groups would benefit most from font personalization in practice, and would the differences show up more strongly in comprehension than in reading speed?
• Q.Between automatic personalization and giving users direct control over settings, which is more likely to drive higher trust and continued use?
• Q.How should we evaluate the impact of optimizing the reading experience—not just on short-term readability, but also on long-term learning outcomes and reading habits?