Improving viewing of webpages for users with assistive needs: insights into offline testing

In this paper, we discuss an application of computer vision for designing systems which can improve the online experience of users with assistive needs; we also delve further into the case of older adult users. We discuss some lessons learned of use for researchers aiming to validate their AI-driven algorithms for web improvements to be experienced by users. Our research is on depicting web pages as visual images, in order to facilitate decluttering and zooming of these pages for assistive needs users. In particular, we elaborate on the ability of our image segmentation model to support smart magnification of content for older adults, as well as a study we designed to examine the effectiveness of these improvements with real participants. We also discuss the possible promise of simulating user behaviour, in what we refer to as offline testing, as an important step in solidifying decisions about interfaces to use with end-users. In particular, this method can serve as a useful stand-in when the appropriate user base is elusive.