School of Information Sciences

Aziz Zeidieh's Preliminary Examination

PhD candidate Aziz Zeidieh will present his dissertation proposal, "Towards an O&M-Grounded AI Navigation Agent: Teaching Blind and Low-Vision Travelers Everyday Routes and Entrances Across Virtual and Physical Environments". His preliminary examination committee includes Chair & Director of Research Assistant Professor JooYoung Seo, Professor Christopher Lueg, Professor Michael Twidale, Professor Yang Wang, and Professor Nicholas A. Giudice.

Abstract

Navigation apps guide blind and low-vision (BLV) travelers turn by turn, and camera-based AI describes their surroundings, but neither is designed to teach routes, leaving travelers dependent on assistance from an orientation and mobility (O&M) specialist, family member, friend, or a remote sighted assistant like those from Aira or Be My Eyes. In this proposed dissertation work, I present a plan to examine whether an AI navigation agent grounded in O&M practice can teach BLV travelers new everyday routes and their entrances. I will pursue this aim in three studies. First, through surveys and focus groups, I will document the practices O&M specialists recommend for teaching their clients such routes, how BLV travelers currently learn them, and both groups' privacy concerns about camera-based AI wearables. These findings will shape the design of an agent for teaching BLV travelers a new route and its terminal point (i.e., a building entrance). Second, with a panel of BLV travelers and O&M specialists, I will lead the co-design of an accessible virtual environment of a real route and identify which features support learning it. Participants will learn the route from the agent, practicing it across two round trips with decreasing support, complete one round trip without guidance in the virtual environment, and then walk the real route, testing the agent's effectiveness and the extent to which virtual route knowledge transfers to the physical environment. Third, in parallel with the virtual-world testing, I will use camera glasses in the real world to identify which agent features translate from the virtual environment and to compare the agent with a general-purpose agent on the same model in a within-subjects study. I will also examine travelers' concerns about their own and bystanders' privacy during real-world use. This work will yield an O&M-grounded agent, evidence on the extent of virtual-to-real transfer, and an account of wearable privacy in practice.

Questions? Contact Aziz Zeidieh.

School of Information Sciences

501 E. Daniel St.

MC-493

Champaign, IL

61820-6211

Voice: (217) 333-3280

Email: ischool@illinois.edu

Back to top