"We are at the mercy of others' opinion": Supporting Blind People in Recreational Window Shopping with AI-infused Technology
Rie Kamikubo, Hernisa Kacorri, Chieko Asakawa · 2024 · Proceedings of the 21st International Web for All Conference (W4A '24) · doi:10.1145/3677846.3677860
Summary
This paper investigates how AI-powered navigation technology can support blind people in recreational window shopping — an activity that sighted people take for granted but remains largely inaccessible to those without vision. Researchers from the University of Maryland and Carnegie Mellon University conducted a formative study with 18 blind participants across two phases: in-person focus groups with 8 participants (ages 39-78, using white canes or guide dogs) and remote one-on-one interviews with 10 additional participants (ages 29-72). The study was grounded in a simulated scenario of walking through a shopping mall with either a guide robot (Cabot) or a turn-by-turn navigation app (NavCog). Participants discussed their current strategies, challenges, and information needs for recreational shopping. The research frames window shopping not as a utilitarian activity but as an important form of leisure and social participation, noting that the UN Convention on the Rights of Persons with Disabilities affirms rights to recreation. A key finding was the deep dependency on sighted companions — participants described relying on friends, family, or store staff to access basic information about their surroundings, with one participant capturing the central tension: "We are at the mercy of others' opinion." The COVID-19 pandemic originally disrupted the planned in-person mall study, leading to the remote interview format.
Key findings
Participants expressed distinct preferences for push notifications (automatic, proximity-triggered) versus pull notifications (user-initiated, on-demand). For shops, all wanted automatic push notifications of shop names as they walked past, with pull access to deeper layers: shop categories, target audience, window displays, sales, and new arrivals. Promotional content like sales and grand openings was universally desired as push notifications. A critical finding was that obstacle and POI notification preferences varied significantly based on mobility aid: white cane users wanted more environmental notifications (columns, planters) as orientation cues, while guide dog users preferred fewer obstacle alerts since their dogs handle avoidance. Pedestrian notifications were context-dependent — most participants only wanted them near crowds, children, or wheelchair users. Participants identified a layered information architecture: a first layer of shop names and categories delivered automatically, with a second layer of detailed information (prices, styles, ratings) available on demand. Interaction preferences favoured voice commands and physical buttons over touchscreens, as participants noted their hands might be occupied with canes, guide dog harnesses, or shopping bags.
Relevance
This research has significant implications for anyone designing navigation or wayfinding technology for blind users. It demonstrates that current assistive navigation systems focus too narrowly on getting from A to B, neglecting the recreational and exploratory aspects of navigating public spaces. The layered information architecture — automatic basic info with pull access to details — provides a practical design pattern for managing cognitive load in audio-based interfaces. The finding that notification preferences differ by mobility aid type is particularly valuable: a one-size-fits-all approach to obstacle alerts will frustrate users. The paper also highlights an underexplored social dimension — blind people face "sighted people interference" and social costs when shopping, suggesting that technology design must account for social acceptability alongside functional utility. The authors note the potential for generative AI to automate the labour-intensive process of collecting and annotating POI data for shopping environments.
Tags: blindness · navigation · guide robots · wayfinding · recreational accessibility · AI assistive technology · social inclusion
Standards referenced: UN Convention on the Rights of Persons with Disabilities