Imagining Artificial Intelligence Applications with People with Visual Disabilities using Tactile Ideation
Cecily Morrison, Edward Cutrell, Anupama Dhareshwar, Kevin Doherty, Anja Thieme, Alex Taylor · 2017 · Proceedings of the 19th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '17) · doi:10.1145/3132525.3132530
Summary
This paper presents a novel tactile ideation methodology for working with people with visual disabilities to imagine future AI applications, and reports on cross-cultural workshops conducted in the UK (6 participants) and India (8 participants). Rather than asking users about current accessibility challenges, the researchers wanted to explore what people with visual disabilities envision as their own technological future. Traditional ideation methods (ideation cards, post-it notes, visual sketching) are inherently visual, so the team developed tactile alternatives drawing on show-and-tell, object brainstorming, and Lego Serious Play principles. The workshops used two activity sets: Activity Set 1 was technology-agnostic, asking participants to imagine a "sixth sense" or superpower they would like, using physical objects as prompts and clay to create accessories representing their ideas. Talking buttons with place names (home, market, temple, work) helped participants situate their ideas in specific contexts. Activity Set 2 was more directed, providing participants with AI and output "widgets" (person recogniser, object recogniser, room mapper; speech, vibration, 3D audio, tactile display) recorded on talking buttons, which they combined with craft materials to design specific applications. The India workshop focused entirely on Activity Set 1 due to resistance observed in the UK workshop toward clay prototyping. Data was synthesised through iterative clustering by the research team.
Key findings
The workshops generated 66 unique ideas across both sites. The most striking finding was that the overwhelming majority of ideas focused on social interaction rather than functional tasks like navigation or object recognition. Forty of 62 ideas related to visual disability mentioned people in some way. Ideas clustered into four categories: identifying people (recognising friends at a temple, knowing who is approaching at work, navigating networking events), managing social interactions (reading facial expressions and emotional cues, establishing private communication channels like eye contact, knowing if someone is paying attention), social stories (participating in visual-reference conversations, recognising coins at a temple without embarrassment, not being cheated by vendors), and social independence (reducing reliance on helpers who may take advantage, having technology for mundane tasks so that social interactions are not driven by need). Remarkably, these themes were consistent across the UK and India despite substantial cultural and infrastructural differences. The researchers distilled four social dimensions for AI assistive technology design: Social Participation (ability to participate in a given social context), Social Navigation (identifying and entering opportunities for social interaction), Social Maintenance (managing interaction through knowledge of others' social cues), and Social Independence (being free from the constraints of social interaction through independent abilities).
Relevance
This paper challenges the dominant framing of AI for visual disabilities as primarily about functional access (reading text, identifying objects, navigating spaces). The finding that users across two very different cultures consistently prioritised social interaction over functional tasks is a powerful message for accessibility practitioners and AI developers: the social dimensions of lived experience with visual disability deserve equal attention in technology design. The four social dimensions framework (participation, navigation, maintenance, independence) provides a practical design lens for evaluating whether new AI tools address the full range of users' needs. The tactile ideation methodology itself is a significant contribution, offering concrete techniques for inclusive co-design workshops with people who cannot use visual ideation tools — physical objects as conversation prompts, talking buttons instead of written cards, and clay as a non-visual prototyping medium (though this worked better for totally blind participants than those with partial sight). Limitations include the small and self-selected samples, the difficulty of prototyping without vision, and the inherent influence of workshop structure on the ideas generated.
Tags: artificial intelligence · blindness · visual impairment · participatory design · design methodology · computer vision · social accessibility · cross-cultural research