Crowdsourcing Subjective Fashion Advice Using VizWiz: Challenges and Opportunities
Michele A. Burton, Erin Brady, Robin Brewer, Callie Neylan, Jeffrey P. Bigham, Amy Hurst · 2012 · ASSETS '12: Proceedings of the 14th International ACM SIGACCESS Conference on Computers and Accessibility · doi:10.1145/2384916.2384941
Summary
This paper investigates the accessibility barriers that fashion and clothing present to people with vision impairments, and explores whether crowdsourcing through VizWiz can provide useful subjective fashion advice. The research is grounded in the recognition that fashion is a primarily visual language — how we dress communicates identity and social belonging — yet this language is largely inaccessible to blind and low-vision people. The study comprises three components. First, a 10-day diary study with eight legally blind women (ages 21-73) explored how fashion is perceived and managed by people with vision impairments. Participants described challenges spanning both objective aspects (identifying colors, reading washing instructions) and subjective ones (whether items coordinate, whether an outfit is age-appropriate, whether clothes have stains or flaws). The desire to "fit in" was universal, with participants acutely aware that sighted people judge their appearance while they cannot reciprocally observe fashion norms. Second, an online survey of 22 adults (with and without vision impairments) confirmed that clothing decisions are collaborative for everyone — only 2 of 22 participants shop alone — and that people with vision impairments particularly want information about wardrobe coordination. Third, a two-week pilot study with seven blind participants (ages 27-59) tested the feasibility of using VizWiz with designated volunteer workers ("trusted strangers") to answer fashion questions via photographs.
Key findings
During the two-week VizWiz pilot, participants submitted 93 questions total, 77 of which were fashion-related. A clear progression emerged: early questions were primarily objective and validating ("What color is this shirt?" — testing whether the system worked), while by mid-first-week, questions became more complex and subjective ("Can I wear these two pieces together?" "Does this polo shirt go with these trousers?"). Questions fell into three categories: Visual/Objective (color identification, reading labels), Fashion/Subjective (matching, appropriateness, style advice), and Both. Participants built trust with the volunteer workers through a deliberate process of asking questions they already knew the answers to, then validating responses with sighted friends and family. Six of seven participants confirmed receiving external validation of the crowd's answers. Trust grew with use — one participant noted "What truly changed how I used the application was the confidence I had using it. I became much more confident with the answers the more I used it." Unexpected uses included asking about a companion's outfit (a grandfather checking if his grandson's shirt was appropriate for his birthday outing). Key challenges included photo quality (lighting affecting color identification, with volunteers sometimes giving conflicting color answers like "brown, tan, and red" vs. "brown, tan, and dark orange"), volunteer coordination and availability (response times ranged from 59 seconds to 30 minutes), and sensitive topics that participants were uncomfortable asking strangers about (undergarments, weight-related questions).
Relevance
This paper broadens the scope of accessibility research beyond functional task completion to encompass social participation and self-expression — areas often overlooked in assistive technology design. Fashion is not trivial; it affects employment, social acceptance, and self-confidence. The finding that blind participants were called "poor blind girl" when clothing flaws were visible illustrates how inaccessible fashion information can reinforce stigma and undermine the independence that assistive technology aims to support. For accessibility practitioners, the study demonstrates that crowdsourcing can work for subjective, opinion-based questions — not just the objective identification tasks VizWiz was originally designed for — but requires careful attention to trust-building, privacy, and cultural sensitivity. The concept of "trusted strangers" as a middle ground between anonymous crowd workers and personal contacts is a valuable design pattern for any accessibility service involving personal or sensitive information. The paper also highlights that color identification — seemingly one of the simplest visual tasks — remains surprisingly unreliable even for sighted people due to lighting, screen calibration, and genuine perceptual disagreement, a finding relevant to anyone designing color-dependent interfaces or accessibility tools.
Tags: blind users · crowdsourcing · fashion accessibility · VizWiz · subjective questions · social participation · assistive technology · trust