Using a participatory activities toolkit to elicit privacy expectations of adaptive assistive technologies
Foad Hamidi, Kellie Poneres, Aaron Massey, Amy Hurst · 2020 · Proceedings of the 17th International Web for All Conference (W4A) · doi:10.1145/3371300.3383336
Summary
This paper addresses a critical but often overlooked tension in assistive technology design: the privacy tradeoff inherent in adaptive assistive technologies (AATs) that must collect and analyze user performance data to function effectively. The authors developed a participatory activities toolkit — a set of tangible, low-fidelity physical materials including scenario cards, data type cards, third-party cards, an expectations chart, a Wheel of Emotions, and privacy standard strips — designed to help non-technical users articulate their privacy preferences and expectations about AATs. The toolkit was used in semi-structured interviews averaging 58 minutes with six older adults (ages 64-87) who have Essential Tremors, a progressive neurological condition causing hand tremors that affects approximately 2% of the U.S. population. All participants experienced pointing and typing difficulties when using computers. Two AAT prototypes were presented: SuperSpeller, a typing-assistance tool modeled on Grammarly that could detect ability-related errors, and PINATA, a pointing-assistance tool with a dynamic bubble cursor that adapts to the user's tremor severity. The four structured activities walked participants through considering what data AATs should collect, who should access it, how they would feel if expectations were violated, and what privacy standards should apply.
Key findings
Participants had nuanced, sometimes conflicting views about AAT data collection. They expected each AAT to collect only data directly relevant to its function — typing data for SuperSpeller, pointing data for PINATA — and were surprised when prompted to consider broader data collection. A pivotal finding was that initially only one of six participants recognized that AAT performance data could reveal health conditions, but after reflection during the activities, all participants made this connection. This shift triggered significant concern. Participants were most comfortable sharing data with the AAT developer and medical professionals, but strongly opposed access by employers (fear of job discrimination), insurance companies (fear of rate increases or incorrect diagnoses), advertisers, and government agencies. Emotional responses were intense: participants expressed anger, disgust, fear, and feelings of violation when imagining unauthorized access, particularly by government or insurance companies. Participants were more protective of typing data than pointing data, viewing it as more personal. Four of six participants wanted GDPR applied to AATs, valuing its emphasis on user agency. The tangible toolkit itself was praised as "thought-provoking" and "eye-opening," with participants creatively using the physical cards — placing them on boundary lines, adding new cards, and physically smacking cards into the "do not access" column.
Relevance
This research raises urgent questions for the accessibility technology field. As assistive technologies become more adaptive and data-driven — incorporating machine learning, behavioral modeling, and cloud processing — the privacy implications grow more serious. Performance data from AATs can inadvertently reveal sensitive health information: typing patterns may indicate neurological conditions, pointing data may expose tremor severity, and this information could be exploited by employers, insurers, or advertisers. For practitioners building adaptive accessibility features, the study provides clear guidance: be transparent about what data is collected, limit collection to what is functionally necessary, give users granular control over sharing, and recognize that ability data is effectively health data deserving strong protections. The participatory toolkit methodology is also valuable — it demonstrates that tangible, low-tech elicitation materials can surface privacy concerns that traditional interviews miss, making it a useful approach for inclusive design research with older adults and people with disabilities.
Tags: privacy · assistive technology · adaptive systems · participatory design · essential tremor · older adults · data ethics
Standards referenced: HIPAA · GDPR