AccessShare: Co-designing Data Access and Sharing with Blind People

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence Ethics
• Accessibility Research

Background:

• Visually impaired individuals are crucial for AI data collection but lack accessible methods for data inspection and control.
• Existing data contribution processes exclude blind individuals from essential oversight.
• There is a need for inclusive mechanisms in AI data stewardship.

Purpose of the Study:

• To address the inaccessibility of visual data inspection for blind AI data contributors.
• To explore the needs and preferences of blind individuals regarding data consent, inspection, and control.
• To co-design accessible interfaces for AI data stewardship.

Main Methods:

• Engaged 10 blind participants using smartglasses and an AI-infused application for home-based image data collection.
• Engineered AccessShare, a novel data access interface, as a design probe.
• Conducted a co-design study to gather insights on consent, data inspection, and control.

Main Results:

• Interactive informed consent significantly impacts data contribution.
• Accessible data inspection systems like AccessShare facilitate communication between data stewards and blind contributors.
• Participants provided valuable insights into preferences for data control and inspection.

Conclusions:

• Accessible data inspection and control are vital for inclusive AI data practices.
• Novel interfaces and interactive consent mechanisms can empower blind data contributors.
• Findings guide the development of responsible and equitable AI data stewardship.