AccessShare: Co-designing Data Access and Sharing with Blind People (2407.19351v1)

Abstract: Blind people are often called to contribute image data to datasets for AI innovation with the hope for future accessibility and inclusion. Yet, the visual inspection of the contributed images is inaccessible. To this day, we lack mechanisms for data inspection and control that are accessible to the blind community. To address this gap, we engage 10 blind participants in a scenario where they wear smartglasses and collect image data using an AI-infused application in their homes. We also engineer a design probe, a novel data access interface called AccessShare, and conduct a co-design study to discuss participants' needs, preferences, and ideas on consent, data inspection, and control. Our findings reveal the impact of interactive informed consent and the complementary role of data inspection systems such as AccessShare in facilitating communication between data stewards and blind data contributors. We discuss how key insights can guide future informed consent and data control to promote inclusive and responsible data practices in AI.

• The paper introduces a novel interface that enables blind users to inspect and control image data using smart glasses and automated descriptors.
• It employs a participatory co-design methodology with ten blind participants in authentic data collection and decision-making sessions.
• Findings emphasize that refined descriptor generation is key for balancing privacy concerns with effective data sharing in AI applications.

Insights on AccessShare: Co-Designing Data Access and Sharing with Blind People

The paper "AccessShare: Co-Designing Data Access and Sharing with Blind People" addresses a fundamental challenge in the intersection of accessibility and AI: enabling blind individuals to access and control their data. The research highlights a critical gap in current practices, where blind contributors often lack the means to inspect the image data they provide, leading to limited agency and control in data stewardship processes. By engaging ten blind participants, the paper explores their needs and preferences regarding data consent, inspection, and sharing.

Overview and Methodology

The authors present a comprehensive approach designed to engage blind participants in the data collection and sharing process. Participants used smart glasses to capture image data in their homes and assessed their photos through a novel data access interface named AccessShare. This paper was structured into multiple sessions, including a system evaluation, semi-structured interviews, and a co-design session. The co-design phase was particularly critical, encouraging participants to reflect on their data-sharing decisions and ideate features for accessible data-sharing systems. The choice to engage participants in real-world data collection scenarios adds depth to the paper’s findings, capturing authentic user experiences and concerns.

Key Findings

One notable finding is the participants' general inclination to contribute their data for research purposes, juxtaposed against the genuine concerns for privacy and the presence of personally identifiable information (PII) in the images. Participants who decided to share their data highlighted the absence of privacy concerns as a primary factor, whereas those who refrained often pointed to the presence of unintended information in the backgrounds, such as visible faces in reflective surfaces or other sensitive content.

The AccessShare interface played a critical role in decision-making, offering participants a summary of their images based on automatically generated descriptors. While some descriptors, like 'face presence,' were directly beneficial for understanding privacy implications, others, such as vague object detection descriptors, were less actionable and occasionally caused confusion. The paper emphasizes the significance of a more nuanced approach to descriptor generation, allowing blind users to make informed decisions about their photo data.

Implications for Data Sharing and AI

This research is timely as it opens the discussion for inclusive data stewardship practices in AI, especially for underrepresented communities. Providing actionable descriptors and enabling better data inspection mechanisms can empower blind users, promoting ethical data use and responsible AI development. Furthermore, the paper sheds light on the importance of participatory frameworks that allow blind individuals to have a voice in data governance, aligning with broader trends in AI for accessible and ethical technology design.

Moving forward, the integration of advanced technologies, such as generative AI for enhanced image descriptions, could potentially bridge some of the current gaps. However, this requires careful consideration to balance the added utility against the risks of overreliance on AI-generated content, which may introduce its own biases and errors.

Future Directions

The paper concludes with several open questions and challenges, emphasizing the need for continued exploration of consent dynamics and the practical integration of accessible data sharing systems. Future research should consider scalable solutions for consent acquisition and ongoing data access, examining how technology can be designed to support flexible, user-centric data governance models.

Overall, AccessShare presents a thought-provoking contribution, highlighting the complex interplay of accessibility, privacy, and technology. It underscores a broader commitment to fostering inclusive environments where all individuals, regardless of disability, can actively participate in shaping AI and data-driven futures.