How GenAI Replaces or Supplements Human/VAT Support for BLV Privacy

Determine how generative AI tools used by blind and low vision individuals supplement or replace tasks that have traditionally been supported by human assistants or visual assistive technologies, and characterize how these users conceptualize and evaluate generative AI tools with respect to their visual privacy concerns.

Background

The paper investigates visual privacy management among blind and low vision (BLV) individuals in the context of rapidly proliferating generative AI (GenAI) tools, which are increasingly used for visual interpretation and related tasks. Prior work has focused largely on human-mediated visual assistance (e.g., volunteers, employees, crowdworkers) and subsequent automation via visual assistive technologies (VATs), but the privacy implications and comparative roles of GenAI versus traditional supports for BLV users have not been clearly established.

The authors explicitly note a gap in understanding both the functional substitution or augmentation that GenAI may provide relative to human assistance or VATs, and the ways BLV users conceptualize and evaluate GenAI tools in relation to privacy. This uncertainty motivates their empirical paper across multiple real-world scenarios to surface current practices and design preferences for privacy-aware GenAI.