Generative AI (GenAI) tools promise to advance non-visual information access but introduce new challenges due to output errors, hallucinations, biases, and constantly changing capabilities. Through interviews with 20 blind screen reader users who use various GenAI applications for diverse tasks, we show how they approached information access with everyday uncertainty, or a mindset of skepticism and criticality towards both AI- and human-mediated assistance as well as information itself. Instead of expecting information to be 'correct' and 'complete', participants extracted cues from error-prone information sources; treated all information as tentative; acknowledged and explored information subjectivity; and constantly adjusted their expectations and strategies considering the politics around access. The concept of everyday uncertainty situates GenAI tools among the interconnected assistive applications, humans, and sociomaterial conditions that both enable and hinder the ongoing production of access. We discuss the implications of everyday uncertainty for future design and research.
https://dl.acm.org/doi/10.1145/3706598.3713433
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)