Expertise and Information Seeking in the Age of Generative AI: New Procedures, New Problematics

As generative AI chatbots reroute procedures around information seeking, challenges increase with assessing and implementing relevant source material. Interactions between users and chatbots differ from previous paradigms of information seeking (i.e., search engines, library stacks), as these chatbots summarize information probabilistically through dialogic interactions, leveraging natural language-based signifiers of expertise and intelligence to persuade users of their utility and trustworthiness. This challenge could present significant risks for users who engage with these chatbots without clear expectations or expertise on the topic of their query. Our methodology in this paper comes from Ian Bogost's concept of procedural rhetoric, which examines how arguments are constructed (and persuasion takes place) through rule-based processes. We investigate how interactions relating to high stakes domains that value accuracy and precision, such as medicine and academic research, are subtly (and not-so-subtly) reshaping users' experiences and expectations surrounding information seeking. While seemingly authorless, AI-generated text nonetheless reflects human motives, values, and ideals that have been embedded here, as they are in all technological objects, systems, and structures. Accordingly, there is a need to understand how the procedures surrounding information seeking practices are informed by users' values, experiences, and expectations-and how this might in future be altered by the emerging turn towards AI "experts" and authority.