Registered Reports emphasize the importance of the research question and the quality of methodology by conducting peer review prior to data collection. High quality protocols are then provisionally accepted for publication if the authors follow through with the registered methodology. Registered Reports are currently only accepted in the Department Human-Centered Information Systems (former Human Computer Interaction and Social Computing department). Further information on Registered reports can be found here.
The following page documents all accepted Registered Reports within BISE.
Beneficial Mistrust in Generative AI: The Role of AI Literacy in Handling Bad Advice
Authors: Dirk Leffrang (Padeborn University), Nina Passlack (University of Bamberg), Oliver Müller (Paderborn University), Oliver Posegga (University of Bamberg)
Acceptance date: 16 Jan 2025
Accepted by: Alexander Maedche (Department Editor)
Abstract: Despite the increasing proliferation of Generative Artificial Intelligence (GenAI),
systems like large language models (LLMs) can sometimes present misleading or false
information as true – a problem known as “hallucinations.” As GenAI systems become
more widespread and accessible to the general public, understanding how AI literacy
influences advice-taking from imperfect GenAI advice is crucial. Drawing on the
correspondence bias, we study how individuals with varying AI literacy levels react to
GenAI providing bad advice. Gathering empirical evidence through an online
programming experiment, we expect AI-literate individuals to take less advice while
receiving an error but relatively more advice in future tasks. We outline how
correspondence bias can explain these variations, reconciling mixed findings of prior
studies on AI literacy. Our research thus contributes a holistic perspective on the
beneficial mistrust through AI literacy to education, integration, and evaluation
programs of AI, highlighting the dangers of naive evaluation strategies