The generative Artificial Intelligence (genAI) innovation enables new potentials for end-users, affecting youth and the inexperienced. Nevertheless, as an innovative technology, genAI risks generating misinformation for end-users that is not recognizable as such. This results in increased trustworthiness of AI outputs. An assessing system for end-users is necessary to expose the unfounded reliance on erroneous responses. This paper identifies requirements for an as- sessing system to prevent end-users from overestimating trust in generated texts. Thus we conducted requirements engineering based on a literature review and two international surveys. With the surveys, we confirmed the requirements which enable human protection, human support, and content veracity in dealing with genAI. High detected trust is rooted in miscalibration; clarity about genAI and its provider is essential to solving this phenomenon, and we detected a demand for human verifications. Consequently, we provide evidence for the significance of future IS research on human-centered genAI trust solutions.

Tomitza, Christoph; Schaschek, Myriam; Straub, Lisa; Winkelmann, Axel: What is the Minimum to Trust AI? - A Requirement Analysis for (Generative) AI-based Texts In: 18th International Conference on Wirtschaftsinformatik (2023), bl under consideration for publication