Publication of new research findings on trust in AI-based coaching systems in the journal Behavior & Information Technology.

The research paper "Do human trust models transfer to AI? A necessary condition analysis of interpersonal trust antecedents in large language model-based coaching" by Sophia Meywirth, Andreas Janson and Matthias Söllner was published in the renowned journal Behavior & Information Technology. With a CiteScore of 5.9, the journal is one of the leading international publications in the field of human-centered information systems, user experience and the social impact of digital technologies and is listed in the Q1 ranking of scientific journals.

With the increasing spread of anthropomorphically designed coaching systems based on Large Language Models (LLMs) for health-related behavior change, the question of whether the trust mechanisms between humans can also be transferred to AI systems is becoming increasingly important. The study therefore investigates whether and to what extent the factors known from interpersonal trust research - ability, benevolence and integrity - influence trust and the intention to use LLM-based fitness and nutrition coaching.

The results of the necessary condition analysis show that the perceived integrity of the system is a central prerequisite for both the perceived usefulness and the intention to use LLM-based coaching systems. Furthermore, the perceived capability of the system significantly influences the assessment of its usefulness. However, contrary to established models of interpersonal trust, no necessary or significant influence of benevolence was found. The results thus partially question existing assumptions of trust research and make it clear that not all mechanisms of interpersonal trust can be directly transferred to AI systems.
The study provides important implications for the design of trustworthy anthropomorphic AI systems. It shows how human trust mechanisms can be used in a targeted manner without losing sight of the special features of human-AI interaction. The research work thus makes an important contribution to the development of effective and trustworthy AI-based coaching and assistance systems.

The publication underlines the international visibility of the WISE team's research at the interface of artificial intelligence, human-computer interaction, user behavior and the design of digital systems.