Trust in Artificial Agents
As artificial intelligence (AI) systems become increasingly embedded in everyday life, effective human–AI interaction critically depends on appropriately calibrated trust—knowing when reliance on AI is warranted and when it is not.
Trust in artificial agents is multidimensional, encompassing not only perceptions of performance (e.g., reliability, competence) but also moral characteristics (e.g., transparency, benevolence). To capture this complexity, Ullman and Malle (2019) developed the Multidimensional Measure of Trust (MDMT), which distinguishes between performance-based and moral trust components. In a preregistered online study with English-speaking participants (N = 692), we found that individuals calibrate their trust in a fine-grained manner for different types of agents (robots, virtual AI agents, humans), providing empirical support for the multidimensional structure of trust.
Building on this work, we translated and validated a German version of the MDMT for use in the D-A-CH region. This involves psychometric evaluation using existing (N = 1305) and newly collected data, including tests of reliability, factorial structure, and convergent and discriminant validity. Beyond validation, we will examine whether trust calibration processes replicate in German-speaking samples. Taken together, the project aims to advance research on human–AI interaction and enables systematic investigation of trust calibration in culturally diverse contexts.