Development Of A Scale For Measuring The Perceived Trustworthiness Of AI Systems

Several scales exist that try to measure trust in or trustworthiness of automated systems, computers, and robots (see Jian et al., 2000; Madsen & Gregor, 2000; Malle & Ullman, 2023; Wojton et al., 2020 for examples), some of which are also used to measure trust and/or trustworthiness in/of AI (Jensen et al., 2025; Perrig et al., 2023; Scharowski et al., 2025). However, there is no scale yet that has been created to explicitly measure the perceived trustworthiness of AI and is at the same time based on the bases of trust as suggested by Mayer and colleagues (1995), and Lee and See (2004). Therefore this research aims to fill this gap and create a measuring instrument that measures the perceived trustworthiness of AI-systems and simultaneously takes into account recent models of trust in AI (see Solberg et al., 2022; Hertel et al., 2025) that are based on the aforementioned frameworks by Mayer et al. (1995), and Lee and See (2004), respectively. In accordance with Solberg et al. (2022) and Lee and See (2004) the model by Hertel et al. (2025) suggests three facets of perceived trustworthiness of an AI-system: Performance, Purpose and Process. Furthermore, Hertel et al. (2025) suggest another construct, Transparency, as essential for the assessment of the trustworthiness of an AI system. As it is one of the goals of this research to align the measuring instrument with the model suggested by Hertel et al. (2025), items are generated via deductive item generation in accordance with those four factors: Performance, Purpose, Process and Transparency. The initial item pool will be tested on participants who are asked to assess the perceived trustworthiness of ChatGPT via the presented items. Afterwards an exploratory factor analysis will be performed. In accordance with the model by Hertel et al. (2025) it is anticipated that factor analysis will show a structure of factors corresponding to the three factors of perceived trustworthiness plus the additional construct Transparency. Also, taking factor loadings into account, the initial item pool will be reduced to form a measuring scale of economically reasonable extent. The ultimate goal of this research is to suggest an item set for a measuring scale that is ready to be confirmed and validated in follow up studies, which are not within the scope of this work.