B1462
Title: Trust in automated systems as a multidimensional psychological construct
Authors: Magdalena Wischnewski - TU Dortmund University (Germany)
Marie Beisemann - TU Dortmund University (Germany)
Philipp Doebler - TU Dortmund University (Germany) [presenting]
Nicole Kraemer - University of Duisburg-Essen (Germany)
Abstract: Trust in a machine-learning-based system is typically justified only to a certain degree. Ideally, trust is calibrated in the sense that a human interacting with a system neither over- nor undertrusts the system. To relate objective reliability measures like classification accuracy, fairness or robustness measures to perceived trust, the latter needs to be quantified. However, trust consists of several related facets and is, hence, multi-dimensional. A theoretically well-founded questionnaire is presented that includes 30 five-point Likert scale items for six dimensions of trust: global trust, integrity, unbiasedness, perceived performance, vigilance and transparency. A large English language sample $(n = 883)$ was used to derive the final TrustSix scale from a larger initial item pool. Perceived trust in three vignettes (fictional automated systems) is measured, e.g., a system for skin cancer detection. Special emphasis has been placed on exploring the exact factorial structure of the latent variables and checking their stability across vignettes. A global trust factor could be discovered with the help of a bifactor rotation, with five additional factors for the more specific trust dimensions. The reliability of each 5-item subscale is satisfactory ($\alpha = .76-.96$), with satisfactory overall reliability for the main factor ($\omega_H = .75-.80$, $\omega_T = .97-.98$). Correlations with adjacent constructs indicate sufficient discriminant validity.
