Trust and Reliance in XAI -- Distinguishing Between Attitudinal and Behavioral Measures

Abstract

Trust is often cited as an essential criterion for the effective use and real-world deployment of AI. Researchers argue that AI should be more transparent to increase trust, making transparency one of the main goals of XAI. Nevertheless, empirical research on this topic is inconclusive regarding the effect of transparency on trust. An explanation for this ambiguity could be that trust is operationalized differently within XAI. In this position paper, we advocate for a clear distinction between behavioral (objective) measures of reliance and attitudinal (subjective) measures of trust. However, researchers sometimes appear to use behavioral measures when intending to capture trust, although attitudinal measures would be more appropriate. Based on past research, we emphasize that there are sound theoretical reasons to keep trust and reliance separate. Properly distinguishing these two concepts provides a more comprehensive understanding of how transparency affects trust and reliance, benefiting future XAI research.CCS Concepts: • Human-centered computing → HCI theory, concepts and models.

It seems smart, but it acts stupid. Development of trust in AI advice in a repeated legal decision-making task

“…Independent of the latter, these findings raise questions on the explanatory power of trust measures and the quality of trust outcomes: are individuals able to express trust adequately or is a behavioral trust measure stronger in predicting trust? Is trusting an AI as similar to relying on AI advice, cf., [6,61].…”

“…As a second measurement, we track self-reported trust. Subjective measures are used to capture inherently subjective trust data, reflecting an individual's perspective [61]. These psychological constructs are usually measured with survey scales.…”

Humans in XAI: increased reliance in decision-making under uncertainty by using explanation strategies

“…Whereas trust is frequently evaluated via self-reported assessments, user reliance is measured by the metric of advice-taking. Scharowski et al (2022) further emphasize the relevance of the metrics employed in XAI research. The way in which researchers operationalize these metrics has consequences for the variable of interest.…”

“…Research in AI and XAI occasionally utilizes the concepts of behavior and attitudes interchangeably, leading to potential confusion (Scharowski et al, 2022). By implementing a clinical DSS interface, Bussone et al (2015) investigated the correlation between explanations, trust, and reliance.…”

Selectively Providing Reliance Calibration Cues With Reliance Prediction

“…Reliance is a concept relevant to trust, and it is sometimes studied inclusively in the field of engineering. Trust is attitudinal and a psychological construct, while reliance focuses on the behaviors of humans, which is directly observable and thus an objective measure (Scharowski, Perrig, von Felten, & Brühlmann, 2022). Although the main focus of this paper is reliance, this section reviews both trust and reliance calibration to highlight our research because of their close relevance.…”