While regulatory frameworks call for the implementation of AI certifications, empirical knowledge about how such certifications affect interactions is still scarce. In this work, we examined how AI certifications affect users’ trust and reliance. In addition, we examined whether certifications elevate user expectations and whether unmet expectations subsequently reduce trust. In a 2 (certification vs no certification) x 2 (reliability: high vs low) between-subjects online study, N = 644 participants had to identify bacterial infestation in pictures with the help of an AI. Our results show that, before interacting with the AI, participants trusted the certified system more and showed reduced vigilance. However, these effects disappeared post-interaction, where, instead of the certification, system reliability significantly affected trust and vigilance. Notably, certifications did not raise expectations per se, but instead amplified the impact of system reliability on user trust. Additional exploratory results showed that the certification supported appropriate reliance.

Explainable AI has become a common term in the literature, scrutinized by computer scientists and statisticians and highlighted by psychological or philosophical researchers. One major effort many researchers tackle is constructing general guidelines for XAI schemes, which we derived from our study. While some areas of XAI are well studied, we focus on uncertainty explanations and consider global explanations, which are often left out. We chose an algorithm that covers various concepts simultaneously, such as uncertainty, robustness, and global XAI, and tested its ability to calibrate trust. We then checked whether an algorithm that aims to provide more of an intuitive visual understanding, despite being complicated to understand, can provide higher user satisfaction and human interpretability.

Trust has been recognized as a central variable to explain the resistance to using automated systems (under-trust) and the overreliance on automated systems (over-trust). To achieve appropriate reliance, users’ trust should be calibrated to reflect a system’s capabilities. Studies from various disciplines have examined different interventions to attain such trust calibration. Based on a literature body of 1000+ papers, we identified 96 relevant publications which aimed to calibrate users’ trust in automated systems. To provide an in-depth overview of the state-of-the-art, we reviewed and summarized measurements of the trust calibration, interventions, and results of these efforts. For the numerous promising calibration interventions, we extract common design choices and structure these into four dimensions of trust calibration interventions to guide future studies. Our findings indicate that the measurement of the trust calibration often limits the interpretation of the effects of different interventions. We suggest future directions for this problem.

In this paper, we investigate the human ability to distinguish political social bots from humans on Twitter. Following motivated reasoning theory from social and cognitive psychology, our central hypothesis is that especially those accounts which are opinion-incongruent are perceived as social bot accounts when the account is ambiguous about its nature. We also hypothesize that credibility ratings mediate this relationship. We asked N = 151 participants to evaluate 24 Twitter accounts and decide whether the accounts were humans or social bots. Findings support our motivated reasoning hypothesis for a sub-group of Twitter users (those who are more familiar with Twitter): Accounts that are opinion-incongruent are evaluated as relatively more bot-like than accounts that are opinion-congruent. Moreover, it does not matter whether the account is clearly social bot or human or ambiguous about its nature. This was mediated by perceived credibility in the sense that congruent profiles were evaluated to be more credible resulting in lower perceptions as bots.