Introduction to the Special Issue on Trust in Artificial Intelligence

Abstract

Trust in artificial intelligence (AI) is becoming an increasingly important topic for the human factors community of practice. We continue to see novel applications of AI, more recently including generative AI, being deployed for more high-consequence work (e.g., national security, healthcare, and transportation). Calibrated trust in AI-based technologies is critical for supporting performance in human-machine teaming constructs and, for adoption of such technologies.

Although trust in AI (and trust in automation, more broadly) has been an increasingly prominent focus of scholarship and practice, some have argued via papers and panel discussions that there are limitations in the extant body of research. This special issue was an attempt to fill gaps and/or push boundaries, with three specific emphases:

1. New concepts, frameworks, and definitions of trust : Trust has long been defined as an attitude held by a human in a situation characterized by uncertainty and vulnerability (Lee & See, 2004); however, some have suggested trust being something else. Thus, we sought research that re-examines or refines our understanding of trust, trustworthiness, and other related concepts, similar to the recent work of Chiou & Lee (2023).

2. Methods and tools for engineering trustworthy AI : Trust is often something that is studied or measured after the development of an AI-based technology, or supported during the design phase with extant human factors tools and principles. As I recently wrote with my colleague (Dorton & Stanley, 2024), there is a dearth of practical tools and techniques for practitioners to employ in “trust engineering” of AI-based systems (see Ezer et al., 2019). Thus, we sought methods and tools that allow for the engineering of more trustworthy AI before it is deployed in real-world operations (often with serious consequences).

3. Naturalistic assessment of trust in sociotechnical systems : Trust in AI and autonomy has largely been studied in controlled laboratory settings with simple synthetic tasks and naïve users. Since trust is defined as an attitude in a situation marked by vulnerability, some have argued (at varying levels of seriousness) that simple studies with students in no real peril lack construct validity. Thus, we sought research that explores trust dynamics with real end users, using real technologies, in real work, with real vulnerability (i.e., in the tradition of the naturalistic decision making movement).

Based on the stated interests for this special issue, we present three papers. Hoffman (2024) addresses the first point of interest, providing an argument for focusing on trusting as an emergent, rather than focusing on trust as an attitude, due to its highly dynamic nature. Hoffman also disentangles trust from reliance and highlights limitations of current design principles regarding trustworthy AI.

Tolk (2024) addresses the second point of interest (and the first, to some extent), encouraging us to apply lessons learned from the modeling and simulation (M&S) community to AI. That is, the M&S community has long dealt with fostering trust in large and complex models, and that AI is merely a specific instance of a model. More specifically, Tolk argues that there may be merit in shifting focus from trust in the model or technology to trust in the results generated by the model, which may be moderated by relationships between the users and providers of AI-based technologies.

Finally, Dorton et al. (2025) address the third (and first) point of interest by providing a case study of examining trust from a more naturalistic perspective by working with submariners. They also demonstrate insights gained from examining trust in the context of the entire sociotechnical system (rather than the human-AI dyad), and present a concept of Systemic Trust—the idea that trust is mediated by all components of the sociotechnical system into which it is integrated: the people, technology, work, and the interactions of these components.

Author note

Approved for public release, distrubution unlimited: PR_25-01912-6.

References

Chiou

E. K.

Lee

J. D.

(2023). Trusting automation: Designing for responsivity and resilience. Human Factors, 65(1), 137–165. https://doi.org/10.1177/00187208211009995

Dorton

S. L.

Korman

Ralph

J. M.

Speaker

S. R.

(2025). Systemic trust in artificial intelligence. Ergonomics in Design, 10648046251366116. https://doi.org/10.1177/10648046251366116

Dorton

S. L.

Stanley

J. C.

(2024). Minding the gap: Tools for trust engineering of artificial intelligence. Ergonomics in Design, 33(3), 142–147. https://doi.org/10.1177/10648046241249903

Ezer

Bruni

Cai

Hepenstal

S. J.

Miller

C. A.

Schmorrow

D. D.

(2019). Trust engineering for human-AI teams. Proceedings of the Human Factors and Ergonomics Society, 63(1), 322–326. https://doi.org/10.1177/1071181319631264

Hoffman

R. R.

(2024). Trusting as an emergent: Implications for design. Ergonomics in Design, 10648046241295270. https://doi.org/10.1177/10648046241295270

Lee

J. D.

See

K. A.

(2004). Trust in automation: Designing for appropriate reliance. Human Factors, 46(1), 50–80. https://doi.org/10.1518/hfes.46.1.50_30392

Tolk

(2024). Can we learn from trust in simulation to gain trust in AI? Ergonomics in Design, 10648046241291832. https://doi.org/10.1177/10648046241291832