Trust in Human-Autonomy Teaming¶

Status: emerging
Last updated: 2026-05-31
Sources: Ergoship2021A.Hynnekleivm.Lutzhoftfinal.Pdf
Tags: [trust, trustworthiness, human-autonomy-teaming, mass, automation-acceptance, humane, remote-operations]

Summary¶

Trust is identified as a central issue for autonomous maritime operations and a precondition for successful remote operations and public acceptance (Hynnekleiv & Lützhöft, 2021). Drawing on HUMANE project expert workshops, Hynnekleiv and Lützhöft distinguish three meta-categories of trust and trustworthiness issues: trust as technology acceptance, operational trust, and trustworthiness. The ideal they describe is calibration — alignment between a human's trust and the objective trustworthiness of the technology — which requires both transparency in the autonomous agent's choices and appropriate training and design.

Body¶

Context¶

Hynnekleiv and Lützhöft (2021) examine trust issues in human-autonomy teaming for autonomous maritime operations. Their evidence is a qualitative analysis of HUMANE project expert-workshop data combined with a data-driven literature review, treating trust as a property of the sociotechnical system rather than a fixed attitude. The conference paper is itself an output of the HUMANE project (Humane Project), and within this knowledge base it supplies the trust dimension of the same workshop programme: it makes transparency a precondition for human-autonomy teaming, linking it to the interface response in Human In The Loop Automation Transparency, and treats trust as trainable, connecting it to the competence agenda in Seafarer Skills And Competence For Mass.

Key Points¶

The paper situates trust within human-autonomy teaming (HAT), one design paradigm for intelligent technology. The ISO term Robotic, Intelligent, Autonomous (RIA) describes technology able to self-regulate and self-govern, enabling more social interaction with humans. Among ISO's listed design paradigms — augmentation, remoting, replacement, teaming, symbiosis — teaming receives the most attention from the Human Factors community, and a prerequisite for HAT is transparency behind the choices an autonomous agent offers, which helps humans decide whether to trust it (PDF p. 2, orig. p. 1). This links trust directly to the automation-transparency design problem examined in Human In The Loop Automation Transparency.

The empirical basis is the HUMANE workshops, which gathered experts in focus-group style across themes of system safety and cybersecurity (2018), rules, regulations and classification (2019), and skills, training and education (2019), involving 59 experts from shipping companies, classification societies, technology manufacturers, ship owners, government agencies, insurers, and academia (PDF p. 2, orig. p. 1). The workshops did not ask direct questions about trust; the trust themes emerged from analysis of data collected on other topics, which the authors treat as evidence of how pervasive the concern is (PDF p. 2, orig. p. 1).

The central finding is a three-part structure of trust issues: trust understood as technology acceptance, operational trust, and trustworthiness, identified as distinct meta-categories (PDF p. 4, orig. p. 3). The authors connect these to existing theory and argue that the desired end state is calibration between human trust and objective technological trustworthiness — neither over-trust, which risks complacency, nor under-trust, which produces unwillingness to operate and excessive supervisory action (PDF pp. 5–6, orig. pp. 4–5).

Conclusion¶

Hynnekleiv and Lützhöft (2021) conclude that trust is a central, pervasive issue for autonomous maritime operations and a designable, trainable property rather than a fixed attitude. The end state they describe is calibration between human trust and objective trustworthiness, achieved through transparency, training, and design. They note that more research is needed to provide guidelines for reaching calibrated trust, which the article does not yet supply.

References¶

Hynnekleiv, A. & Lützhöft, M. (2021) 'Designing for trustworthiness, training for trust. An overview of trust issues in human autonomy teaming', Ergoship 2021. Western Norway University of Applied Sciences. hynnekleiv2021trust

Open Questions¶

The paper identifies the need for guidelines to achieve calibrated trust but does not yet provide them; this remains an open research task.
How does the three-category trust model relate to Endsley-style situation awareness loss under automation? Cross-domain evidence is held in the unverified Raw/Meta/Kognitive Grenzen Fernueberwachung Sa Assistenzsysteme.Md.Md digest and not yet compiled.