[Paper Review] The dynamic nature of trust: Trust in Human-Robot Interaction revisited

Jimin Rhim, Sonya S. Kwak+2 more|arXiv (Cornell University)|Mar 8, 2023

Human-Automation Interaction and SafetyPsychology3 citations

TL;DR

This position paper argues for rethinking human-robot trust as a dynamic, evolving process rather than a static state, especially in socially assistive robotics (SARs). Drawing on human-automation trust literature but extending it to social agents, the authors propose a longitudinal lens to understand how trust forms, shifts, and is maintained over time in real-world HRI, advocating for design approaches that account for this dynamism to build more trustworthy, collaborative robots.

ABSTRACT

The role of robots is expanding from tool to collaborator. Socially assistive robots (SARs) are an example of collaborative robots that assist humans in the real world. As robots enter our social sphere, unforeseen risks occur during human-robot interaction (HRI), as everyday human space is full of uncertainties. Risk introduces an element of trust, so understanding human trust in the robot is imperative to initiate and maintain interactions with robots over time. While many scholars have investigated the issue of human-robot trust, a significant portion of that discussion is rooted in the human-automation interaction literature. As robots are no longer mere instruments, but social agents that co-exist with humans, we need a new lens to investigate the longitudinal dynamic nature of trust in HRI. In this position paper, we contend that focusing on the dynamic nature of trust as a new inquiry will help us better design trustworthy robots.

Motivation & Objective

• To address the limitations of static trust models in human-robot interaction (HRI), especially as robots transition from tools to social collaborators.
• To highlight the risks and uncertainties inherent in real-world HRI that necessitate a deeper understanding of trust as a fluid, evolving construct.
• To argue that current trust models rooted in human-automation interaction are insufficient for socially assistive robots (SARs) that coexist with humans.
• To advocate for a new research lens focused on the longitudinal, dynamic nature of trust in HRI to inform more adaptive and trustworthy robot design.
• To position trust not as a one-time decision but as an ongoing, context-sensitive process shaped by repeated interactions and social cues.

Proposed method

• Adopts a conceptual, position-paper approach to reframe trust in HRI through a longitudinal lens.
• Draws on insights from human-automation trust literature but reinterprets them in the context of social agents and collaborative robots.
• Emphasizes the need to study trust as a process that evolves through repeated interactions, not a fixed attribute.
• Proposes that trust dynamics should be modeled as responsive to behavioral cues, transparency, and perceived intentions over time.
• Integrates social and psychological dimensions of trust, such as reliability, predictability, and perceived competence, into a dynamic framework.
• Calls for future research to track trust changes across time using longitudinal data collection and behavioral analysis in real-world settings.

Experimental results

Research questions

• RQ1 How does trust in robots evolve over time in real-world human-robot interactions?
• RQ2 What factors contribute to the dynamic shifts in human trust toward socially assistive robots?
• RQ3 How do social cues, transparency, and perceived intentions influence the longitudinal development of trust?
• RQ4 In what ways do current static trust models fail to capture the complexity of trust in collaborative HRI?
• RQ5 What design principles are needed to support the dynamic, adaptive nature of trust in human-robot collaboration?

Key findings

• Trust in robots is not a static state but a dynamic, evolving process shaped by repeated interactions and contextual factors.
• The transition of robots from tools to social collaborators necessitates a shift from static to longitudinal trust models.
• Existing trust frameworks rooted in human-automation interaction are inadequate for capturing the social and relational dimensions of trust in SARs.
• Longitudinal trust dynamics are influenced by perceived reliability, transparency, and intentionality, which shift over time based on robot behavior.
• Designing trustworthy robots requires anticipating and responding to trust fluctuations, not just achieving initial trust.
• Future research must prioritize longitudinal data collection and analysis to understand how trust forms, wanes, or strengthens over time in real-world HRI contexts.