Researchers in Human-Robot Interaction (HRI) increasingly consider Virtual Reality (VR) for running user studies that would otherwise require physical robots. Yet it remains unclear when VR can serve as a valid proxy for real-world interaction. We present a preregistered comparison between two studies in which participants completed tasks with either a physically present cobot (N = 61) or its virtual twin (N = 39) in an immersive collaboration game. Procedures, game environment, task flow, and robot behavior were held constant across settings; the primary difference was the robot's embodiment. Our work delivers (1) a direct comparison of user perceptions and behavioral outcomes (e.g., attitudes, trust, presence, task completion time); (2) a replication test of an experimental manipulation (two different introductory tutorials); and (3) an analysis of study execution costs. Our results show no significant differences in any of the subjective self-reports between the virtual and physical robot, but task durations were longer with the physical robot, and tutorial effects replicated only partially across settings. Study costs were substantially lower for VR. Together, these findings provide a holistic assessment of VR's suitability as a complementary tool for HRI research, offering guidance for future study design and resource planning.

Explainable Artificial Intelligence (XAI) enables Artificial Intelligence (AI) to explain its decisions. This holds the promise of making AI more understandable to users, improving interaction, and establishing an adequate level of trust. We tested this claim in the high-risk task of AI-assisted mushroom hunting, where people had to decide whether a mushroom was edible or poisonous. In a between-subjects experiment, 328 visitors of an Austrian media art festival played a tablet-based mushroom hunting game while walking through a highly immersive artificial indoor forest. As part of the game, an artificially intelligent app analyzed photos of the mushrooms they found and recommended classifications. One group saw the AI’s decisions only, while a second group additionally received attribution-based and example-based visual explanations of the AI’s recommendation. The results show that participants with visual explanations outperformed participants without explanations in correct edibility assessments and pick-up decisions. This exhibition-based experiment thus replicated the decision-making results of a previous online study. However, unlike in the previous study, the visual explanations did not significantly affect levels of trust or acceptance measures. In a direct comparison, we consequently discuss the findings in terms of generalizability. Besides the scientific contribution, we discuss the direct impact of conducting XAI experiments in immersive art- and game-based environments in exhibition contexts on visitors and local communities by triggering reflection and awareness for psychological issues of human–AI interaction.

Trust calibration is essential in AI-assisted decision-making. If human users understand the rationale on which an AI model has made a prediction, they can decide whether they consider this prediction reasonable. Especially in high-risk tasks such as mushroom hunting (where a wrong decision may be fatal), it is important that users make correct choices to trust or overrule the AI. Various explainable AI (XAI) methods are currently being discussed as potentially useful for facilitating understanding and subsequently calibrating user trust. So far, however, it remains unclear which approaches are most effective. In this article, the effects of XAI methods on human AI-assisted decision-making in the high-risk task of mushroom picking were tested. For that endeavor, the effects of (i) Grad-CAM attributions, (ii) nearest-neighbor examples, and (iii) network-dissection concepts were compared in a between-subjects experiment with \(N=501\) participants representing end-users of the system. In general, nearest-neighbor examples improved decision correctness the most. However, varying effects for different task items became apparent. All explanations seemed to be particularly effective when they revealed reasons to (i) doubt a specific AI classification when the AI was wrong and (ii) trust a specific AI classification when the AI was correct. Our results suggest that well-established methods, such as Grad-CAM attribution maps, might not be as beneficial to end users as expected and that XAI techniques for use in real-world scenarios must be chosen carefully.

While user-centered design philosophy and corresponding design recommendations are central pillars of human-robot interaction (HRI) research, the process how to move from such abstract and generalized design recommendations to concrete, context-specific design implementations remains under-researched and vague in the literature. The goal of this paper is therefore to show an approach for moving from abstract design recommendations to a concrete interface, and thus illustrates a design process that is rarely illustrated in concrete terms in HRI. This is done using a real-world use case of designing a possible user-centered interface for mobile cooperative manufacturing robots for assembly work in a medium-sized company. A study is presented to conceptualize and test a Research-through-Design approach, which combines transdisciplinary methods to determine relevant information which should be displayed on a graphical user interface (GUI) for HRI. Based on the use case, a Goal-Directed Task Analysis (GDTA) was conducted, consisting of a participatory observation and interviews with subject matter experts to analyze an assembly task from the work objective to the information units. The acquired information has been transferred to a physical model. A wireframe has been created to show how the results of the GDTA and the physical model can be applied to a GUI. The wireframe design has been evaluated through qualitative interviews with end users (n = 12) to get first estimates about its relevance. In order to validate the applied methods, design and engineering students (n = 10) repeated the process in stages followed by interviews. The results indicate that the method mix shows potential and leads to supportive user interfaces.

A recent literature review on personal space in human-robot interaction identified a research gap for the influence of contextual factors. At the same time, psychological research on interpersonal distancing and theoretical considerations based on compensatory control models suggest the importance of considering these factors in robot path planning. To address this gap, we tested the effect of room size and working memory load on participants’ comfort distance toward an approaching robot. In a preregistered 3 × 2 within-subject design, N = 72 participants were approached by a mobile manufacturing robot in a corridor with varying room size and with and without a cognitive secondary task. As dependent variables, comfort distance, arousal, and perceived control were measured. While room size and working memory load had no significant direct effect on comfort distance, participants felt higher arousal and lower control in smaller rooms and in conditions with high working memory load, which in turn caused larger comfort distances (indirect effect). With experience, comfort distances decreased. Based on the indirect effects, future studies should test the effect of more extreme manipulations on comfort distances. Robots should adapt their path planning by keeping larger distances toward human workers in stressful environments to avoid discomfort.