Thumbnail Image
Publication

Collabor-Action! Evaluating an adaptive model of robot performative autonomy for human-robot collaboration

Bezerra, Lara
; 
Smith, Cailyn
; 
Sousa Silva, Rafael
; 
Higger, Mark
; 
Williams, Thomas
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2025-04
Date Submitted
Keywords
Collections
2025 Spring Undergraduate Research Symposium
Show all metadata
Files
Thumbnail Image
FerreiraBezerra_Lara_UGRS2025.pdf
Adobe PDF, 2.63 MB
Research Projects
Organizational Units
Journal Issue
URI
https://hdl.handle.net/11124/180540
Embargo Expires
Abstract
In safety-critical human-robot collaboration contexts like space exploration, maintaining high human Situational Awareness (SA) is a crucial yet challenging task. Prior work has shown that Performative Autonomy (PA), a strategy in which a robot intentionally lowers its apparent autonomy, can raise SA while not substantially impacting Cognitive Workload (CW). However, research on PA has only investigated static autonomy strategies in which a robot remains at the same level of autonomy throughout an interaction. In this study, we thus investigate how robots can dynamically change their level of PA to maintain high SA and low CW during a collaborative interaction. We first present the results of a human-subjects experiment (n = 120) that shows the impact of PA strategies on SA and CW. We then use the data from this experiment for training three Reinforcement Learning (RL) algorithms to predict the appropriate levels of PA that robots should use during interactions. Finally, we present the results of a second human-subject experiment (n=120) comparing the best-performing RL algorithm to a heuristic baseline, with respect to SA, CW, and other important human-robot teaming metrics. We hypothesize that an adaptive communication style will improve human SA, especially when under increased levels of CW. By revealing how adaptive communication can mitigate the cognitive challenges of remote collaboration, we aim to identify the best design and algorithm choices for high-load human-robot applications within this paradigm.
Associated Publications
Rights
Copyright of the original work is retained by the author.
Embedded videos