Presentation
Calibrated vs. Overestimating Initial Information: Effects on Mental Model Accuracy and Performance in Human-Robot Collaboration
DescriptionThis study examines how initial information about robotic capabilities shapes mental models and performance in human–robot collaboration. Prior research suggests that humans tend to overestimate robotic abilities, leading to biased expectations that may impair collaboration outcomes. To investigate this, we experimentally manipulated initial system descriptions, comparing calibrated versus overestimating information in a simulated search-and-rescue task (N = 61). Mental model accuracy was assessed using a questionnaire, and task performance was measured across two rounds of increasing complexity.
Results showed that initial information significantly influenced both cognition and performance. Contrary to expectations, participants who received overestimating information exhibited smaller discrepancies in their mental models of collaboration. However, these participants showed a greater decline in performance under higher task demands compared to those receiving calibrated information. No meaningful relationship was found between mental model accuracy and task performance.
These findings suggest that initial information affects human–robot collaboration through mechanisms beyond immediate mental model accuracy. The results highlight the importance of empirically validating onboarding and training materials, as more calibrated system descriptions do not necessarily lead to better cognitive representations or improved performance.
Results showed that initial information significantly influenced both cognition and performance. Contrary to expectations, participants who received overestimating information exhibited smaller discrepancies in their mental models of collaboration. However, these participants showed a greater decline in performance under higher task demands compared to those receiving calibrated information. No meaningful relationship was found between mental model accuracy and task performance.
These findings suggest that initial information affects human–robot collaboration through mechanisms beyond immediate mental model accuracy. The results highlight the importance of empirically validating onboarding and training materials, as more calibrated system descriptions do not necessarily lead to better cognitive representations or improved performance.
Event Type
Lecture
TimeTuesday, October 20th11:20am - 11:40am PDT
Location
Similar Presentations