Expert Power Dynamics in Human–Robot Team Decision-Making

As artificial intelligence and autonomous systems increasingly participate in operational teams, humans may perceive these systems as holders of authority or expertise that influence group decision-making. Classic theories of social power suggest that expert, legitimate, reward, coercive, and referent power shape how teams evaluate information and select courses of action. When AI-enabled systems provide recommendations, team members may defer to those recommendations despite limited understanding of system capabilities. Importantly, these power dynamics may depend on interaction context, as cues signaling authority and expertise can differ across virtual and physically embodied environments and across team structures. Understanding how these contextual factors influence perceptions of machine expertise is therefore important for maintaining decision advantage in future HMT.

This research investigates how perceived expert power demonstrated by a designated expert influences team dynamics and decision-making in human–robot interaction contexts. The study examines two key factors: expert type (human versus robot) and interaction environment (virtual reality versus physical embodiment). Teams composed of two cadet participants complete a structured Desert Survival decision task in which they first rank survival items individually and then collaborate to produce a team ranking. During team discussion, a designated expert advisor provides recommendations and rationale for item prioritization.

In embodied conditions, participants interact with either a human confederate or a humanoid robotic agent. To maintain consistency across conditions, the confederate follows scripted responses and operates the robot using a Wizard-of-Oz methodology. Data collection integrates survey measures of team perceptions, cohesion, and responsibility attribution with behavioral interaction logs and objective decision quality based on established expert survival rankings.

The study examines whether robot experts exert influence comparable to human experts and whether physical embodiment amplifies perceived authority and directive impact. Findings will provide insight into power dynamics in HMT and inform the design of AI-enabled decision-support systems for operational environments.