← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

Why Don't People Follow Robot Leaders? Understanding the Effects of Power Legitimacy on Compliance with Agents

Huajie Jay Cao , Minrui Chen , Wei Peng , Hee Rin Lee

DOI PDF Program page Page JSON

This is a strong honorable-mention paper because it turns a messy empirical question—do people follow robot leaders?—into a cleaner theoretical one about legitimacy. The three-study package is persuasive, especially because it combines explicit manipulations with a mediation account in a more ecologically grounded setting. The main caution is scope: the evidence is strongest for assigned roles, controlled decision tasks, and mediated collaboration rather than for all forms of robot authority in the wild.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

What makes this paper notable is that it does not merely add another result to the literature on robot influence; it explains why the literature has been inconsistent. The authors argue that robot power is not a direct cause of compliance. Instead, legitimacy is the key psychological condition that determines whether people accept a robot’s authority. That is a strong conceptual move because it converts a binary debate—robots in power work versus do not work—into a conditional account that better matches the mixed prior findings the paper reviews. The empirical package is also well matched to that claim. The paper reports three preregistered experiments with a total N of 431. The first two studies manipulate power assignment and legitimacy through competence and procedural fairness, which gives the authors two distinct ways to test whether justified versus unjustified authority changes compliance. The third study is especially important because it moves beyond direct manipulation and asks whether legitimacy emerges naturally in a more ecologically grounded collaboration scenario. There, the mediation result is the clearest theoretical payoff: the robot leader is perceived as less legitimate than the human leader, and that lower legitimacy accounts for reduced compliance. That is exactly the kind of evidence needed for a theory contribution rather than a one-off behavioral effect. The paper is also commendably self-limiting. It does not pretend to have solved robot authority in general. The authors explicitly note that power is operationalized narrowly through role assignment and resource control, that the task paradigm is limited, that the setting is simulated, and that the interaction is primarily with a virtual rather than tangible robot. They also acknowledge that the legitimacy manipulation is relatively weak. Those limitations matter because they keep the contribution in the right register: this is a strong causal explanation for compliance in constrained collaborative settings, not a universal law of human-robot governance. Overall, I would place the paper as a valuable theory-building contribution in HRI and CHI. Its biggest payoff is giving the field a better explanatory variable. Future work on robot leadership, persuasive agents, workplace automation, and human-AI authority can use legitimacy to distinguish mere capability from socially accepted power. That makes the paper more than an isolated experiment set; it is a useful conceptual anchor for subsequent research and design.

What Changed

Canon before

Prior HRI and social-influence work had already shown that robots can sometimes persuade, direct, or outperform humans in collaborative settings, but findings about robot power and compliance were inconsistent. The field had examples where robot leaders increased compliance, others where authority backfired, and little agreement on whether the decisive factor was role, competence, embodiment, or simple deference to automation.

Departure from common sense

The paper argues against the simple intuition that assigning a robot authority should automatically make people follow it. Instead, power only works when people accept that power as legitimate. That is a meaningful departure from a naive “authority causes obedience” story, because the studies show that the same nominal leader role can produce different compliance outcomes depending on whether the robot’s position is justified through competence, fair procedure, or naturally inferred legitimacy.

Actual novelty

The paper’s real novelty is not a new robot artifact but a theory-level explanation for mixed findings in robot authority research. It introduces legitimacy as the psychological acceptance mechanism that conditions whether robot power translates into compliance, then tests that idea across three preregistered experiments. The contribution is strongest as a causal account: robot leadership is not inherently effective or ineffective; its effect depends on whether people see the robot’s power as proper, fair, and justified. The third experiment extends this beyond explicit manipulation by showing that lower legitimacy of a robot leader, relative to a human leader, mediates reduced compliance in a more ecologically grounded collaboration scenario.

Evidence

The evidence base is substantial for a CHI theory contribution. The paper reports three preregistered experiments with a combined N = 431. Experiments 1 and 2 manipulate power assignment and legitimacy using competence and procedural fairness, while Experiment 3 examines whether legitimacy emerges naturally in a more ecologically valid collaboration setup and tests mediation. Across these studies, the authors repeatedly show that legitimacy changes how robot power affects compliance, and they also report downstream effects on perceived social attributes and affect. The paper further strengthens credibility by discussing where the evidence is bounded, including narrow power operationalization, limited task structure, simulated collaboration, mediated interaction, and relatively weak legitimacy manipulation.

Limits

Method limits

The methods are careful but bounded. Power is defined as control of resources and operationalized mainly through role assignment, control over final decisions, and bonus allocation, which captures one organizational form of power rather than the full range of human-robot authority relations. The task paradigm is also narrow: participants make uncertain allocation decisions on a continuous scale in a controlled web setting. In addition, the paper notes that the legitimacy manipulation was relatively weak, which may explain why some attitudinal outcomes were less responsive than compliance itself.

Deployment limits

The findings transfer most directly to settings where authority is assigned, explained, and interpreted socially, such as workplace-like collaborations with explicit leader and follower roles. They are less directly applicable to embodied field deployments, dynamic teams where leadership shifts across tasks, multi-human teams coordinated by one robot, or high-stakes environments where compliance is constrained by policy, safety rules, or institutional accountability rather than perceived legitimacy alone.

Boundary conditions

The claims are bounded by several conditions the paper itself makes visible. Effects depend on how power is framed, whether legitimacy cues are provided or inferred, and whether the robot is compared with a human teammate or leader. The studies also suggest that competence cues, procedural fairness, and no-explanation defaults matter differently across leader and follower roles. More implicit dependence relations, discrete or binary decision tasks, physically collaborative work, and dynamic redistribution of power could all change how legitimacy forms and how compliance is expressed.

Manual check: No major manual check is required for the repaired narrative fields, but publication workflows should still verify the quoted spans and offsets against the source text because publish gating depends on exact evidence localization.

Position in field

This paper sits at the intersection of HCI, HRI, and social-psychological work on power, legitimacy, and compliance. Its main value is as an organizing explanation for a fragmented literature on robot authority: instead of asking only whether robots in power increase compliance, it asks when that power is accepted as proper. That reframing helps connect prior findings on robot expertise, role appropriateness, persuasion, and resistance to automation. In the field, the paper is best understood as a theory-building contribution that gives researchers a sharper construct for studying agent authority and gives designers a more grounded basis for thinking about when robot leadership will be accepted rather than merely imposed.

Abstract

Artificially Intelligent systems such as robots are increasingly integrated into the workplace and gaining more power. Yet studies on robot power and compliance report mixed findings. To address these inconsistencies, we introduced legitimacy as people's psychological acceptance of power. Three preregistered experiments were conducted (N = 431). In Experiment 1 and 2, we manipulated power assignment (robot power vs. human power), and legitimacy of power (legitimate, illegitimate, no explanation) through competence and procedural fairness. The results showed that participants complied more to the legitimate robot power than illegitimate one. In Experiment 3, we examined whether perceptions of legitimacy would emerge naturally in more ecologically valid collaboration. Results of multigroup mediation model showed that the robot leader was perceived as less legitimate than the human leader, which accounted for the reduced compliance to the robot’s decisions. In all three experiments, people’s perceived social attributes of robots with power and their affective responses were negatively affected. Theoretical and design implications are discussed.