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

Paper Title

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

Publication Info

• Topic area: Human-robot collaboration and power dynamics
• Keywords: Human-robot interaction, power legitimacy, compliance, leadership, collaboration, social psychology, automation, artificial intelligence, workplace robots, human-computer interaction

Background and Problem

• Problem / challenge: Studies on robot power and compliance show inconsistent results, with unclear factors influencing whether humans comply with robot leaders.
• Significance: Understanding compliance with robot leaders is critical as robots increasingly assume leadership roles in workplaces, affecting collaboration efficiency and human acceptance.
• Motivation and related work: Prior research has explored robot power through trust, persuasion, and shared autonomy but has not systematically examined legitimacy as a moderating factor. Mixed findings suggest legitimacy might explain compliance variability, but this has not been directly tested.

Solution

• Proposed approach: Investigate the role of power legitimacy in human-robot collaboration through three preregistered experiments manipulating power assignment and legitimacy.
• Novelty:
  1. Introduced legitimacy as a psychological construct moderating compliance with robot leaders.
  2. Conducted three experiments to systematically test legitimacy's role in shaping compliance.
  3. Developed a multigroup mediation model to explore legitimacy's emergence in natural human-robot collaboration.
  4. Highlighted the broader implications of legitimacy for human-robot interaction (HRI) theory and design.
• Procedure and key techniques:
  • Experiment 1: Manipulated power assignment (robot vs. human leader) and legitimacy (legitimate, illegitimate, no explanation) based on competence.
  • Experiment 2: Controlled competence and introduced procedural fairness (rock-paper-scissors) to manipulate legitimacy.
  • Experiment 3: Examined natural emergence of legitimacy in ecologically valid settings, manipulating teammate type (human vs. robot), power assignment, and task ability.
  • Measured compliance, perceived legitimacy, social attributes, and affective responses.

Results

• Concrete findings:
  • Compliance with robot leaders was higher when their power was perceived as legitimate.
  • In no-explanation conditions, participants defaulted to perceiving human power as legitimate and robot power as illegitimate.
  • Multigroup mediation analysis showed that lower perceived legitimacy of robot leaders fully mediated reduced compliance.
• Advantage over baselines:
  • Legitimacy emerged as a critical factor explaining compliance variability, reconciling prior inconsistent findings.
  • Procedural fairness (e.g., rock-paper-scissors) effectively conferred legitimacy on robot leaders.
• Experiments / evaluation:
  • Experiment 1 (N=131): Tested legitimacy via competence; found legitimacy influenced compliance but also noted competence effects.
  • Experiment 2 (N=137): Controlled competence and used procedural fairness; confirmed legitimacy's role in compliance.
  • Experiment 3 (N=163): Explored natural legitimacy emergence; found robot leaders perceived as less legitimate than human leaders, reducing compliance.
  • Metrics: Compliance scores, perceived legitimacy, social attributes (warmth, competence, discomfort), affective responses.
• Limitations and future work:
  • Focused on power acquisition legitimacy; future work should explore legitimacy during power enactment.
  • Limited task and power scope; broader social relations and dynamic power redistribution need investigation.
  • Virtual robot interaction may not generalize to physical robots; stronger legitimacy manipulations could enhance findings.

Summary

This study highlights legitimacy as a critical moderator of compliance in human-robot collaboration, showing that robot power elicits compliance only when perceived as legitimate. Through three experiments, the authors demonstrated that legitimacy can be conferred through competence, procedural fairness, or naturally inferred in collaborative contexts. The findings reconcile prior inconsistent results on robot power and compliance, offering theoretical insights and practical design implications for fostering effective and psychologically acceptable human-robot interactions. Future research should explore legitimacy in broader contexts, dynamic power roles, and physical robot interactions.