“Should I Follow the Human, or Follow the Robot?” — Robots in Power Can Have More Influence Than Humans on Decision-Making

Title of the Paper

“Should I Follow the Human, or Follow the Robot?” — Robots in Power Can Have More Influence Than Humans on Decision-Making

Paper Information

• Research Area: Studies on power influence in Human-Robot Interaction (HRI) and Human-Agent Interaction (HAI)
• Keywords: Robots, intelligent agents, human-robot interaction, social power, authority, influence, human-robot hybrid teams, decision-making

Research Background and Problem

• Identified Problem or Challenge: As artificial intelligence and robots increasingly occupy significant roles in work environments, they may compete with humans for power and influence. However, how power affects human decision-making in this competition remains unclear.
• Significance: When robots and intelligent agents gain power, they may impact team collaboration, decision-making processes, and even social structures. Understanding this influence is critical for designing controllable intelligent systems and mitigating social and ethical risks.
• Research Motivation and Related Work:
  • Previous studies have shown that power and authority can make robots more influential, but most of these studies focus on single human-robot interactions, lacking exploration in group (multi-human-robot hybrid) scenarios.
  • Related research has found that power can significantly alter human acceptance of robots when using support systems to enhance collaboration, but there is a lack of direct comparison of the interaction effects between power and robot types.

Solution

• Proposed Methods and Solutions:
  • Developed an experimental framework where participants were divided into two "power assignment conditions" (robot as leader, human as leader) and a "no power difference" control group to examine the effects of power and agent type on decision-making.
  • Leveraged the "Bases of Power Framework" by French and Raven, operationalizing power through assigning team leaders "bonus allocation authority" and "legitimate leadership roles."
• Innovations:
  • Unique experimental design that simulates a complex yet controllable "three-member group" (two humans and one robot) to directly compare the influence differences between humans and robots when power is assigned.
  • Introduced an innovative method for quantifying power effects by measuring the degree of change in participants' decisions from initial to final stages to assess the impact of power and agent type.
• Specific Steps and Key Techniques:
  • In the task, participants first submitted personal suggestions, then viewed suggestions from the robot and human partners, and finally made a final suggestion under the power conditions. The changes in participants' decisions were recorded to quantify the influence.
  • The experiment utilized automated virtual operations and system feedback to ensure logical and scientifically valid task processes.

Research Findings

• Specific Findings:
  • Role of Power: Robots or humans assigned power had significantly more influence on participants' decisions than those without power.
  • Robot Influence: Even though participants had more positive attitudes toward humans, high-power robots could still be more influential than low-power humans.
  • Human Attitude Superiority: Humans were perceived as more intelligent, warmer, and less likely to cause physiological discomfort. However, high-power robots were also seen as more "capable" and could suppress human influence.
  • Comparisons with existing studies revealed that power is a more decisive factor than agent type in determining the direction of influence.
• Experimental or Evaluation Results:
  • Single-factor ANOVA and multiple group t-tests revealed the significant effect of power on the "leadership influence index."
  • In the control condition, where power differences were absent, there was no significant difference in influence between humans and robots.
  • A quantitative "team member perception" analysis model confirmed that power altered participants' perception of the robot's capability but did not change their perception of the robot's intelligence.
• Limitations and Future Directions:
  • The current experimental scenario is limited to a three-member group in a consulting context; future research should expand to more complex team structures, diverse task types, or scenarios involving physical robots.
  • The operationalization of power focused on "legitimate power" and "reward power," lacking exploration of other power types (e.g., coercive power, expert power). Future studies should investigate the interaction effects of these power types and agent types.
  • The absence of physical robot representations in the virtual experiment may limit the generalizability of conclusions to real-world interactions. Future research should use physical robots to enhance ecological validity.

Discussion and Design Implications

• Theoretical Contributions:
  • Demonstrated that power is a more critical determinant than agent type, influencing not only decision-making behavior but also participants' judgments of the agent's capability.
  • Cross-disciplinary insights: Although the experiment focused on robots, similar mechanisms and conclusions may apply to AI-supported systems. Further integration of research on their social-level impacts is needed.
• Design Recommendations:
  • Proactively consider power factors in intelligent system design: Explicit power settings (e.g., reward systems) can enhance the acceptance and trust of robots or AI.
  • Avoid excessive power reinforcement in design to prevent AI or intelligent robots from threatening human authority in human-machine collaboration. Additional regulations and constraints should address potential social and ethical issues.
• Social Risks:
  • Robots or AI as leaders may have advantages in certain scenarios, but their potential social impact is broad and profound. Further research is needed to determine optimal power configurations in real-world teams.
  • Researchers in the HRI/HAI field are urged to strengthen insights and exploration of future trends in robot power and corresponding ethical issues, such as team autonomy and fairness.