Controllable Complementarity: Subjective Preferences in Human-AI Collaboration

2025-03-07

AI Summary

Human-AI collaboration research has long overlooked subjective human preferences—such as sense of control and perceived enjoyment—despite their critical role in effective teamwork. Method: This work proposes explicit user controllability over AI behavior as a core mechanism to enhance complementarity and user experience. We adopt a Behavior Shaping reinforcement learning framework, enabling real-time user intervention and policy guidance during shared tasks. Contribution/Results: We introduce, for the first time, a systematic integration of subjective preferences into human-AI complementarity evaluation—moving beyond conventional objective-performance-only metrics. Empirical results demonstrate that controllable AI significantly improves users’ subjective ratings of AI effectiveness and enjoyment. Notably, AI policies retain robustness even under hidden-control conditions. These findings confirm that aligning AI behavior with human control preferences yields synergistic gains exceeding those captured by objective metrics alone.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Intelligent Robots: Behavior Learning & Control
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Economics, Online Markets and Human Computation: Research challenges in human and human-AI computation
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Research on human-AI collaboration often prioritizes objective performance. However, understanding human subjective preferences is essential to improving human-AI complementarity and human experiences. We investigate human preferences for controllability in a shared workspace task with AI partners using Behavior Shaping (BS), a reinforcement learning algorithm that allows humans explicit control over AI behavior. In one experiment, we validate the robustness of BS in producing effective AI policies relative to self-play policies, when controls are hidden. In another experiment, we enable human control, showing that participants perceive AI partners as more effective and enjoyable when they can directly dictate AI behavior. Our findings highlight the need to design AI that prioritizes both task performance and subjective human preferences. By aligning AI behavior with human preferences, we demonstrate how human-AI complementarity can extend beyond objective outcomes to include subjective preferences.

Problem

• Understanding human subjective preferences in human-AI collaboration
• Investigating controllability in shared workspace tasks with AI partners
• Designing AI that balances task performance and human preferences

Innovation

• Behavior Shaping for human-AI collaboration
• Reinforcement learning with human control
• Aligning AI behavior with subjective preferences