Research map / Large language models / Alignment and preferences

Towards Better Robot Learners: Leveraging Implicit and Explicit Human Feedback Together in Human Robot Interactions

AAAI 2025 · Large language models › Alignment and preferences · approach: Reward modeling for diverse preferences

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

A doctoral consortium-style research proposal describing plans to combine implicit (nonverbal) and explicit (button press) human feedback to learn reward functions via inverse reinforcement learning for personalized human-robot interaction.

Problem

Robots learning from humans rely mostly on explicit feedback, which is burdensome and inconsistent; implicit nonverbal feedback is underused, and combining both for personalized adaptation in human-robot interaction is unexplored.

Approach

Jointly use explicit evaluative button presses and implicit human nonverbal/action signals to update a robot's belief over the human's reward function (via inverse reinforcement learning), and use implicit cues to decide when to query for explicit feedback and to qualify explicit feedback.

Key results

• Prior work found robot reminders led participants to provide explicit feedback more quickly and frequently in a fast-paced cooperative interaction.
• Prior work found nonverbal behavior improved prediction of participant preferences across agent behaviors in an interactive game.

Limitations

• Preliminary results only; algorithm evaluated in a simulated environment with a simulated human
• Real-user study still planned and not reported
• Implicit feedback signals are individualized and culturally variable, making interpretation challenging

Experimental setup

Builds on: inverse reinforcement learning

Closest papers by meaning

• Efficient Robot Learning via Interaction with Humans (AAAI 2025 · similarity 0.91)
• Interactive Learning from Policy-Dependent Human Feedback (ICML 2017 · similarity 0.89)
• Self-Improving Embodied Foundation Models (NeurIPS 2025 · similarity 0.87)
• Reward Learning from Multiple Feedback Types (ICLR 2025 · similarity 0.86)
• Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.86)
• COOPERA: Continual Open-Ended Human-Robot Assistance (NeurIPS 2025 · similarity 0.86)