Reproduce “Understanding Large-Language Model (LLM)-powered Human-Robot Interaction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Callie Y. Kim, Christine P. Lee, Bilge Mutlu
Researchers at the University of Wisconsin–Madison empirically investigated how Large Language Model (LLM)-powered robots are perceived by users across different tasks compared to text and voice-only LLM agents. Their study found that embodied robots enhance social engagement and rapport in instructional and negotiation tasks, while highlighting challenges in creative collaboration and information retrieval due to communication inefficiencies and social pressure from the robot's presence.
Key insights
- LLM-powered robots elevate user expectations for sophisticated non-verbal cues; a mismatch between verbal sophistication and non-verbal expression can lead to discomfort.
- Task characteristics are crucial for determining the suitability of embodied LLM agents, with social and deliberative tasks benefiting more from embodiment than those requiring high precision, efficiency, or uninhibited creativity.
- The social presence of an embodied robot, when powered by an LLM, can introduce unexpected social pressure and accountability on users, affecting task performance in creative or open-ended scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.