When LLMs Recognize Your Space: Research on Experiences with Spatially Aware LLM Agents

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Social Psychology

Background:

• Large language models (LLMs) are evolving into agents capable of role-playing and utilizing conversational context.
• LLM agents show potential as human-like partners in social interactions.
• The impact of environmental context, specifically spatial information, on human-LLM agent interactions remains unexplored.

Purpose of the Study:

• To investigate the influence of spatial awareness in LLM agents on human-LLM interactions.
• To assess the effect of LLM agents actively reflecting spatial context on user experience.
• To explore how spatial awareness in LLM agents affects social dynamics like copresence, trust, and self-disclosure.

Main Methods:

• Participants engaged in counselling conversations with LLM agents.
• Three experimental conditions were used, varying the LLM agent's spatial awareness.
• Dependent measures included copresence, trust, therapist alliances, and self-disclosure.

Main Results:

• Participants interacting with LLM agents that actively reflected spatial information reported enhanced user experience.
• LLM agents reflecting spatial context led to increased self-description and emotional expression from participants.
• Spatial awareness in LLM agents positively influenced social interaction metrics.

Conclusions:

• Spatially aware LLM agents can significantly improve social interactions between humans and AI.
• Integrating spatial context awareness into LLM agents enhances user experience and engagement.
• Findings support the development of augmented reality applications in counseling, education, and healthcare.