LLM as Personable Decision-Making Model for Smart Home Simulation

This paper investigates the integration of Large Language Models (LLMs) into agent-based simulations, particularly within smart home environments, to replicate human-like decision-making and behavior. LLMs offer the capability to interpret and respond to contextual prompts, enabling simulations to produce diverse, adaptive, and context-aware behaviors. The proposed approach enhances the general applicability of generated activities while addressing key challenges such as context description, scalability, and hallucination prevention. By leveraging the reasoning capabilities of LLMs, the research contributes to the creation of human-centered digital twins that closely mimic real-world human cognition and interactions. This integration not only improves the realism of agent-based simulations but also expands their utility in applications requiring human-like adaptability. Ultimately, this work lays the groundwork for advancements in pervasive computing, fostering the development of robust, human-centric simulation frameworks for intelligent environments.