ELDER-SIM Platform Enables Personality-Stable Digital Twins for Elderly Care Using LLMs

A new platform called ELDER-SIM has been developed to create personality-stable digital twin agents for older adults, addressing the challenge of personality drift in conversational agents. The system uses large language models to capture lived experiences and behavioral responses over time, which is crucial for reliable geriatric care simulations. Implemented through n8n workflow orchestration with local LLM inference via Ollama/vLLM, it integrates Big Five (OCEAN) trait specifications for personality consistency. A Cognitive Conceptualization Diagram grounded in Beck's CBT framework and a MySQL-based long-term memory module are key components. The research proposes a psychometric validation framework specifically designed to quantify personality consistency in LLM-based agents. This development responds to the central barrier of inconsistent trait expression across repeated interactions that undermines simulation reliability. The platform represents a significant advancement in creating patient-facing conversational agents that maintain stable personality characteristics. The work was documented in arXiv preprint 2604.16343v1, which announced the cross-disciplinary research.