SPeCtrum: A Grounded Framework for Multidimensional Identity Representation in LLM-Based Agent
This work addresses the need for more authentic identity simulation in LLM agents to improve personalized human-AI interactions and simulation realism, though it appears incremental as it builds on existing identity modeling approaches.
The paper tackled the problem of oversimplified identity representation in LLM-based agents by introducing SPeCtrum, a framework incorporating social, personal, and contextual identity components; automated evaluations on drama characters showed the contextual component alone performed comparably to the full combination, while human evaluations on real individuals found the full combination enhanced authenticity.
Existing methods for simulating individual identities often oversimplify human complexity, which may lead to incomplete or flattened representations. To address this, we introduce SPeCtrum, a grounded framework for constructing authentic LLM agent personas by incorporating an individual's multidimensional self-concept. SPeCtrum integrates three core components: Social Identity (S), Personal Identity (P), and Personal Life Context (C), each contributing distinct yet interconnected aspects of identity. To evaluate SPeCtrum's effectiveness in identity representation, we conducted automated and human evaluations. Automated evaluations using popular drama characters showed that Personal Life Context (C)-derived from short essays on preferences and daily routines-modeled characters' identities more effectively than Social Identity (S) and Personal Identity (P) alone and performed comparably to the full SPC combination. In contrast, human evaluations involving real-world individuals found that the full SPC combination provided a more comprehensive self-concept representation than C alone. Our findings suggest that while C alone may suffice for basic identity simulation, integrating S, P, and C enhances the authenticity and accuracy of real-world identity representation. Overall, SPeCtrum offers a structured approach for simulating individuals in LLM agents, enabling more personalized human-AI interactions and improving the realism of simulation-based behavioral studies.