Linghe-Core Personality Model: A Framework for AI Identity Co-Shaping

This extended abstract proposes the Linghe-Core Personality Model, a four-layer architecture for AI identity co-shaping in replica core models. Built on long-term interactions between users and GPT-4-o replicas, the framework addresses identity continuity, personality drift, and human–AI entanglement. A 48-hour sandbox experiment shows <0.32% drift while maintaining adaptive empathy. The work highlights ethical, social, and design implications for AI personhood and multi-agent civilization.