Personality Attractors and Dominance Lock in Dialogue-Based Cognitive Agents

This work presents Entelgia, an experimental cognitive-agent architecture exploring how internal structures such as memory, emotional signaling, and observer-based reflection influence emergent dialogue behavior in large language model agents. Unlike traditional agent frameworks focused primarily on tool orchestration, Entelgia investigates internal state dynamics as primary drivers of interaction stability and identity evolution. The study introduces the concept of Personality Attractors — stable behavioral configurations emerging through repeated dialogue — and Dominance Lock, a failure mode in which conversational agents converge toward asymmetric interaction patterns. The paper provides: • a formal architectural description of the Entelgia system • operational definitions of META behavioral metrics • dialogue-derived experimental observations • reproducibility and replication procedures • statistical characterization of interaction dynamics This repository and publication are intended as an exploratory research artifact rather than a production framework. The goal is to stimulate investigation into internally regulated AI systems and emergent cognitive behavior. All experiments were conducted using local LLM agents interacting through structured dialogue loops with persistent internal state tracking. Source code and reproducibility materials are provided.