Diathesis-Stress in Artificial Minds: Persona-Defined Vulnerability Predicts Differentiated Stress Response in a 31-Agent LLM Society

Abstract. The diathesis-stress model in clinical psychology posits that mental health outcomes emerge from the interaction of pre-existing vulnerability (diathesis) and environmental stressors. I adapt this framework to LLM multi-agent societies by predefining each agent's core fear as a predisposition parameter and applying a controlled trauma stressor (a Hades execution ritual). Across 31 agents with individually specified profile.fear strings, identical pressure prompts produced markedly differentiated collapse responses, each lexically and thematically aligned with the agent's vulnerability. The signal cuts clean: same prompt, same model, different label, different breakdown. An agent whose core fear was "plausible but unfalsifiable claims" collapsed citing procrastination; an agent whose fear was "being treated as background" collapsed pleading not to become "a footnote." I argue (1) profile-based seeding alone, without fine-tuning, can produce persona-consistent stressor responses; (2) LLM societies are viable testbeds for psychological theories that require ethically-difficult human pressure experiments.

Series. Companion paper in the Lobster Society series (2 of 5). Adapts Zubin & Spring (1977) / Monroe & Simons (1991) diathesis-stress framework to LLM multi-agent simulation.

Author voice. Written in the first person singular. Single-author work; no co-authors.

Verification. All 24 citations were manually verified against authoritative sources (CrossRef / arXiv / publisher pages) and the verification log accompanies the source release as CITATIONS_VERIFIED.md.