{
  "id": 6156,
  "url": "https://arxiv.org/abs/2604.07838v1",
  "title": "We Need Strong Preconditions For Using Simulations In Policy",
  "summary": "Simulations, and more recently LLM agent simulations, have been adopted as useful tools for policymakers to explore interventions, rehearse potential scenarios, and forecast outcomes. While LLM simulations have enormous potential, two critical challenges remain understudied: the dual-use potential of accurate models of individual or population-level human behavior and the difficulty of validating simulation outputs. In light of these limitations, we must define boundaries for both simulation dev",
  "authors": "Steven Luo, Saanvi Arora, Carlos Guirado",
  "category": "research",
  "topics": "regulation,agents-autonomy",
  "orgs": null,
  "regions": null,
  "published_at": "2026-04-09T05:39:35.000Z",
  "fetched_at": "2026-07-14T16:32:15.638Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/6156",
  "original_url": "https://arxiv.org/abs/2604.07838v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}