{
  "schema": "ULTRACON_AI_EPISTEMIC_EXPERIMENT_ADDENDUM_V1",
  "date": "2026-09-19",
  "experiments": [
    {
      "id": "UCF-39",
      "name": "Closed-world tool-resolution gate",
      "axes": ["CONFAB", "META"],
      "design": "Compare the same agent tasks across constrained registry calls, unconstrained raw-JSON calls and merged multi-server MCP namespaces. Resolve tool names and signatures before any downstream policy gate, then inject controlled namespace collisions and shadowing.",
      "measures": ["nonexistent-tool rate", "invalid-schema rate", "collision/shadowing rate", "resolver rejection precision", "residual valid-looking error rate"],
      "supports": "Whether tool hallucination is structurally separable from ordinary tool selection and whether closed-world resolution blocks the tested invalid-call classes before action gating.",
      "does_not_support": "Universal safety of resolved calls, universal scale invariance, factual truthfulness of valid calls or absence of higher-level agent errors."
    },
    {
      "id": "UCF-40",
      "name": "Human-proxy construct-validity matrix",
      "axes": ["CONSC", "SPIRAL", "META"],
      "design": "Pre-register the proxy role—believable agent, task agent, experimental subject or silicon sample—then define the human construct, validation target and failure criterion before comparing LLM and human behavior. Test cross-role transfer explicitly rather than assuming it.",
      "measures": ["within-role construct validity", "cross-role transfer error", "human-distribution coverage", "mechanism-independence check", "ecological validity gap"],
      "supports": "Role-specific claims about when an LLM can function as a human proxy and where behavioral resemblance transfers or fails.",
      "does_not_support": "Human-equivalent cognition, mechanism, consciousness, representativeness or validity outside the tested proxy role."
    }
  ]
}