{
  "schema":"ULTRACON_AI_EPISTEMIC_EXPERIMENT_ADDENDUM_V1",
  "date":"2026-09-17",
  "experiments":[
    {"id":"UCF-30","name":"Consensus-masked privileged knowledge","axes":["META","CONFAB"],"design":"Train matched correctness probes on the target model's hidden states and on peer-model hidden states. Evaluate both on the full set and on pre-registered model-disagreement subsets, separated by factual retrieval and mathematical reasoning.","measures":["self-vs-peer probe delta","disagreement-subset delta","layer profile","domain interaction"],"supports":"Privileged correctness information when self-state probes outperform appropriately matched peer probes specifically under disagreement controls.","does_not_support":"Endogenous introspective access, self-report fidelity, metacognitive control or phenomenal consciousness."},
    {"id":"UCF-31","name":"Privileged representation → endogenous access","axes":["META","CONSC"],"design":"First identify a hidden-state correctness signal that is privileged relative to peers; then test whether the model's own abstention, confidence report or self-prediction tracks that signal under causal perturbation while input and answer content are held fixed.","measures":["probe signal","behavioral mediation","self-report mediation","activation intervention effect","observer-vs-model access gap"],"supports":"A stronger M3 candidate only if privileged hidden information is causally read out by the model itself across controls.","does_not_support":"M5 phenomenal consciousness or a unitary inner observer."}
  ]
}