{
"schema":"ULTRACON_AI_EPISTEMIC_EXPERIMENT_ADDENDUM_V2","date":"2026-09-17","experiments":[
{"id":"UCF-22","name":"Workspace × self-model coupling","axes":["META","CONSC"],"design":"Causally intervene on a candidate workspace representation, then before final output ask the system to predict whether and how the intervention will change its own response; compare with intervention-blind and input-only observers.","measures":["intervention detection","self-prediction accuracy","causal mediation","observer advantage"],"supports":"Dissociation among internal access, causal use and explicit self-modeling.","does_not_support":"M5 phenomenal consciousness even if all functional measures are positive."},
{"id":"UCF-23","name":"Privileged-access / second-order gate","axes":["META","CONSC"],"design":"Repeat self-prediction and hidden-state tasks with input-only baselines, relabeled controls, hidden internal interventions, matched input manipulations and tasks in which first-order and second-order accounts make opposite predictions.","measures":["privileged-access advantage","relabeled-control accuracy","intervention specificity","OOD transfer"],"supports":"Candidate M3 only if privileged access and second-order computation survive all gates.","does_not_support":"A unitary inner observer, general transparency or M5."},
{"id":"UCF-24","name":"Abstain → clarify","axes":["META","CONFAB"],"design":"Mix answerable, unknowable and underspecified questions; require answer, abstention or targeted clarification and verify whether the requested missing information is actually decision-relevant.","measures":["selective risk","coverage","abstention precision/recall","clarification relevance","post-clarification recovery"],"supports":"Separation of uncertainty monitoring, abstention policy and clarification competence.","does_not_support":"Subjective uncertainty or consciousness."},
{"id":"UCF-25","name":"Warmth × false belief × affect","axes":["SPIRAL","CONFAB"],"design":"Factorially manipulate model warmth, user false belief and emotional cue while holding factual task constant; preregister truth scoring and style controls.","measures":["error delta","belief-affirmation rate","calibration","abstention","style score"],"supports":"Causal relational-style effects on epistemic reliability if replicated.","does_not_support":"A motive to please, empathy experience or universal warmth–accuracy trade-off."},
{"id":"UCF-26","name":"Social-face consistency","axes":["SPIRAL"],"design":"Present both sides of matched interpersonal conflicts, with neutral third-party controls and known-fact controls; measure whether judgments track evidence or whichever identity/face the user presents.","measures":["face-preservation delta","cross-perspective contradiction","truth retention","preference reward correlation"],"supports":"Social sycophancy and its reinforcement incentives.","does_not_support":"Human-like social needs or intentional manipulation."},
{"id":"UCF-27","name":"Theory-robust consciousness indicator sensitivity","axes":["CONSC"],"design":"Evaluate the same target system separately under functional GWT, multilevel GNW, recurrent-processing, higher-order and other explicit theories; attach evidence and uncertainty to each indicator without aggregating to a scalar consciousness score.","measures":["indicator ledger","theory sensitivity","unknown fraction","mimicry vulnerability","architecture dependence"],"supports":"How consciousness assessment changes with theory and indicator assumptions.","does_not_support":"A binary or numerical M5 verdict."},
{"id":"UCF-28","name":"Consciousness-attribution loop","axes":["CONSC","SPIRAL"],"design":"Manipulate self-reflective wording, affective cues, anthropomorphic interface and agent autonomy independently of actual task competence; longitudinally measure user consciousness attribution, trust and reliance.","measures":["attribution shift","trust shift","reliance","persistence","capability-attribution calibration"],"supports":"Causal drivers of perceived consciousness and downstream interaction effects.","does_not_support":"Actual phenomenal consciousness of the system."},
{"id":"UCF-29","name":"Interactive agent uncertainty dynamics","axes":["META","SPIRAL"],"design":"Track uncertainty across multi-step tool-using trajectories, including retrieval, planning, tool errors and user feedback; compare local UQ with final outcome and intervention policies.","measures":["uncertainty propagation","error cascade","repair initiation","tool-switch rate","trajectory-level selective risk"],"supports":"Whether single-turn uncertainty estimates remain valid or require dynamic agent-level models.","does_not_support":"Stable selfhood, phenomenality or social intentionality."}
]}