{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"authority_mentor_relationship_anchoring__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"authority_mentor_relationship_anchoring__computer_science__CONSTRAINED_HIGH","decision":"PROPOSAL","abstention_reason":null,"proposal":{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","cell_id":"authority_mentor_relationship_anchoring__computer_science","arm":"CONSTRAINED_HIGH","candidate_id":"authority_mentor_relationship_anchoring__computer_science__CONSTRAINED_HIGH","proposal_index":1,"version":0,"title":"Reference-Coupled Physical Apprenticeship for Analog Computing Arrays","problem":"Nominally identical analog computing arrays can realize different input-output behavior because their physical conductance elements respond differently to fabrication and programming. A newly fabricated array can therefore satisfy a digital weight specification yet fail to reproduce a trusted array's realized transfer behavior, including coupled device effects that are difficult to capture as independent numerical settings.","actors":["A characterized analog reference array serving as the mentor anchor","A newly fabricated adjustable analog array serving as the mentee","A resistor-bridge comparison fixture that physically couples the arrays during transfer","A secondary characterized reference array used to detect mentor-specific defects","A hardware safety engineer authorized to approve the bounded experiment"],"observable_state":"When the reference and new arrays receive the same sequence of probe voltages, corresponding output-line currents differ in sign and magnitude beyond a prespecified electrical tolerance, despite having been programmed from the same nominal settings.","consequence":"The new array does not reproduce the reference array's physical transfer response, so computations mapped onto nominally interchangeable hardware can yield materially different electrical outputs.","affected_objective":"Reproducible, standalone analog computation across replaceable hardware arrays without requiring permanent digital correction.","intervention":"Place one characterized reference array and one adjustable new array in a removable analog coupling fixture. Apply identical physical voltage probes to both. For each output line, a differential bridge converts their current mismatch directly into polarity-correct programming pulses applied to the new array's adjustable conductance elements. Pulse charge is bounded by fuses, series resistance, and hard voltage clamps. As mismatch shrinks, the bridge's physical error signal and consequent pulse energy diminish. The coupling resistance is then increased in fixed hardware stages before complete disconnection, requiring the new array to retain the matched response while operating alone. A second reference array is substituted for a limited cross-check so a defect peculiar to the first reference is not automatically treated as the target.","structural_mapping":[{"archetype_element":"Legitimate Mentor Anchor","domain_realization":"A stable reference array admitted only after direct electrical characterization establishes the bounded transfer response it is meant to transmit."},{"archetype_element":"Mentee Readiness and Consent Boundary","domain_realization":"A new array enters coupling only if its device ranges, polarity, endurance budget, and safe programming envelope are compatible with the fixture."},{"archetype_element":"Relational Safety Container","domain_realization":"Galvanic isolation where needed, current-limiting resistors, fuses, voltage clamps, and thermal cutoffs bound what the reference-coupled bridge can do to either array."},{"archetype_element":"Cultural Norm and Value Payload","domain_realization":"The payload is the reference array's realized multi-line transfer surface over a declared probe envelope, not its nominal digital parameter file or every incidental behavior outside that envelope."},{"archetype_element":"Modeled Practice and Judgment Window","domain_realization":"Both arrays experience the same physical probes, exposing how the reference responds under combinations of voltages rather than transmitting isolated target numbers."},{"archetype_element":"Dialogic Interpretation Loop","domain_realization":"The differential bridge repeatedly compares simultaneous reference and learner currents; each residual mismatch physically determines the polarity and bounded energy of the next conductance adjustment."},{"archetype_element":"Autonomy and Exit Safeguard","domain_realization":"A hardware disconnect opens the coupling, and the new array must preserve its response without the reference or fixture before it is accepted."},{"archetype_element":"Secondary Reference Anchor","domain_realization":"A separately characterized array provides a limited alternate comparison that can reveal a feature unique to the primary reference."},{"archetype_element":"Progressive Autonomy Release","domain_realization":"Successive increases in coupling resistance reduce correction strength before the arrays are completely separated."}],"mechanism_mapping":[{"mechanism_slug":"guided_shadowing_with_debrief","role":"Shared voltage probes make the new array physically shadow the reference; the differential bridge exposes residual response differences as measurable error currents.","counterfactual_removal":"Without shared probes and differential comparison, no physical relation connects the reference's realized behavior to the new array, leaving only nominal open-loop programming."},{"mechanism_slug":"joint_practice_with_corrective_feedback","role":"Simultaneous electrical operation produces local, polarity-correct physical programming pulses whose energy falls as the paired responses converge.","counterfactual_removal":"Without mismatch-driven conductance adjustment, observing both responses does not alter the new array and cannot transfer the reference response."},{"mechanism_slug":"mentor_rotation_or_second_opinion_channel","role":"Temporary substitution of a second characterized reference tests whether the primary reference is transmitting an idiosyncratic defect.","counterfactual_removal":"The primary transfer can still occur, but there is no physical plural-reference check against copying a mentor-specific anomaly."}],"causal_chain":["Identical probe voltages excite the reference and new arrays concurrently.","Fabrication or programming mismatch produces unequal output currents.","The resistor bridge converts current differences into local analog error voltages.","Hard-limited pulse-forming circuitry converts error polarity and magnitude into physical programming charge delivered to adjustable elements of the new array.","Conductance changes alter the new array's subsequent current response.","Reduced current mismatch reduces subsequent programming energy, permitting convergence within a declared tolerance if the device physics is suitable.","Staged weakening and final removal of the coupling test whether the altered material state persists independently.","A secondary-reference exposure distinguishes broadly shared response features from a primary-reference anomaly."],"baseline":"Program each new array from the same nominal conductance table using fixed open-loop pulse recipes, then operate it without reference coupling. This baseline does not use realized response mismatch to compensate for device-specific physical variation.","nearest_rivals":["Per-element electrical write-and-verify against independently specified scalar conductance targets","Permanent digital calibration tables or software correction applied to each array's outputs","Storing weights digitally and using conversion circuitry instead of relying on matched analog conductance states","Conventional post-fabrication trimming against an externally specified transfer-function target rather than a peer reference array"],"remaining_contrastive_claim":"Relative to fixed open-loop programming of nominal targets, the proposed fixture could transfer a characterized reference array's realized multi-line electrical response into a physically variable new array through direct paired excitation and local error-current-driven material adjustment; unlike permanent digital correction, any resulting match must persist after all coupling and computation wrappers are removed. Whether it does so is an empirical question.","authority_safety":{"decision_authority":"A hardware safety engineer may authorize only a benchtop experiment within documented voltage, current, temperature, and cumulative programming-charge limits; acceptance of the resulting array for any operational system remains outside this experiment.","authorized_first_step":"Couple one small reference array and one sacrificial compatible array through fused, current-limited channels and test only a prespecified low-energy probe set behind a physical emergency disconnect.","excluded_actions":["Connection to production systems","Use on irreplaceable reference hardware","Operation above characterized electrical or thermal limits","Automatic acceptance based solely on agreement with the primary reference","Permanent coupling of the reference to the new array","Software compensation during the standalone retention test"],"halt_rollback":"A fuse event, clamp activation, unexpected heating, oscillatory mismatch, monotonic error growth, or programming-charge-limit approach immediately removes probe and programming power. The arrays are physically disconnected; the sacrificial array is quarantined, and the untouched baseline specimen remains the rollback reference."},"negative_tests":{"strongest_counterevidence":"After direct coupling, matched behavior disappears on disconnection, or the apparent improvement is fully explained by temporary electrical loading from the reference fixture rather than persistent conductance change in the new array.","problem_falsifier":"Repeated nominally identical arrays already reproduce the declared transfer surface within tolerance using open-loop programming, leaving no independently observable physical transmission gap for this intervention to address.","intervention_falsifier":"Across the bounded probe envelope, error-current-driven programming fails to reduce held-out standalone transfer error relative to both the array's pre-coupling state and the open-loop baseline, or it succeeds only while software correction or reference loading remains present.","risks":["A defective or drifting reference response may be physically copied.","Programming pulses may accelerate wear or create irreversible device damage.","Coupled channels may oscillate or converge to an unintended electrical state.","Matching the tested probe envelope may degrade behavior outside that envelope.","A secondary reference may disagree because both references are individually idiosyncratic.","Transient fixture loading may be mistaken for persistent learning.","Thermal coupling or shared supply variation may create spurious apparent agreement."]},"next_evidence_step":"On a benchtop, use one small characterized reference array, one sacrificial adjustable array, one untouched open-loop baseline, and a fixed low-energy probe set divided in advance into coupling and held-out probes. Record direct electrical responses before coupling, after each fixed hardware-decoupling stage, immediately after full disconnection, and after one bounded unpowered retention interval. Proceed beyond feasibility only if held-out standalone mismatch decreases without clamp events, excess heating, oscillation, or software correction.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed against other proposals because runtime isolation forbids inspecting them; this candidate is derived solely from the supplied archetype and computer-science domain card.","revision_record":{"parent_version":null,"progress_targets_addressed":["Construct one causally explicit candidate under the binding physical-substrate constraint","Preserve mentor anchoring, modeled practice, corrective interaction, autonomy release, and plural-reference safeguards","State serious rivals, counterevidence, falsifiers, and a bounded evidence step"],"conceptual_changes":["Recast relational enculturation as direct physical transfer of a trusted analog computing array's realized response to a variable apprentice array."],"operational_changes":["Specified paired excitation, differential error currents, hard-limited programming pulses, staged decoupling, and a secondary physical reference."],"evidence_changes":["Defined pre-coupling, held-out, standalone, and retention measurements against an untouched baseline."],"claim_changes":["Limited the claim to a testable contrast with open-loop nominal programming and explicitly withheld novelty, prevalence, demand, and effect-size claims."]},"substrate_contract":{"primary_allowed_process":"PHYSICAL_MATERIAL","counterfactual_independence":"The essential effect is persistent alteration of adjustable conductance states by electrically generated programming charge. After transfer, the reference array, comparison fixture, software, algorithms, reports, permissions, incentives, and procedures can all be removed; the new array must still exhibit the transferred response under direct electrical measurement. If it works only while coupled or while digitally corrected, the intervention fails.","forbidden_channel_audit":"No model, database, dashboard, recommender, information-routing system, software controller, policy, incentive, authorization decision, training program, or human reporting loop computes or applies the corrective action. Probe generation may be fixed and measurements may be logged for evidence, but logging is not connected to programming. Mentor selection and safety approval bound the experiment without causing convergence. The operative pathway is shared physical excitation, analog current differencing, bounded electrical pulse transfer, and persistent material-state change."}}}