{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","cell_id":"bounded_rivalry_governance__mathematics","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"math_bounded_compute_window_auction_v0","proposal_index":4,"version":0,"title":"Bounded Credit Auction for Scarce Mathematical Computation Windows","problem":"A mathematics institute operates a shared system capable of long, uninterrupted exact computations for activities such as exhaustive case analysis, symbolic derivation, and counterexample search. When two or more research teams request the same indivisible compute window, an informal queue gives each team incentives to overstate urgency, reserve more capacity than its job can use, split requests across nominal projects, or negotiate reciprocal priority. Administrators cannot directly compare the prospective mathematical value of unrelated jobs, while unused reservations, overruns, storage, and operator interventions impose costs on rival teams.","actors":["Mathematical research teams requesting uninterrupted computation","Facility operator controlling the shared system","Eligibility reviewer assessing whether jobs are executable and within scope","Auction administrator separated from eligibility review","Independent process and appeal reviewer","Operations staff handling checkpoints, overruns, and cleanup","Research teams displaced or delayed by a winning job","Authorized anti-collusion monitor"],"observable_state":"For each allocation round, the facility can observe registered team identity, eligibility disposition, requested resource envelope, sealed bid in nontransferable credits, clearing debit, declared urgency rank, job readiness, reservation start and release times, utilization, checkpoints, cancellation, overrun, storage retained, operator interventions, standby activation, rule violations, appeals, and delayed aggregate bidding patterns. Mathematical content beyond the minimum confidential eligibility description and private team communications remains outside routine monitoring.","consequence":"Scarce uninterrupted windows can be reserved by teams best able to navigate or manipulate the queue rather than by teams willing to sacrifice their own future access for the current job. Capacity may sit idle or overrun, operational burdens may be shifted to other users, reciprocal bidding arrangements may make the apparent contest noncompetitive, and administrators may become de facto judges of incomparable mathematical research claims.","affected_objective":"Allocate each indivisible uninterrupted computation window through an observable and contestable process that elicits teams' relative time priority without requiring administrators to rank mathematical importance, while bounding reservation waste, operational spillovers, collusion, and repeated dominance.","intervention":"Replace discretionary priority claims for the specified uninterrupted windows with repeated sealed-bid auctions using nonpurchasable, nontransferable facility credits. The facility freezes a rulebook before each allocation season. Stable registered teams receive a disclosed credit endowment and may bid for an upcoming window only after an independent eligibility check confirms that the job is within mathematical scope, ready to run, checkpointable where technically possible, and described by a bounded resource envelope. The highest qualified bid wins and is debited the second-highest qualified bid, with a predeclared lottery for exact ties. A separate, precomputed surcharge covers declared storage, operator attention, restart risk, and other measured spillovers. The winner also posts a refundable credit bond against avoidable overrun, nonuse, and cleanup. Bids remain sealed; only delayed aggregate results are disclosed. A standby bidder receives the window if the winner releases it before the cutoff. Multiple identities, bid sharing, side payments, deliberate nonuse, false resource declarations, tampering, and retaliation are prohibited. Fixed screens examine cross-round patterns for possible bid rotation or cover bidding, but flags only initiate an independent inquiry. Credits, surcharges, caps, and screens are recalibrated or the auction is retired after a scheduled impact review.","structural_mapping":[{"archetype_element":"Rivalry purpose statement","domain_realization":"Use rivalry to reveal how strongly each team prioritizes the current computation relative to its own future access, without asking a central panel to compare the mathematical promise of unrelated research programs."},{"archetype_element":"Scarce prize or selection constraint","domain_realization":"One indivisible, uninterrupted facility-controlled computation window in each allocation round."},{"archetype_element":"Competitor eligibility boundary","domain_realization":"A stable registered team may bid only for an executable mathematical job with a declared resource envelope, responsible operator, checkpoint or abort plan, and no unresolved facility-safety violation."},{"archetype_element":"Contest arena boundary","domain_realization":"Teams may choose bids, conserve credits, revise jobs before the deadline, withdraw, or release a won window. They may not create sham identities, exchange bids, make side payments, reserve without a ready job, falsify resource needs, tamper with rival jobs, or retaliate against other bidders."},{"archetype_element":"Performance metric and scoring basis","domain_realization":"Among qualified jobs, the sealed credit bid determines rank; measured externality surcharges and bond adjustments remain separate from the bid so a nominally high valuation cannot erase operational obligations."},{"archetype_element":"Fair process and due process layer","domain_realization":"Published eligibility rules, identical bid deadlines, sealed submissions, deterministic clearing rules, recorded decisions, a predeclared tie procedure, and an independent time-boxed appeal bind teams and administrators."},{"archetype_element":"Anti-sabotage and anti-collusion guardrail","domain_realization":"Access controls, stable team identities, bid confidentiality, delayed aggregate disclosure, operational logs, fixed cross-round screens, and a separate investigation process address tampering, bid rotation, cover bids, and side agreements."},{"archetype_element":"Externality and spillover boundary","domain_realization":"Storage, operator intervention, restart risk, avoidable overrun, cleanup, and nonuse are assigned predeclared credit charges or secured by a refundable bond rather than shifted to later users."},{"archetype_element":"Escalation and arms-race damper","domain_realization":"Credits cannot be purchased or transferred, each registered team faces a seasonal bid ceiling, and project splitting does not create additional endowment."},{"archetype_element":"Winner power and lock-in review","domain_realization":"A win grants only the specified window; it creates no permanent queue priority, additional voting power, or authority over future eligibility and auction rules."},{"archetype_element":"Learning and recalibration loop","domain_realization":"Periodic review compares auction behavior with utilization, overruns, standby transfers, spillovers, concentration, appeals, and investigation outcomes before parameters are renewed."}],"mechanism_mapping":[{"mechanism_slug":"auction_with_eligibility_and_externality_rules","role":"Allocates each window through a sealed second-price credit bid after job qualification, then adds separately specified spillover charges.","counterfactual_removal":"Without the governed auction, administrators would have to rank incomparable mathematical narratives or revert to arrival time, while teams would not express current priority through an explicit sacrifice of future facility access."},{"mechanism_slug":"contest_rulebook","role":"Freezes team identity rules, eligibility, bidding, clearing, ties, standby transfer, disclosure, prohibited conduct, appeals, and bond treatment before bids are visible.","counterfactual_removal":"Without the rulebook, administrators could alter qualifications or clearing treatment after learning team identities, bids, or preferred projects."},{"mechanism_slug":"externality_bond_or_liability_rule","role":"Requires a refundable credit bond and predeclared charges for avoidable nonuse, overrun, retained storage, cleanup, and operator intervention.","counterfactual_removal":"Without secured liability, a winner could bid aggressively and leave later users or facility staff to absorb the operational costs after its credits were spent."},{"mechanism_slug":"anti_collusion_monitoring","role":"Applies the same structural and behavioral screens across delayed cross-round data to identify possible bid rotation, stable share division, or cover bidding and refer anomalies for independent inquiry.","counterfactual_removal":"Without cross-round monitoring, reciprocal priority agreements could make repeated auctions appear competitive even when teams had privately divided the windows."},{"mechanism_slug":"spending_cap_or_resource_cap","role":"Uses nonpurchasable credits, a seasonal bid ceiling, stable team identities, and aggregation of related projects to constrain repeated dominance and credit arms races.","counterfactual_removal":"Without the cap, teams able to obtain additional credits or manufacture identities could dominate successive windows independently of their opportunity cost."},{"mechanism_slug":"sabotage_or_foul_penalty_schedule","role":"Publishes graduated responses to negligent resource misstatement, repeated nonuse, sham identities, bid sharing, side payments, retaliation, and facility tampering, subject to notice and review.","counterfactual_removal":"Without predeclared sanctions, prohibited conduct could become a tolerable cost for some teams or be punished selectively through administrator discretion."},{"mechanism_slug":"post_contest_impact_review","role":"Tests whether the arena's rules still serve the allocation purpose by examining realized utilization, burdens, concentration, false flags, strategic adaptation, and user appeals.","counterfactual_removal":"Without ex-post review, poorly calibrated credits, charges, caps, or screens could persist even after teams learned to circumvent them or operational conditions changed."}],"causal_chain":["Indivisible overlapping requests make teams rivals for the same computation window.","A readiness and resource-envelope gate excludes requests that cannot use the prize as specified without asking reviewers to predict mathematical importance.","A finite nontransferable credit budget makes a high current bid consume the same team's ability to win later rounds.","Sealed second-price clearing provides a fixed comparative allocation rule while limiting opportunities for real-time signaling among bidders.","Separating the externality surcharge from the competitive bid prevents teams from winning by understating storage, staff, restart, or cleanup burdens.","The refundable bond preserves recourse when actual nonuse or avoidable overrun differs from the declared envelope.","Seasonal caps and stable identities reduce the value of purchasing dominance, splitting projects, or escalating bids through nominal teams.","Uniform cross-round screens can surface patterns consistent with false competition, while an independent inquiry prevents a statistical flag from becoming a verdict.","A one-window award and standby transfer prevent victory from creating durable control or leaving released capacity idle by rule.","Post-round operational evidence supports recalibration or abandonment if bidding ceases to track the arena's stated purpose."],"baseline":"Teams submit narrative priority requests to facility administrators, who combine arrival time, asserted urgency, relationships, and informal negotiation to schedule uninterrupted jobs. Reservations, overruns, project identity, unused time, externality charges, appeals, and recurring reciprocal arrangements are handled case by case.","nearest_rivals":["First-ready, first-served queue: the earliest executable job receives the next window. This is simple and avoids bidding administration, but arrival time can become the target and does not express the opportunity cost of displacing other ready jobs.","Independent scientific peer review: a panel ranks requests by expected mathematical value and feasibility. It can incorporate substantive judgment but requires comparisons among uncertain and unrelated research claims and may favor familiar fields or investigators.","Equal round-robin quotas: registered teams receive windows on a fixed rotation. This protects access and is difficult to manipulate after identity is fixed, but it cannot respond when one team's current job is urgent and another has no ready use.","Lottery among qualified jobs: every ready request receives an equal or weighted chance. This limits lobbying and fine-grained gaming but does not let teams trade present access against their own future opportunities.","Preemptible shared scheduling: all jobs run concurrently or in short slices under technical scheduling rules. This is preferable if jobs can be divided and checkpointed without invalidating the mathematical computation, eliminating the indivisible prize that motivates the auction."],"remaining_contrastive_claim":"The candidate's testable structural distinction is to govern repeated rivalry through a sealed internal-credit auction whose result is conditioned on readiness, externality charges, secured cleanup duties, anti-collusion referral, and recurrent reopening. It does not assert that auction allocation is preferable to peer review, rotation, lottery, first-ready scheduling, or technical sharing for a particular facility.","authority_safety":{"decision_authority":"The facility operator may control access to its own hardware, issue nonmonetary facility credits, enforce operational limits, and record resource metadata. Separate reviewers may decide eligibility, process appeals, or investigate a specific flag. They may not judge the truth or importance of the resulting mathematics, control publication or attribution, impose employment consequences, purchase access to private communications, or regulate computation performed on resources outside the facility.","authorized_first_step":"Authorize only a no-stakes shadow auction using scripted and voluntarily supplied deidentified job cards. Shadow bids must not change the live queue, consume real credits, determine funding, or produce public team rankings.","excluded_actions":["Selling credits or permitting credit transfers, side markets, or monetary bids","Changing endowments, eligibility, surcharges, caps, or clearing rules after bids or team identities are visible","Treating a statistical collusion flag as proof or punishment without an independent inquiry","Inspecting private research communications or undisclosed mathematical content through general surveillance","Using auction outcomes in hiring, promotion, publication, attribution, or research-quality assessment","Allowing the winner to alter rival jobs, future eligibility, credit issuance, monitoring thresholds, or appeal decisions","Charging a team for facility faults, ordinary uncertainty, or harms outside the predeclared liability boundary","Interrupting a running job outside the agreed safety, overrun, or emergency conditions","Creating additional compute load merely to make a shadow test consequential"],"halt_rollback":"Pause a round after bid leakage, identity duplication, unequal submission access, clearing error, conflicted review, unsafe resource declarations, or a monitor defect capable of changing an outcome. Preserve the sealed record, restore the window to the pre-auction queue or an already qualified standby under the frozen rule, and appoint an unconflicted reviewer. A live operational emergency permits safe termination of the job but does not establish misconduct or automatic bond forfeiture."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence would be that first-ready scheduling, round-robin access, or independent peer review achieves at least as much usable window time with fewer overruns and lower administrative burden, while credit bidding tracks team size, risk tolerance, or strategic hoarding rather than declared time priority. Evidence that repeated bidding suppresses legitimate collaboration or produces stable reciprocal allocation would also weigh strongly against the intervention.","problem_falsifier":"The inferred problem is falsified if eligible requests rarely overlap, windows are divisible or safely preemptible, unused capacity is readily reassigned, administrators do not compare strategic urgency claims, teams cannot affect one another's access, or existing quotas already bound reservation waste and repeated dominance without meaningful contest behavior.","intervention_falsifier":"The intervention is falsified for its intended mechanism if shadow bids are unstable under inconsequential framing changes, teams create effective side transfers, project aggregation cannot prevent identity splitting, externality schedules systematically misclassify facility costs, scripted innocent bidding repeatedly triggers collusion screens, scripted bid rotation escapes every screen, or auction administration consumes more scarce staff capacity than the allocation dispute it replaces.","risks":["Credit bids may reflect team size, patience, or risk preference rather than mathematical urgency.","Repeated second-price auctions can still invite bid shading, signaling, retaliation, or strategic budget hoarding.","Stable team registration may entrench incumbents or misclassify genuine new collaborations as identity splitting.","Externality surcharges may deter legitimate resource-intensive mathematics.","A bond may punish uncertain runtimes even when uncertainty was disclosed honestly.","Eligibility descriptions and operational metadata may expose confidential research direction.","Delayed aggregate disclosure may still enable coordination in a small repeated field.","Anti-collusion screens can mistake common deadlines or similar job cycles for coordination.","Seasonal caps may leave capacity unused when capped teams have the only ready jobs.","A standby rule may encourage low-cost speculative bids if qualification is weak.","Teams may move workloads outside the governed facility, shifting rather than resolving operational harms.","Auction administration may become a new source of favoritism through eligibility, team-definition, or surcharge decisions."]},"next_evidence_step":"Pre-register one no-stakes shadow exercise spanning ten scripted allocation rounds. Use deidentified job cards with known readiness, urgency ordering, resource envelopes, overruns, releases, innocent correlated demand, identity-splitting attempts, and bid-rotation scripts. Give each simulated team the same nontransferable credit budget and compare the auction outcome with first-ready, round-robin, and lottery allocations without changing any live schedule. Record rule ambiguities, clearing reproducibility, unused windows, externality recovery, standby transfers, scripted abuses detected, innocent patterns flagged, reviewer time, and whether simulated teams' bids contradict their preregistered priority ordering. Stop before any live pilot if clearing cannot be independently reproduced, an unsafe or unready job can win, bond treatment depends on undisclosed discretion, innocent scenarios repeatedly generate unresolved collusion allegations, scripted coordination remains operationally indistinguishable from competition, or administration exceeds its preset staff-time ceiling.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 governs rival proof routes seeking one formal-verification slot; its causal path uses proof reproduction, correctness gating, route-quality scoring, and staged formalization. Proposal 4 instead governs research teams bidding for a temporary computation window without comparing the correctness or prospective importance of their mathematics; its path uses nontransferable opportunity-cost credits, sealed price clearing, externality charges, bonds, standby transfer, and cross-round collusion screens. Proposal 2 governs foundational interfaces competing for persistent default status in a shared library; its hazards are semantic migration loss, rule control, and network-effect lock-in, addressed through interoperability trials, secured exit duties, competition review, and challenger access. Proposal 4 awards no standard or durable default: every computation window is consumed and the arena resets. Proposal 3 governs authors competing for six live problem-set slots; its path uses confidential validation, pilot solving, grading audits, and portfolio complementarity to address leakage, ambiguity, and redundancy. Proposal 4 uses repeated single-window auctions rather than portfolio judging and does not create or evaluate contest problems. It is independently adoptable by a shared mathematical computing facility without adopting any proof-route, foundation-interface, or problem-set intervention.","revision_record":{"parent_version":null,"progress_targets_addressed":["Initial complete formulation for proposal index 4","Material separation from sealed proposal indices 1, 2, and 3"],"conceptual_changes":["Instantiated bounded rivalry as repeated bidding for indivisible mathematical computation windows.","Used internal opportunity-cost revelation rather than mathematical merit scoring as the allocation logic.","Located the main spillovers in nonuse, overrun, storage, operator intervention, and queue displacement."],"operational_changes":["Specified stable eligibility identities, sealed second-price credit clearing, externality surcharges, a refundable bond, standby transfer, seasonal caps, delayed disclosure, anomaly referral, appeals, and rollback.","Limited the first evidence step to a scripted no-stakes shadow auction that cannot alter the live queue."],"evidence_changes":["Defined observable bid, readiness, utilization, spillover, concentration, investigation, and process records.","Specified abuse and innocent-correlation scenarios plus explicit stop conditions.","Kept prior art unsearched and made no novelty, prevalence, demand, or effect-size claim."],"claim_changes":["Restricted the claim to the structural behavior of a governed internal-credit auction.","Explicitly retained queueing, peer review, rotation, lottery, and preemptible sharing as potentially preferable alternatives."]}}