{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp06_four_proposal_generalization60_20260803","cell_id":"bounded_rivalry_governance__human_computer_interaction","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"brg_hci_accessibility_discovery_portfolio_004","proposal_index":4,"version":0,"title":"Accessibility Barrier Discovery Portfolio Challenge","problem":"An organization offers a fixed reward pool for external testers to find accessibility barriers in a digital service before release. When payment and recognition favor the first valid report or the largest issue count, testers can improve their standing by flooding the queue with automated findings, reserving likely issues without complete evidence, splitting one barrier into several reports, withholding reproduction details, or probing live systems for an advantage. Rivalry can broaden discovery, but unmanaged rivalry can select fast claimants rather than reports that reveal distinct, reproducible barriers and support repair.","actors":["Disabled people whose interaction requirements the testing is intended to represent","Independent accessibility testers","Testing teams and their disclosed affiliates","Accessibility program owner","Product designers and engineers responsible for remediation","Security and privacy reviewers","Independent report-verification panel","Maintainers operating the test environment","Appeal reviewer separate from initial scoring"],"observable_state":"A fixed bounty pool is depleted sequentially; the report queue contains near-duplicate scanner output, incomplete reproduction steps, fragmented versions of one underlying barrier, and disputes over who reported first; testers use unequal automation or test traffic; live or personal data can enter evidence attachments; easy-to-count defects dominate while end-to-end task failures remain weakly documented; and repeat winners gain early access or procedural knowledge unavailable to challengers.","consequence":"Maintainers spend the review period deduplicating and reconstructing findings, the reward portfolio can concentrate on redundant or readily automated issues, task-blocking barriers can remain uncharacterized, unsafe testing can expose data or degrade the service, and procedural advantage can become more important than accessibility contribution.","affected_objective":"Discover a complementary set of reproducible accessibility barriers across important user journeys and interaction modes while producing safe evidence that maintainers can use to verify and repair each barrier.","intervention":"Replace first-to-file bounty allocation with a bounded, batch-scored portfolio challenge conducted in a synthetic-data test environment. Before each round, the sponsor publishes the discovery purpose, fixed reward pool, eligible service scope, user journeys, assistive-technology and input-mode coverage goals, legal tests, prohibited actions, submission cap, evidence format, scoring rubric, conflicts, tie-breaks, verification process, and appeals. Qualified individuals or disclosed teams receive equal test-window access and scoped accounts. Submissions remain sealed until the batch closes, so filing minutes earlier does not determine payment. Every report must identify the affected task, starting state, interaction sequence, observed barrier, expected accessible behavior, reproducible evidence, environment, remediation-relevant trace, and any security or privacy concern. Privacy, authorization, fixture integrity, and reproducibility are noncompensable gates. Verified reports are scored for observed task obstruction, evidentiary clarity, reproducibility across the declared environment, distinctness from other findings, and usefulness of the remediation handoff. Awards are then selected as a portfolio that covers nonredundant journeys, interaction modes, and failure classes within the fixed pool rather than by taking reports in queue order. Submission and automated-request caps limit flooding. Pseudonymous standings and scoring rationales are released after verification; accepted evidence becomes shared with maintainers and, where safe, other entrants. Duplicate identities, fixture alteration, report fragmentation, interference with other testers, concealed coordination, unsafe scanning, and fabricated evidence are predefined fouls with graduated consequences. New qualified entrants may join later rounds, and a post-round review examines missed barriers, tester burden, reward concentration, gaming, exported harms, and whether the contest should be revised or replaced by a noncompetitive audit.","structural_mapping":[{"archetype_element":"Rivalry purpose statement","domain_realization":"Use competition to draw out independently discovered accessibility barriers and strong reproduction evidence, not to maximize report volume or reward filing speed."},{"archetype_element":"Scarce prize or selection constraint","domain_realization":"A fixed round-specific reward pool supports only a bounded portfolio of awards, so verified reports compete for inclusion."},{"archetype_element":"Competitor eligibility boundary","domain_realization":"Entrants must disclose team and affiliate identities, accept the safe-testing terms, use scoped accounts, and demonstrate that they can submit evidence in the required accessible format."},{"archetype_element":"Contest arena boundary","domain_realization":"Entrants may inspect and exercise the designated synthetic-data service but may not probe production, access other testers' accounts, alter shared fixtures, degrade availability, collect personal data, conceal joint authorship, or split one barrier into artificial reports."},{"archetype_element":"Performance metric and scoring basis","domain_realization":"Reports pass authorization, privacy, fixture-integrity, and reproducibility gates, then receive declared scores for task obstruction, evidence clarity, cross-environment reproduction, distinctness, and remediation handoff."},{"archetype_element":"Fair process and due process layer","domain_realization":"Rules and coverage goals are frozen before access, submissions are batch-sealed, verification is separated from entrants, scoring rationales are disclosed, and factual or procedural errors may be appealed to a separate reviewer."},{"archetype_element":"Anti-sabotage and anti-collusion guardrail","domain_realization":"Immutable access and submission logs support investigation of fixture interference, duplicate identities, coordinated fragmentation, reciprocal report ownership, or sham independence; screens create referrals rather than verdicts."},{"archetype_element":"Externality and spillover boundary","domain_realization":"Synthetic data, scoped accounts, request limits, stop conditions, and noncompensable privacy and security gates keep discovery costs from being exported to service users, other testers, or production maintainers."},{"archetype_element":"Escalation and arms-race damper","domain_realization":"Each entrant or disclosed team receives the same test window, submission ceiling, automated-request allowance, account set, and evidence-storage quota."},{"archetype_element":"Prize decomposition or multiple-winner design","domain_realization":"The fixed reward is divided among a complementary portfolio of verified findings rather than assigned to one highest-count tester or exhausted in filing order."},{"archetype_element":"Winner power and lock-in review","domain_realization":"Winning does not grant exclusive access to test fixtures or future rounds; accepted evidence enters the shared remediation record, identities are aggregated across affiliates, and entry reopens under published criteria."},{"archetype_element":"Learning and recalibration loop","domain_realization":"Post-round review examines unrepresented journeys, rejected and duplicate reports, safety events, award concentration, circumvention, repair utility, and participant feedback before revising or retiring the next round."}],"mechanism_mapping":[{"mechanism_slug":"prize_challenge","role":"Posts bounded accessibility-discovery goals and a fixed reward paid only for verified findings that satisfy predeclared evidence and safety conditions.","counterfactual_removal":"Without outcome-contingent awards, the intervention would become an ordinary commissioned audit and would no longer use rival discovery effort to surface alternatives from a qualified field."},{"mechanism_slug":"contest_rulebook","role":"Freezes eligibility, testing scope, legal moves, coverage goals, evidence requirements, scoring, tie-breaks, fouls, disclosure, and appeals before entrants inspect the test environment.","counterfactual_removal":"Without the rulebook, the sponsor could favor known testers, reinterpret duplicate or severity standards after seeing submissions, or enforce unsafe-testing boundaries selectively."},{"mechanism_slug":"multiple_award_or_portfolio_selection","role":"Selects the set of reports for complementary coverage across tasks, interaction modes, and failure classes rather than treating each report's isolated score as sufficient.","counterfactual_removal":"Without portfolio selection, the pool could be consumed by several high-scoring descriptions of similar barriers while distinct but less easily counted interaction failures receive no award."},{"mechanism_slug":"ranked_leaderboard_with_audit","role":"Publishes pseudonymous standings after verification and audits leading reports by reproducing their interaction path, inspecting evidence provenance, and checking claimed distinctness.","counterfactual_removal":"Without verification of leaders, fabricated traces, environment-specific artifacts, or strategically fragmented findings could outrank reproducible accessibility evidence."},{"mechanism_slug":"spending_cap_or_resource_cap","role":"Caps submissions, automated requests, scoped accounts, test time, and evidence storage per entrant or disclosed team.","counterfactual_removal":"Without caps, the field could become an automation and traffic race in which well-resourced entrants fill the queue and burden the service before manual or assistive-technology testing is completed."},{"mechanism_slug":"sabotage_or_foul_penalty_schedule","role":"Defines graduated responses to fixture alteration, unauthorized scanning, identity duplication, evidence fabrication, harassment, report fragmentation, and interference with another tester.","counterfactual_removal":"Without predefined fouls and consequences, damaging the test environment or manufacturing more apparent findings could remain a viable way to capture the finite pool."},{"mechanism_slug":"anti_collusion_monitoring","role":"Pools identity, timing, content-similarity, and award patterns across rounds to flag concealed joint authorship, reciprocal report allocation, sham duplicates, or coordinated division of coverage for investigation.","counterfactual_removal":"Without cross-round monitoring, affiliated entrants could appear independent, rotate awards, or divide one finding among identities in patterns not visible from a single report."},{"mechanism_slug":"challenger_access_window","role":"Reopens qualification and test access for each new round while preventing previous winners from controlling shared fixtures or eligibility.","counterfactual_removal":"Without recurring access windows, early winners could convert familiarity and privileged test access into durable ownership of the discovery arena."},{"mechanism_slug":"post_contest_impact_review","role":"Checks whether awarded findings supported reproduction and remediation, what barrier classes were missed, what harms or burdens escaped the arena, and whether reward concentration or gaming distorted the result.","counterfactual_removal":"Without post-round review, a rubric that rewards reportability rather than meaningful barrier discovery could be repeated after its mismatch becomes observable."}],"causal_chain":["A fixed reward pool and recognition make accessibility testers strategically dependent rivals rather than independent reporters.","Sealed batch submission removes filing speed as the automatic route to payment while retaining pressure to discover and document strong findings.","A common synthetic environment, equal access, and resource caps reduce advantages from production access, automation scale, and queue flooding.","Noncompensable authorization, privacy, integrity, and reproducibility gates exclude findings produced through exported harm or unverifiable evidence.","Audited report scoring couples individual standing to observable task barriers and usable reproduction evidence.","Portfolio selection makes complementary coverage, rather than redundant issue accumulation, the route by which the field captures the finite reward pool.","Foul enforcement and cross-round monitoring make identity splitting, sabotage, fragmentation, and concealed coordination contestable without treating statistical patterns as guilt.","Shared accepted evidence and recurring challenger entry prevent an award from granting permanent control of fixtures, knowledge, or future eligibility.","Post-round comparison with remediation work and missed-barrier review reveals whether the arena continues to serve accessibility discovery and supplies grounds to revise or retire it."],"baseline":"A conventional accessibility bounty accepts reports continuously and pays the first valid submission according to an issue classification until the budget is exhausted. Reports are reviewed sequentially, duplicate disputes are resolved after filing, automated and manual submissions share the same queue, entrant resource use is weakly bounded, and awards are assessed individually rather than as a coverage portfolio.","nearest_rivals":["Contracted accessibility audit: can provide disciplined expert evaluation and clear accountability but uses commissioned work rather than governed rivalry among independent discoverers.","Participatory accessibility testing: grounds evaluation in disabled participants' experiences but is primarily collaborative research and does not govern competition for a fixed discovery reward.","Automated accessibility scanning: produces repeatable checks at scale but cannot by itself establish end-to-end task obstruction, manage rival reporting behavior, or assemble a complementary human-evidence portfolio.","Conventional first-valid-report bug bounty: uses competitive discovery and outcome payments but rewards timing at the report level and does not inherently cap resource escalation or optimize the awarded set for accessibility coverage.","Internal heuristic review and usability testing: can identify interaction problems within a planned study but does not create an open, contestable field with scarce awards and recurring challenger access."],"remaining_contrastive_claim":"Even if a conventional bounty adds safe-harbor terms, duplicate checks, severity labels, and reproduction audits, the remaining distinction is governance of the reward portfolio as a whole: sealed batch entry, equal resource ceilings, noncompensable harm gates, complementarity-based multiple awards, affiliate aggregation, cross-round abuse screens, appeals, and reopening jointly make distinct accessibility contribution the route to winning. If every valid report can be rewarded without scarcity, testers cannot affect one another's outcomes, or competitive discovery adds no useful independent coverage, the proposal collapses into ordinary accessibility testing or commissioned audit.","authority_safety":{"decision_authority":"The accessibility program owner may authorize a challenge round only after the product owner, security lead, privacy lead, and test-environment maintainer approve its scope and stop conditions. An independent verification panel may recommend awards, while a separate reviewer decides appeals. Production operators retain control of all live systems.","authorized_first_step":"Run one nonmonetary mock round against an isolated prototype containing synthetic accounts and a preregistered set of seeded and unseeded accessibility barriers. Invite a bounded group of consented testers, issue fictitious points instead of payments, and test the submission, portfolio-scoring, verification, explanation, appeal, and foul-detection procedures.","excluded_actions":["Testing production services, live accounts, or real personal data","Running denial-of-service, credential, social-engineering, or destructive tests","Contacting uninvolved service users or using them as test participants","Paying or ranking a report that fails authorization, privacy, fixture-integrity, or reproducibility gates","Publishing exploit details, participant identities, or sensitive traces before remediation approval","Treating automated similarity or collusion screens as proof without investigation and appeal","Using challenge standings in employment, contracting retaliation, or public shaming","Allowing prior winners to control entrant eligibility, verification fixtures, or appeal decisions","Changing the scoring rubric, pool allocation, or coverage goals after submissions are opened"],"halt_rollback":"Halt the round if any request reaches production, real personal or credential data appears, the test environment becomes accessible outside the approved scope, a shared fixture is altered, a security issue presents risk beyond safe reproduction, participant harassment occurs, or scoring materials are exposed before the batch closes. Disable challenge accounts, stop the test gateway, preserve audit records, quarantine sensitive submissions, restore the environment from its clean image, notify the responsible reviewers, and resume only after independent verification of the corrected scope."},"negative_tests":{"strongest_counterevidence":"A noncompetitive paid panel using the same tasks and environment produces equally distinct and repair-usable findings without queue flooding, concealment, interference, or resource escalation, while competitive incentives add only administrative burden. That would indicate that structured accessibility evaluation, not rivalry governance, is the operative solution.","problem_falsifier":"There is no scarce reward, status, access, or selection opportunity; every valid report can be accepted independently; testers cannot strategically affect one another's outcomes; or the sponsor needs cooperative longitudinal research rather than competitive discovery.","intervention_falsifier":"Batch portfolio selection still rewards redundant or readily automated findings, reproducibility scoring is inconsistent across reviewers, caps disproportionately exclude manual or assistive-technology investigation, sealing delays urgent disclosure, affiliate aggregation is evaded, unsafe testing moves outside the sandbox, or accepted reports do not provide maintainers with usable verification evidence.","risks":["Portfolio scoring can encode the sponsor's incomplete assumptions about which journeys or interaction modes matter.","A submission cap may cause entrants to withhold uncertain findings that would have been useful when combined with other evidence.","Batch sealing can delay communication of a barrier or security issue requiring immediate attention.","Equal automated-request limits do not equalize prior tooling, expertise, hardware, or access to assistive technology.","Pseudonymous standings can still create status competition or expose identities through writing style and specialized evidence.","Complementarity can become a pretext for rejecting a strong report after the pool is informally reserved for favored categories.","Manual verification can reproduce evaluator bias or fail when assistive-technology configurations differ.","Identity aggregation can mistakenly combine independent testers or fail to detect coordinated affiliates.","Testers may optimize for the published rubric and avoid barriers that are difficult to document or remediate.","Synthetic journeys may omit contextual, longitudinal, or socially mediated accessibility barriers.","Publishing accepted evidence can disclose attack paths or enable copycat submissions in later rounds.","Disabled participants may be treated as sources of competitive evidence without appropriate compensation, consent, or authority if the program expands beyond qualified testers."]},"next_evidence_step":"Conduct a preregistered two-week mock round in one isolated web-service prototype with synthetic data, four user journeys, a declared interaction-mode coverage matrix, six consented testers or teams, equal scoped accounts, and fictitious points. Seed several reproducible barriers, leave additional interface states unseeded, and keep the seed register hidden from entrants and initial scorers. Compare filing-order selection with the proposed sealed portfolio selection using the same reports. Record duplicate and fragmentation decisions, independently reproduced findings, scorer agreement, coverage selected, request-cap exceptions, unsafe-action attempts, time-sensitive disclosure handling, explanation comprehension, appeals, and missed seed categories. Red-team identity splitting, automated flooding, reciprocal submissions, fixture alteration, fabricated traces, and rubric-targeted report splitting. Use the exercise only to assess arena coherence, safety, and observable selection differences, not to infer production effectiveness.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 governs internal interface-design teams competing for one default AI-assisted review interface and awards a reversible deployment pilot after common usability evaluation. Proposal 4 instead governs external or cross-organizational testers competing to discover barriers in an existing service and distributes a fixed portfolio of evidence rewards; it neither chooses an interface design nor grants deployment control. Proposal 2 governs continuous runtime bidding by applications for a user's interruptive attention through publisher credits and user-weighted notification slots. Proposal 4 has no attention allocation, notification issuer, recurring display slot, or user-priority auction; its causal path runs through sealed discovery, reproduction gates, submission caps, and complementary report selection. Proposal 3 governs operational teams competing among incident hypotheses for one expiring production command lease, using safety gates, sandbox rehearsal, and prediction-triggered reopening. Proposal 4 grants no production command authority and does not serialize remediation actions; it evaluates evidence about interface barriers in an isolated test environment and pays several complementary findings. The challenge can be adopted as an accessibility assurance program without changing default-interface selection, notification infrastructure, or incident-command procedure, making it independently adoptable from all three earlier proposals.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}