{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp05_complete_proposal_portfolio20_20260803","cell_id":"computability_boundary_mapping__economics_finance","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_id":"cbm-ef-runtime-trading-conduct-shield-005","proposal_index":5,"version":0,"title":"Runtime Conduct Shield for Unrestricted Trading Strategies","problem":"A securities exchange or broker gateway is asked to certify, before deployment, whether an arbitrary executable trading strategy will ever emit an order sequence forbidden by a machine-specifiable conduct rule, for every admissible market-data stream and execution history. Strategies may contain unbounded loops, recursion, persistent state, or adaptive logic. Backtests and code review cover only selected executions, while a timeout or absence of a discovered violation can be converted into a SAFE_STRATEGY label. For an unrestricted strategy language, a proposed total analyzer must distinguish a strategy that never emits a forbidden event from one that simulates an arbitrary computation and emits the event only if that computation halts.","actors":["Exchange conduct-rule owners","Exchange and broker order-gateway operators","Algorithmic trading firms and strategy developers","Market-surveillance analysts","Risk-control and market-operations staff","Market participants exposed to accepted orders","Independent formal-methods reviewer"],"observable_state":"A strategy passes sampled replays and receives permission to send orders directly to the market. The analyzer either did not explore a later adaptive branch or stopped at its resource limit. When an unseen input activates that branch, the gateway has no mechanism that checks the next order against the formal rule; the earlier pre-deployment label is expected to stand in for runtime enforcement.","consequence":"A forbidden machine-detectable order prefix can reach the market because incomplete pre-deployment evidence was treated as a universal behavioral guarantee, while strategies that merely resist analysis can be rejected without a proved violation.","affected_objective":"Correct enforcement of formally specified trading-conduct safety rules without claiming that arbitrary strategy programs can be universally classified from finite tests or bounded analysis.","intervention":"Replace universal strategy-safety certification with complete runtime mediation for a declared fragment of conduct rules. First freeze the strategy execution model and the ex-ante property, then construct and independently check a proposed reduction mapping a source computation to a strategy that emits a designated forbidden order event if and only if the source halts. Until that reduction and its assumptions pass review, label the unrestricted ex-ante status UNRESOLVED rather than undecidable. Separately define an enforceable finite-state rule language over normalized order events such as submission, modification, cancellation, account, instrument, and gateway state. Compile each in-fragment safety rule into a terminating reference monitor. Every outbound event updates finite monitor state; an event entering a forbidden state is blocked with a checkable prefix witness, and any permitted event advances atomically. The shipped guarantee is only that no event accepted through complete mediation violates the versioned monitored safety rule—not that the underlying strategy is benign, lawful in every respect, or will never attempt a violation. Rules involving intent, unobservable cross-venue conduct, future outcomes, or liveness remain OUT_OF_FRAGMENT. Gateway bypass, missing state, stale rule versions, and monitor failure produce MONITOR_FAILURE and fail closed under an approved continuity procedure rather than silently allowing orders.","structural_mapping":[{"archetype_element":"Unsupported total analysis of arbitrary programs","domain_realization":"The exchange seeks a correct terminating verdict about every future execution of unrestricted adaptive trading code."},{"archetype_element":"Constructive and impossibility evidence separated","domain_realization":"A checked reduction evaluates the unrestricted pre-certification claim, while a separate constructive proof establishes total per-event monitoring for finite-state safety rules."},{"archetype_element":"Language restriction applied to the guarantee","domain_realization":"The conduct-rule language, rather than the trading-strategy language, is restricted to safety properties recognizable from finite order prefixes."},{"archetype_element":"Weaker fallback with semantic fidelity","domain_realization":"The system guarantees that mediated accepted events satisfy the monitored rule, not that the strategy is universally safe or free of unformalized misconduct."},{"archetype_element":"Explicit scope and status behavior","domain_realization":"MONITORED_ALLOWED, BLOCKED_WITH_WITNESS, RULE_OUT_OF_FRAGMENT, UNRESOLVED_ANALYSIS, and MONITOR_FAILURE remain distinct."},{"archetype_element":"Termination contract","domain_realization":"Finite monitor state and a total transition function make each order-event decision terminate under the declared event schema."},{"archetype_element":"Recheck on semantic or architectural drift","domain_realization":"Changes to event normalization, conduct rules, account linkage, gateway routing, monitor state, or bypass paths invalidate the earlier guarantee."}],"mechanism_mapping":[{"mechanism_slug":"halting_problem_reduction","role":"Tests the feasibility of the unrestricted pre-deployment analyzer by mapping a source program to a trading strategy that emits a forbidden event exactly when the source program halts.","counterfactual_removal":"Without the checked mapping, failed static analysis and difficult strategies would not justify retiring the universal analyzer requirement."},{"mechanism_slug":"language_fragment_restriction","role":"Limits machine-enforced rules to a syntactically checkable finite-state safety fragment whose violations have finite order-prefix witnesses.","counterfactual_removal":"Without the rule-language boundary, intent, future effects, and other non-monitorable properties could inherit a runtime guarantee the gateway cannot enforce."},{"mechanism_slug":"constructive_algorithm_and_correctness_proof","role":"Supplies the finite-state reference monitor and proves that it terminates on every well-formed event, blocks every transition into a forbidden state, and permits exactly the transitions allowed by the encoded rule.","counterfactual_removal":"Without a total-correct monitor witness, runtime mediation would be another implementation claim supported only by tests."},{"mechanism_slug":"proof_checking","role":"Independently verifies the reduction, monitor termination and safety invariant, rule compilation, and correspondence between normalized events and gateway behavior.","counterfactual_removal":"Without independent checking, a reduction error or mismatch between the formal monitor and actual gateway could support an invalid boundary or enforcement guarantee."},{"mechanism_slug":"bounded_domain_exhaustive_search","role":"Exhaustively checks every event sequence within a small pilot alphabet and trace bound to test the compiler, monitor transitions, witnesses, and state restoration behavior.","counterfactual_removal":"Without complete bounded testing, the pilot could miss a transition-table or state-synchronization defect while still appearing well covered."},{"mechanism_slug":"fallback_mode_router","role":"Routes in-fragment safety rules to runtime mediation, out-of-fragment concerns to surveillance or human review, and infrastructure failures to the approved fail-closed continuity path while preserving each label.","counterfactual_removal":"Without explicit routing, an out-of-fragment rule or failed monitor could be silently treated as satisfied."},{"mechanism_slug":"computability_boundary_decision_record","role":"Records the rejected or unresolved universal claim, exact runtime guarantee, rule fragment, proof artifacts, complete-mediation assumptions, failure policy, and recheck triggers.","counterfactual_removal":"Without the record, later rule or gateway changes could leave an obsolete runtime guarantee attached to a materially different enforcement system."}],"causal_chain":["Formalizing the ex-ante claim exposes its universal quantification over unrestricted strategies and market inputs.","A valid checked source-to-strategy reduction can rule out the matching total analyzer; if the reduction fails, the unrestricted status remains unresolved rather than being decided by timeout.","Restricting the conduct-rule language to finite-prefix safety properties creates a mechanically enforceable boundary without restricting how strategies internally compute.","The constructive monitor checks each actual outbound event under a total finite-state transition function.","Complete mediation ensures that an accepted order prefix cannot cross the monitored forbidden boundary, regardless of whether the strategy was analyzable in advance.","A blocked transition carries the concrete event prefix and rule state that justify the action, while permitted events receive no broader SAFE_STRATEGY label.","Intent-based, future-dependent, and unobservable cross-venue properties route to separate surveillance or review instead of being laundered through the monitor.","Gateway failures and missing state follow an explicit continuity rule rather than silently bypassing enforcement.","Versioned proof and architecture records cause changes in rules, event semantics, or routing completeness to trigger re-verification."],"baseline":"Evaluation baseline: strategy code review, sampled historical or synthetic replays, bounded static analysis, and post-event surveillance. Admission produces a Boolean permission to trade, while the order gateway relies on that pre-deployment assessment and general operational limits rather than evaluating every event against the specific formal safety rule.","nearest_rivals":["Ban customer-supplied strategies and permit only fixed exchange-approved templates.","Increase the number and duration of strategy backtests and adversarial market simulations.","Require strategy developers to submit proofs of future rule compliance before deployment.","Use sound static analysis that may reject strategies when it cannot eliminate a possible violation.","Rely on post-trade surveillance, cancellation, sanctions, and restitution after a prohibited event is observed.","Apply generic order-rate and price limits without compiling the specific conduct rule into the gateway."],"remaining_contrastive_claim":"The intervention changes the object of the guarantee from an arbitrary program's unbounded future behavior to the finite sequence of events the gateway actually accepts. Template restrictions, proof-carrying submissions, static analysis, and expanded simulation remain pre-deployment judgments about strategies; post-trade surveillance reacts after acceptance. Complete runtime mediation instead enforces a narrowly specified safety invariant on each observable event while explicitly declining to certify intent, liveness, future market effects, or the strategy as a whole.","authority_safety":{"decision_authority":"The exchange or broker conduct-and-market-operations authority retains control over rule definitions, trading permissions, blocks, suspensions, and continuity policy. The formal-methods team may validate monitor artifacts but cannot create conduct rules or authorize production enforcement.","authorized_first_step":"The head of market-operations risk may authorize a mirror-only offline gateway exercise using synthetic strategies and synthetic order events. The monitor records the decisions it would have made but cannot transmit, modify, cancel, delay, or block a live order.","excluded_actions":["Transmit, modify, cancel, delay, or block live orders","Approve, suspend, restrict, or terminate a trading participant or strategy","Change conduct rules, price collars, account linkages, or market parameters","Label an underlying strategy SAFE because its emitted pilot events were allowed","Treat RULE_OUT_OF_FRAGMENT, UNRESOLVED_ANALYSIS, or MONITOR_FAILURE as rule compliance","Infer legal intent or unformalized manipulation from a finite-state block witness","Enable production fail-closed behavior before continuity and recovery procedures are separately approved","Reuse monitor proofs after an unreviewed event-schema, gateway, compiler, or routing change"],"halt_rollback":"Halt the exercise if the monitor permits a synthetic prefix that the same versioned rule marks forbidden, blocks a prefix the rule permits, loses state synchronization, accepts traffic through an unmediated path, produces a witness inconsistent with its decision, or drops result labels downstream. Withdraw the affected monitor version and return the gateway to its unchanged pre-pilot configuration; mirror-only operation causes no live order action to reverse."},"negative_tests":{"strongest_counterevidence":"The strongest counterevidence to the unrestricted-boundary claim would be a checked specification showing that every admissible strategy is already finite-state with a computable global execution bound and that every market-input domain is finite and enforced. A total exact pre-deployment analyzer could then exist for the declared class, leaving complexity and architecture as the relevant questions.","problem_falsifier":"The problem is falsified if no universal SAFE_STRATEGY claim is made from bounded analysis and every relevant outbound order already passes through a verified complete-mediation monitor whose rule fragment, witnesses, failure states, and limits are preserved downstream.","intervention_falsifier":"The intervention fails if the target conduct requirement cannot be expressed as a finite-prefix safety property, the gateway cannot observe all events or state needed by the rule, any bypass path exists, or an accepted in-scope trace violates the compiled rule. It also fails semantic fidelity if users interpret a monitored-event guarantee as evidence about trader intent or compliance with out-of-fragment obligations.","risks":["The formal rule may omit account, venue, timing, or beneficial-owner context needed for the intended conduct constraint.","Event normalization may differ between the monitor and matching engine.","State races, replayed messages, or recovery procedures may break atomic transition assumptions.","Fail-closed behavior may interrupt permitted trading during monitor or state-service failure.","Fail-open recovery would invalidate the complete-mediation guarantee.","Per-event checking may introduce operational latency or capacity constraints.","A strategy may distribute behavior across accounts or venues outside the monitor's observable boundary.","Rule compilation may be incorrect even if the reference monitor is sound.","Operators may treat BLOCKED_WITH_WITNESS as proof of intent rather than a mechanical rule event.","Out-of-fragment conduct concerns may receive less attention because the runtime shield appears authoritative.","A valid impossibility result for unrestricted programs may be overstated as applying to bounded strategy languages."]},"next_evidence_step":"Conduct one mirror-only exercise with an event alphabet of at most five normalized event types, traces of at most six events for exhaustive testing, four synthetic finite-state conduct rules, and eight synthetic strategies including a divergent strategy, an adaptive strategy, an explicit forbidden-prefix generator, and a strategy that never attempts a violation. Produce the unrestricted-class reduction artifact, rule grammar, compiler, reference monitor, termination and safety proof, exhaustive trace table, block-witness checker, state-loss and gateway-bypass tests, and boundary decision record. An independent reviewer must verify reduction direction, compilation fidelity, acceptance invariants, witness correctness, complete pilot coverage, and preservation of out-of-fragment and failure labels. No live order, participant account, market-data feed, or enforcement decision enters the exercise.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Proposal 1 addressed pre-listing solvency analysis for programmable derivatives through a bounded contract fragment, exhaustive or sound abstract analysis, and guarantee-labelled routing. Proposal 2 addressed truthfulness and termination of executable auction mechanisms through proof-carrying admission and certificate-versus-counterexample search. Proposal 3 addressed finite intraday collateral assignment by constructing a total exhaustive decider and separating decidability from operational complexity. Proposal 4 addressed macroeconomic policy-model conclusions that depend on external solvers or humans through oracle-relative passports, promise enforcement, and capability-withdrawal testing. Proposal 5 addresses machine-specifiable conduct constraints on outbound trading events and replaces whole-strategy prediction with complete runtime mediation of finite-prefix safety rules. It restricts the enforceable property rather than the strategy, provides a per-event enforcement guarantee rather than a listing certificate, allocation verdict, or model passport, and can be adopted independently at an exchange or broker order gateway.","revision_record":{"parent_version":null,"progress_targets_addressed":[],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}