The Conflictoring protocol

Take a concrete decision. The treasurer of Silicon Valley Bank keeps a large, concentrated, unhedged portfolio of long-duration bonds because hedging would dent this quarter's earnings. Before committing, the system-welfare question — Field 17 — asks for the consequence beyond the two parties: if interest rates rise, the unhedged concentration threatens the bank's solvency and its depositors, a system cost, not just a private one. Recording that honestly is not free. It costs the treasurer κ: a contemporaneous account now proves he saw the depositor risk and accepted it for a quarter of earnings, and that account can be used against him. The single-audience trap is Prat's conformism. Suppose the only reader he expects is the internal model-validation team, whose standards he knows well. He writes the account to please that team — compliant, reassuring, the systemic risk buried under caveats the validators habitually accept. This is Andrea Prat's result exactly: when a decision-maker knows who is judging, transparency induces conformism, because to that single known judge non-conformism looks like incompetence. With one manageable reader the detection probability p is low, pL falls below κ, and shading the account is his best response. He shades it, the risk stays invisible, and the bank fails — the Silicon Valley Bank outcome¹. The Conflictoring² fix replaces the single judge with several whose objectives cannot be jointly satisfied. The board wants the private payoff maximized; the plaintiff bar wants proof that system welfare was not knowingly destroyed; a whistleblowing risk analyst wants documentation honesty; a regulator in the next examination wants the risk disclosed. These are not the same standard at different strengths — they are structurally opposed. The very account that satisfies the board — maximize the private payoff, keep the negative system-welfare effect out of sight — is precisely the account that hands the plaintiff bar its case and the whistleblower a cause. There is no single story that pleases incompatible judges at once. And because he cannot predict which of them will pull which document, he cannot even optimize for one. Two forces do the work: the audiences' objectives are adversarial, so pleasing one activates another; and the audience is uncertain, so no account is safe. Change the architecture from one known judge to many incompatible unknown ones with divergent objectives, and the equilibrium flips from conformism to candor. Read as pL ≥ κ: p is the combined chance that at least one of the seven Conflictoring lanes — an employee or whistleblower, a CEO or board, a plaintiff litigator, a shareholder, a regulator, a policymaker, or a community or member of the affected public — surfaces a false account. Because their objectives conflict, the account that hides from one exposes it to another, so p stays high. L is whatever the biting channel imposes: under the European Union's Digital Operational Resilience Act³ and the United Kingdom's Senior Managers and Certification Regime⁴, a personal fine and possibly criminal liability on the named senior manager; in litigation, damages and a fraud finding; from the board, termination and clawback; from the market, a repriced stock and a ruined career. When that expected penalty pL exceeds the cost κ of telling the truth, honest disclosure is the decision-maker's best response. The crown jewel is the architecture, not the arithmetic. Prat proved that pointing transparency at one known judge makes things worse. The Conflictoring architecture does not restrict the information or optimize the contract — every prior remedy kept the single-audience architecture and so kept the conformism. It changes who the audience is: many principals whose objectives cannot all be met at once. That is why “any of them, and you cannot tell which” is not a weakness — it is the mechanism.

The transparency trap the example turns on is Andrea Prat's. In The Wrong Kind of Transparency (American Economic Review, 2005)⁵, Prat proved that making an agent's actions transparent to a single principal induces conformism rather than candor: the agent ignores private signals and performs the competence the one watcher expects. That result has stood for roughly two decades as a central obstacle in accountability theory, and every proposed remedy — reducing observability, optimal disclosure, mandatory disclosure to one regulator — keeps the single-audience architecture and so keeps the pathology. The Conflictoring resolution is not “add more readers.” Multiple audiences with aligned objectives collapse back to a single effective judge, and Prat's conformism returns — the common-agency result of Bernheim and Whinston⁶. The escape requires audiences whose objectives are structurally incompatible, where the record that satisfies one arms another. That condition — divergent, adversarial objectives, not mere plurality — is what dissolves the trap, and the Conflictoring mechanism shows that under stated activation and payoff conditions it does. This is offered as a candidate resolution of Prat's problem, pending peer review — not a settled result.

Chapter 2 places Pigou, Coase, and Ostrom as three switches on one defect — the exclusion of system welfare from the payoff space — each with a failure mode: the Pigovian planner fails under capture, when capture intensity φ exceeds the threshold φ*; Coasean bargaining fails when no party owns the support system; Ostromian governance fails as institutions erode. Set those three switches against the seven Conflictoring lanes. The classical trio occupies only two lanes cleanly: the Pigovian imposed price⁷ is the regulator-and-policymaker lane, and Ostrom's self-governing community⁸ is the community lane. Coase straddles⁹ — it runs through whoever defines and enforces the property right, the regulator and the courts. The other lanes — the whistleblower, the board, the plaintiff litigator, the shareholder — sit outside the Pigou-Coase-Ostrom externality trio. Outside that trio is not the same as undiscovered, and honesty requires the distinction. Each of these channels has its own literature, and several are already framed in externality terms: private liability as decentralized internalization of harm is the native language of the economics of tort (Calabresi; Landes and Posner; Shavell)¹⁰; the diversified shareholder as an internalizer of portfolio-wide externalities is the universal-owner hypothesis (Hawley and Williams)¹¹; the employee as the actual detector of corporate fraud is documented empirically (Dyck, Morse, and Zingales)¹². Pluralist accounts of enforcement beyond the state also exist — Yandle's bootlegger-and-baptist coalitions¹³, Ayres and Braithwaite's tripartism¹⁴, Gunningham and Grabosky's surrogate regulators¹⁵. The framework is not the first to notice that correction runs through many hands. What it adds is a formalization, not a discovery, and it is threefold. First, a defined target: a system-welfare coordinate W that is off both transacting parties' payoff ledgers by construction — the Missing System Theory¹⁶ — which the governance-studies typologies, being catalogues of regulatory strategy rather than mechanisms indexed to a welfare coordinate, do not supply. Second, a coverage claim: that these seven lanes jointly span the ways that coordinate can be forced into the open. Third, the mechanism — divergent, adversarial audiences and the unpredictable reader, so no single account survives all of them, with pressure concentrated until the coalition reaches k^∗, which is also what makes the whole capture-robust when the Pigovian regulator is captured or the Ostrom community has eroded. One lane needs its tension named. The board's instrument — competitors jointly petitioning a regulator for industry-wide rules, lawful under the Noerr-Pennington doctrine¹⁷ — is, in the mainstream literature, the very signature of capture and rent-seeking (Stigler; Yandle)¹⁸. The claim that the same joint petitioning can serve system welfare, by binding every rival to a standard no firm can afford to meet alone, is a reading that must be defended against that skeptical default, not assumed. The more than thirty cases documented later populate all seven lanes. In one line: these are harms off the payoff ledger by construction — the Missing System Theory — that the imposed price, the tradable boundary, and the self-governing community reach for only part of, and that a wider set of heterogeneous, adversarial lanes resolve in practice; the framework's contribution is to define that coordinate and unify the lanes that correct it, not to claim the lanes were unknown. Offered as a positioning claim for scrutiny, pending peer review, not a settled result.

The escape from Prat's trap has a name: Conflictoring. It replaces the single known judge with many independent, adversarial audiences whose objectives cannot be jointly satisfied, so no single account survives all of them and candor becomes the writer's best response. The account the treasurer wrote in the example — the system-welfare impact, Field 17, captured at the moment of decision — is what a firm keeps when it practices Decision Accounting. In time, that standardized record will grow into the bedrock of Conflictoring: a clean, contemporaneous account written when the decision is made, the ideal input for the lanes to act on. But adoption is years off. It waits on the theory surviving peer review, on standard-setters and courts recognizing the record, and on organizations building it into how they govern. Conflictoring does not wait. In the meantime it is implemented by the seven lanes acting on what already exists: the ordinary paper trail — emails, memos, filings, minutes — and the lane's own estimate of the system-welfare impact, Field 17 computed as a single coordinate from the published β_W library, without the firm's cooperation. Estimating that one coordinate is not adopting Decision Accounting; it is the one number a lane needs to act. The rest of this chapter is that near-term implementation: how the seven lanes, working today, concentrate enough pressure to dissolve a Hollow Win.

The seven lanes have already changed systems for the better, and none of it waited on Decision Accounting. Whistleblowers and regulators exposed the manipulation of the LIBOR interest-rate benchmark and forced its replacement by SOFR, the Secured Overnight Financing Rate¹⁹. International regulatory action under the Montreal Protocol phased out the chemicals that were thinning the ozone layer, which is now recovering²⁰; on the harder problem of greenhouse gases, the Paris Agreement now coordinates national emissions commitments²¹, a live mechanism whose result is not yet settled. Dodd-Frank and its whistleblower program were built after the 2008 crisis to raise trust in the financial system, and by many measures did²². Each of these is a Conflictoring mechanism already at work — an employee, a regulator, a policymaker, a litigator, the affected public imposing a consequence — and none of them required Decision Accounting as a prerequisite. What Decision Accounting adds is not the mechanism but its input. Today the lanes reconstruct a decision from fragments after the harm is done: emails, trading logs, meeting minutes, subpoenaed files. A Decision Accounting record would instead give them a contemporaneous account written at the moment of decision — the system-welfare consequence priced, the alternatives rejected, the prediction on the record — rather than a reconstruction assembled years later. That standardized record is the forward assumption in this chapter. It does not yet exist for most organizations, because Decision Accounting has not been adopted. The lanes work now; the record that would sharpen them may lie ahead. We also have direct evidence on what makes such a record bite. The strongest to date comes from Standardize the Record, Not the Reader: Decision Records for Consumer AI Agent Markets²³, which tests the mechanism on AI pricing agents rather than people, inside a reproduced Calvano, Calzolari, Denicolò, and Pastorello (2020)²⁴ algorithmic-collusion environment. Filing a decision record alone left the agents' conduct near baseline — a null result — even as their forecasts improved. Conduct changed only when an authorized reader could attach a consequence, and the effect grew with audit exposure. A record is inert until it faces readers who can impose a cost — which is exactly what the LIBOR, ozone, and Dodd-Frank lanes already do. The paper frames this as a practical path for Prat's conformism problem — standardize the record, preserve reader diversity, and let heterogeneous readers attach consequences — the multi-audience, divergent-objectives mechanism this chapter turns on. What we do not yet have is a human-based study; those are expensive and slow, and Decision Accounting is a novel concept, so the AI-agent evidence stands in for now.

None of the following required Decision Accounting. Each is an instance of one or more of the seven lanes repairing a shared system, drawn from the public record. The list is a sample, not a census. Employees and whistleblowers • Jeffrey Wigand's disclosures on nicotine manipulation (1996) helped states win the tobacco settlement.²⁵ • Sherron Watkins' internal memo (2001) exposed Enron's accounting fraud and helped drive Sarbanes-Oxley.²⁶ • Cynthia Cooper's audit team (2002) uncovered roughly $3.8 billion in fraud at WorldCom.²⁷ • Frances Haugen's leaked research (2021) forced congressional scrutiny of platform harm to teenagers.²⁸ • The SEC and CFTC whistleblower program (2010 onward) has paid over $2 billion in awards and recovered billions for harmed investors.²⁹ Boards and executives • Johnson & Johnson recalled 31 million bottles after the 1982 Tylenol tampering and made tamper-evident packaging an industry norm.³⁰ • Maple Leaf Foods recalled 220 products and publicly accepted responsibility during a 2008 listeria outbreak.³¹ • Perrier voluntarily pulled about 160 million bottles worldwide over benzene traces (1990).³² Plaintiff litigators • The Tobacco Master Settlement (1998) forced at least $206 billion in payments and ended cartoon and billboard cigarette marketing.³³ • The Johns Manville asbestos trust (1988) has paid over $5 billion to more than a million victims.³⁴ • Anderson v. PG&E (1996) won $333 million for residents poisoned by chromium-6 in their groundwater.³⁵ • The national opioid settlement (2021-22) secured $26 billion for addiction treatment and abatement.³⁶ • The Flint water settlement (2021) delivered $626 million to residents harmed by lead-contaminated tap water.³⁷ Shareholders • Engine No. 1 won three ExxonMobil board seats (2021) to force climate and capital discipline.³⁸ • The Dodd-Frank say-on-pay rule (2011) gave shareholders a standing advisory vote on executive pay.³⁹ • The Boardroom Accountability Project (2015) drove more than 100 firms to adopt proxy access for long-term owners.⁴⁰ • The CalPERS Focus List (1987 onward) publicly named poorly governed firms and produced measurable governance gains.⁴¹ Regulators • The EPA phase-out of leaded gasoline (final rule 1996) drove a fall of over 90% in children's blood-lead levels.⁴² • CFPB and OCC action fined Wells Fargo $185 million (2016) over roughly two million unauthorized accounts.⁴³ • EPA and DOJ enforcement over Volkswagen's defeat devices produced a settlement of about $14.7 billion (2016).⁴⁴ • Regulator settlements over LIBOR rigging (2012-2023) drove the switch to transaction-based SOFR.⁴⁵ • The Siemens foreign-bribery case (2008) imposed about $1.6 billion and reset global anti-corruption compliance norms.⁴⁶ • The Deepwater Horizon settlement (2016) imposed about $20.8 billion and funded Gulf restoration.⁴⁷ • After two crashes, FAA and congressional action overhauled Boeing 737 MAX certification oversight (2018-2020).⁴⁸ Policymakers • The Montreal Protocol (1987) phased out about 99% of ozone-depleting substances; the ozone layer is now recovering.⁴⁹ • The Paris Agreement (2015) established five-year national climate commitments and a global stocktake.⁵⁰ • The Clean Air Act Acid Rain Program (1990) cut power-plant sulfur-dioxide emissions about 94% through cap-and-trade.⁵¹ • CERCLA and the Superfund (1980) created polluter-pays liability to clean up abandoned hazardous-waste sites.⁵² • Sarbanes-Oxley (2002) strengthened auditor independence and executive certification of financials after Enron.⁵³ • Basel III (2010) raised bank capital and liquidity standards to make banking more resilient after 2008.⁵⁴ • The Kimberley Process (2003) cut conflict diamonds to under about 1% of world production.⁵⁵ • The Post Office (Horizon System) Offences Act (2024) quashed hundreds of wrongful UK convictions caused by faulty accounting software.⁵⁶ Communities and the affected public • Lois Gibbs' Love Canal campaign (1978-80) forced federal action and directly spurred the Superfund law.⁵⁷ • The Panama Papers (2016) triggered resignations and more than 150 investigations across 79 countries.⁵⁸ • Reporting and employee accounts exposed the Theranos fraud (2015), ending the company.⁵⁹ • The Rana Plaza collapse (2013) produced a legally binding Bangladesh factory-safety accord signed by more than 200 brands.⁶⁰ • The dolphin-safe tuna boycott (1988-90) forced the three largest brands to change sourcing, later codified in law.⁶¹

A practitioner working inside an organization that is producing a Hollow Win can observe the records, the approvals, and the compliance sign-offs that do not change the outcome. Both parties to the transaction gain privately while the system they depend on degrades. This is not an accident of bad management: it is the structural exclusion the Missing System Theory identifies (Chapter 1). The standard regulatory remedy is endogenous to the disease, because the single regulatory node is itself a capturable target of the incumbent. The game's payoff space does not carry the coordinate that would register, and therefore stop, the damage.

The Conflictoring protocol is an eight-step procedure for diagnosing and escaping Hollow Win situations. It uses an intent-free detection standard: it does not require proof that anyone intended to harm the system. If the outcome is classified as a Hollow Win (0,1,1), the protocol applies. The first five steps are diagnostic. Steps six through eight are resolution. The protocol is designed so that one uncaptured, privately motivated mover can force the game to carry W.

The first step confirms that the transaction or practice in question is a Hollow Win. The test has two parts: do both parties to the transaction gain privately, and does the system that supports the transaction degrade as a result? When both answers are yes, the outcome is a Hollow Win. The system-welfare coordinate C equals 0; the private coordinates A and B equal 1; the outcome is (0,1,1). This is not a judgment about anyone's intent. It is a structural classification.

The second step establishes the decision's impact on system welfare. Be clear about what this requires. It does not require the firm to adopt Decision Accounting or to keep a seventeen-field record. It requires only that the lane pursuing the Conflictoring claim estimate a single coordinate — the system-welfare impact, Field 17 — for the decision in question. Field 17 is one field; the other sixteen fields of a full record are not needed, and computing this one coordinate is not implementing Decision Accounting. The lane can estimate the impact from the domain's β_W — welfare destroyed per dollar of annual industry revenue — in the published library, without the firm's cooperation. The estimate should be documented so that an independent reader can check it: the affected system named, its boundary, the evidence, and the uncertainty. Putting that one missing coordinate on the table, where the lanes can act on it, is the whole of step two.

The third step assesses which capture technologies the incumbent controls and which lane can act at the lowest total prevention cost. The Conflictoring mechanism identifies seven implementation lanes: employees or whistleblowers, CEOs and boards, plaintiff litigators, shareholders, regulators, policymakers, and communities and the affected public. Each faces a different capture technology, and the incumbent's lobbying budget does not control all of them equally. The regulator is the most capturable single node in the standard single-planner remedy, under the model's assumptions, because the incumbent can shape the regulator's information and incentives through the same influence technology that produces the Hollow Win. Employees or whistleblowers and plaintiff litigators face different capture technologies, because they are not single nodes that can be neutralized by lobbying expenditure. The Calabresi-style question (after legal scholar Guido Calabresi's cheapest-cost-avoider rule) is practical: which lane can most cheaply prevent, reduce, expose, insure, regulate, litigate, or reprice the harm? The paper models capture resistance as a foreclosure problem with convex suppression costs (blocking rivals from the market, where each added unit of suppression costs more than the last).

The fourth step identifies which actor has a private incentive to move first. The Conflictoring mechanism does not rely on collective benevolence. It relies on independently profitable unilateral movers. Employees and whistleblowers have a financial incentive under applicable federal programs. Plaintiff litigators have a damages claim. Shareholders have an alignment incentive from disclosure and restoration repricing. CEOs and boards have a competitive-parity incentive: doing the right thing alone raises their own costs, so instead they can lead a coordinated petition for industry-wide regulation — lawful under the Noerr-Pennington doctrine, which immunizes concerted petitioning of government from antitrust liability — that binds every rival at once and removes the first-mover disadvantage. Regulators have a jurisdictional incentive. Policymakers have a constituency incentive. Communities and the affected public bear the harm directly and can mobilize where the loss is concentrated. The protocol works when at least one of these actors has a private return from moving that exceeds the private cost of staying silent.

Calabresi's cheapest-cost-avoider logic belongs upstream. Tort law asks after an injury who could have prevented or reduced the harm at the lowest cost. Conflictoring asks that question before the harm hardens and across the seven lanes: employees and whistleblowers, CEOs and boards, regulators, policymakers, shareholders, plaintiff litigators with lawful access in the relevant setting, and communities and the affected public. The question becomes operational: which actor can most cheaply prevent, reduce, expose, insure, regulate, litigate, or reprice the harm? The answer may differ by domain. Employees may see hidden facts first. Boards may change strategy fastest. Shareholders may reprice capital. Regulators may alter standards. Policymakers may change statutes or procurement rules. Plaintiff litigators may create liability where the old game hid the cost. Conflictoring uses that distribution of powers to make the Hollow Win privately expensive enough to change.

The fifth step calculates whether the protocol will bite. In the Decision Accounting framing, κ was the cost of truthfully recording the impact — the exposure an honest record creates. Without a firm keeping that record, there is no recording cost to weigh; what deters the incumbent is the sanction on the conduct itself. The condition is pL ≥ κ, where p is the probability the lanes detect and act on the harm, L is the liability they can impose, and κ is the incumbent's private gain from the Hollow Win — what it pockets by degrading the system. When pL ≥ κ, the expected sanction exceeds that private gain, so continuing the destructive game costs more than reforming, and changing conduct becomes the incumbent's best response. The detection probability p is not fixed: it rises with the number and heterogeneity of the Conflictoring lanes, and the unpredictable-reader property makes it hard to game. The mechanism does not require p to approach one. It turns on k^∗, the minimum coalition size: the smallest set of the seven independent lanes that must activate together to weaken or escape the Hollow Win under the stated conditions. A low k^∗, one or two lanes, means the game is cheap to change; a high k^∗ means reform needs coordinated pressure across several lanes. This is how Conflictoring makes the destructive game more expensive than reform: by reaching k^∗, not by driving detection to certainty.

The sixth step files the claim with the applicable authority. The Conflictoring protocol specifies that the claim should reach the actor with the strongest private incentive and cheapest practical path to act on it. For employees and whistleblowers, this means filing under the applicable federal program. For plaintiff litigators, it means filing a civil claim. For shareholders, it means filing a disclosure demand. For regulators or policymakers, it means using the administrative or legislative channel that can change the payoff structure. The claim should include the system-welfare impact — the Field 17 estimate from step two, with its boundary, evidence, and uncertainty — not a full Decision Accounting record. It should be filed before the investigation concludes, because the detection probability p is highest when the Conflictoring actor is the first mover.

Step seven concentrates the pressure. Its goal is to bring the independent lanes to bear on the same conduct at the same time, so their power lands together rather than split across months and forums where an incumbent can absorb each blow separately. This step does not depend on Decision Accounting; it needs only the seven lanes — an employee, a board, a litigator, a shareholder, a regulator, a policymaker, and the affected public — converging at once. Simultaneous coalescence is the point: it turns scattered, sequential complaints that a well-resourced incumbent can pick off one by one into a single review that no single actor controls. That concentrated review is capture-resistant for two structural reasons, neither of which requires the lanes to merely pile up. First, their objectives are incompatible: the account that satisfies one audience is the very account that arms another — what placates a board exposes it to the plaintiff bar and the whistleblower — so there is no single story that survives all of them at once. Second, the mover cannot predict which lane will examine which part, and cannot silence lanes it does not control. Each additional independent, adversarial lane raises the detection probability p that the conduct is exposed. When enough of them activate together — the coalition reaching k^∗, the minimum number that must move as one — the pressure is concentrated past the point the incumbent can suppress, and continuing the destructive game costs more than reforming it.

The eighth step collects the return from the protocol. The Conflictoring mechanism is designed so that the first mover receives a private return that exceeds the private cost of moving. For employees and whistleblowers, the return comes from the applicable federal award program; for the plaintiff litigators, from damages; for the shareholders, from the restoration repricing of the asset. The return is not charity. It is the mechanism's way of making the first move privately profitable. The Calabresi-style test asks whether that actor is also the cheapest practical cost avoider: the lane that can prevent, reduce, expose, insure, regulate, litigate, or reprice the harm at the lowest total cost. The paper shows that when the expected return exceeds the expected cost, the first mover has a dominant strategy (the best choice regardless of what others do) to move. This is how the protocol solves the collective action problem (a benefit spread thinly across many, so each has too little stake to act) that keeps Hollow Wins in place.