The Behavioral Blind Spot in Accountability: Decision Provenance, Hollow Wins, and the Missing System Theorem

Accountability fails when records cannot reconstruct the decision

The paper argues that overconfidence, present bias, and loss aversion matter less than a missing information structure. Accountability requires decision provenance: who acted, what evidence was available, which alternatives were rejected, and when the choice was executed.

• Behavioral explanations enter after the record has already decayed
• S(D) = actor, evidence, rejected alternatives, time is the minimum provenance structure
• Without S(D), assessors cannot separate good-faith error from negligent choice

SVB disclosed the risk but did not preserve the accountable choice chain

Silicon Valley Bank collapsed in March 2023 after a 48-hour run, destroying 209 billion in market value and requiring emergency Federal Reserve intervention. The paper uses SVB to show that disclosure is not provenance: the 91 billion held-to-maturity concentration was visible, but the decision chain around maintaining it was not.

• Missing record: specific meeting, evidence presented, alternatives rejected, and predicted system-welfare impact
• Fed post-mortem named supervisory insufficiency but could not reconstruct the decision provenance
• The failure was not that the risk was hidden; it was that the accountable chain was not preserved

The bilateral accountability game cannot see system welfare

In SAPM, the accountability game has three coordinates: A is the decision-maker, B is the monitor, and C is the system. Principal-agent models focus on A and B, so a system-damaging outcome can still appear as a win-win inside the bilateral frame.

• G1 = A, B, C : decision-maker, monitor, system
• Bilateral G(A,B) sees four outcomes; adding C expands the taxonomy to eight
• The Hollow Win is (C=0, A=1, B=1): both parties win while the system degrades

Conflictoring makes under-disclosure rational for both regulated parties

The paper names Conflictoring as the structure in which formally adversarial parties align around under-disclosure. A manager can receive a bonus while an auditor or regulator receives a clean report, even though unpriced risk, externalized cost, or institutional erosion accumulates in C.

• From inside G(A,B), Hollow Win (0,1,1) looks like Win-Win (1,1)
• System-welfare impact is not a required field in traditional reporting
• The omitted coordinate makes (0,1,1) indistinguishable from (1,1,1)

β_W measures unpriced system damage per revenue dollar

The paper defines β_W = -dW/dΠ, where W is system welfare and Π is annual industry revenue. The denominator must be revenue, not profit; using profit inflates β_W by 5–20x in low-margin industries.

• Global financial services revenue: about $12.5 trillion annually
• Amortized crisis cost: about $1.33 trillion per year over a 15-year cycle
• Financial-sector β_W ≈ 0.10, or 0.10 of unpriced system risk per 1 of revenue

High-risk domains make the Hollow Win measurable

The paper uses β_W to compare domains where bilateral payoffs omit systemic costs. Financial services has β_W ≈ 0.10 overall, but the ratio exceeds 5.0 in shadow banking and high-use crypto-asset domains. PFAS manufacturing is bounded at β_W ≥ 71.4 in the formal theorem.

• Shadow banking and high-use crypto-assets: β_W > 5.0
• PFAS manufacturing: 28 billion revenue against estimated 2 trillion cleanup costs
• The β_W lower bound is β_W ≥ (Σᵢ Rᵢ) / Π

Provenance preservation explains more failure variance than bias prevalence

The Behavioral Blind Spot Theorem states that accountability failure varies mainly with preservation of S(D), not with the prevalence of individual cognitive bias. The paper formalizes this as Var(Accountability Failure) = α·Var(S(D)) + β·Var(B) + ε, with α greater than β by at least one order of magnitude.

• Unstructured records decay over time without contemporaneous preservation
• Post-hoc narratives import hindsight bias, self-serving attribution, and outcome bias
• Aviation black boxes, nuclear operations, and Estonia's X-Road have 2–3 orders lower accountability failure rates

The Hollow Win is a strict Nash equilibrium when W is excluded

Under the Bilateral Payoff Exclusion Axiom and Record-Equilibrium Coupling Axiom, the Hollow Win is stable. A maximizes private payoff while ignoring W; B approves actions that satisfy bilateral compliance while ignoring W. Deviating imposes cost without improving either bilateral payoff.

• A2: W is not an argument in u_A or u_B unless imposed from outside
• A3: equilibrium selection depends on the information structure available at choice time
• Iterated elimination of dominated strategies converges on (0,1,1) under I that excludes W

MCAS certification converted avoided training and redesign into β_W = 2.36

Boeing and the FAA achieved a Hollow Win / management failure when the 737 MAX was certified with MCAS as a single-channel system requiring no additional pilot training. Boeing gained program revenue and avoided certification costs; the FAA preserved certification throughput and delegated authority.

• Program revenue used in the paper: $37 billion for 2013–2019
• System-welfare loss: $87.3 billion, including crashes, grounding costs, fines, compensation, market-value loss, and human-capital loss
• Two crashes killed 346 people: Lion Air Flight 610 and Ethiopian Airlines Flight 302

Unstructured records decay before accountability can act

The Provenance Decay Axiom says reconstructibility falls over time when no structured contemporaneous record exists. For typical corporate decisions, the paper bounds decay by the mean tenure of decision participants, at no more than 3–5 years.

• Decay accelerates with turnover, restructuring, litigation risk, and narrative reconstruction
• Low psychological safety reduces record completeness below the Five-Minute Test threshold
• The later inquiry receives S'(D), not the original S(D)

Decision Accounting makes system welfare a recorded payoff coordinate

The paper's remedy is mandatory Decision Provenance through a 17-field Decision Accounting protocol. Field 17, SYSTEM_WELFARE , requires a contemporaneous estimate of W, changing G = A,B into G' = A,B,C .

• The record must capture actor, evidence, rejected alternatives, time, and system-welfare impact
• The monitor is evaluated on record integrity and W accounting, not only bilateral compliance
• With enforceable Field 17 penalties, the Hollow Win becomes strictly dominated

The paper estimates $73.8T per year from fixing the institutional constraint

The Reform Dividend Axiom treats the gain as an opportunity-cost lower bound, not a point forecast. The estimate comes from observed welfare losses across 49–52 confirmed domains with zero counterexamples to the General Game-Change Conjecture.

• Disclosure-only reforms fail under MST because they do not change the payoff space
• Mandatory provenance changes the information structure at the moment of choice
• The paper's policy target is the accounting system, not debiasing the individual decision-maker