TLDR
The cross-domain synthesis script generates three Analysis of Competing Hypotheses (ACH) tables — structured comparison tables that pit competing explanations against each other — testing spoliation (the destruction or concealment of evidence, score: +5.20), willful blindness (+6.40), and asset concealment (+4.70) against their respective null hypotheses. All three strongly favor the alternative explanation. The scores remained stable across every data improvement in the project, indicating the hypotheses rest on structural evidence, not marginal data points (PAPER TRAIL Project, 2026).
What ACH Does
Analysis of Competing Hypotheses (ACH) is an intelligence methodology developed by Richards Heuer at the CIA (Heuer, 1999). Instead of collecting evidence that supports a preferred theory, ACH forces the analyst to evaluate how consistent the evidence is with each competing hypothesis simultaneously. The result is a score: positive means the evidence favors the alternative hypothesis over the null; negative means the null is better supported.
The cross-domain synthesis script's ACH module produces three structured comparison tables, each testing a primary hypothesis of the investigation against an innocuous alternative. The evidence feeding each table comes from multiple analytical domains — wire transfers, corporate registrations, FedEx shipments, bank documents, flight logs, and the NYDFS consent order — ensuring that no single data source drives the result (PAPER TRAIL Project, 2026).
Table 1: Spoliation (+5.20)
The spoliation hypothesis asks whether gaps in the documentary record result from deliberate destruction or concealment of evidence rather than natural loss or administrative error.
The evidence includes the October 2005 FedEx cutoff (shipping records end four months before arrest), sequential gaps in Bates-numbered document series flagged by the German Tank Problem estimator (a statistical method that estimates how many total items exist from the serial numbers of observed samples), the Account Review and Resolution Committee meeting with no minutes taken, and the 42% gap between identified and released pages. The null hypothesis — that these gaps are the product of routine document attrition over two decades — must account for the consistency of the pattern: gaps cluster around legally sensitive periods and affect specific document types more than others (PAPER TRAIL Project, 2026).
The score of +5.20 means the evidence is substantially more consistent with deliberate spoliation than with natural loss. It does not prove documents were destroyed. It says the pattern of what is missing does not look accidental.
Table 2: Willful Blindness (+6.40)
The willful blindness hypothesis asks whether Deutsche Bank's anti-money laundering (AML) compliance failures were systematic institutional blindness rather than isolated individual mistakes.
This table scores the highest of the three, at +6.40. The evidence is drawn directly from the NYDFS consent order: the phantom authorization email cited for six years without verification, the ARRC committee that took no minutes, the "normal for this client" alert clearances, the "sent to a friend for tuition" explanation accepted without follow-up, the 97 structured cash withdrawals of exactly $7,500 each, the threshold-awareness structuring incident, and the reference letters facilitating migration to the next bank (NYDFS, 2020).
The null hypothesis — that these were isolated lapses by individual employees rather than systemic institutional behavior — must explain why seven different categories of compliance failure occurred across multiple departments, over multiple years, involving multiple staff members, and were never caught by internal controls. The consent order itself found that the failures were systemic. The ACH score quantifies how strongly the documentary evidence supports that finding (NYDFS, 2020; PAPER TRAIL Project, 2026).
Table 3: Asset Concealment (+4.70)
The asset concealment hypothesis asks whether the Butterfly Trust structure and post-death entity activity represent deliberate asset concealment rather than routine estate administration.
The evidence includes the Butterfly Trust's $2.65 million in disbursements to beneficiaries including named co-conspirators, the $14.66 million Interest on Lawyer Account (IOLA) flow on the day the last will was signed, the $2 million wire to the previously unknown "Representation Trust" for "legal fees" on August 8, 2019, the $200,000 criminal defense retainer replenishment, and Kahn's $62,400 in pre-arrest cash withdrawals. The null hypothesis — that these were routine legal and financial activities of a wealthy estate — must account for the timing (concentrated in the 35 days between arrest and death), the recipients (entities created specifically for post-death asset management), and the structure (IOLA processing multi-million-dollar flows typically reserved for client escrow) (PAPER TRAIL Project, 2026).
The score of +4.70 favors concealment over routine administration, though with less certainty than the willful blindness table, reflecting the inherent difficulty of distinguishing aggressive estate planning from deliberate asset hiding without access to privileged attorney-client communications.
Stability Across Data Improvements
The most telling feature of these scores is what did not change them. Over the course of the project, VLM processing (using an AI system that reads text from images of documents) recovered 47 wire dates, 44 originator names, and 2 beneficiaries from degraded bank documents. Flight log extraction produced 4,286 flights with 392 unique passenger names. Beneficiary mapping resolved 90 of 104 null beneficiary fields (PAPER TRAIL Project, 2026).
After each improvement, the cross-domain synthesis script was re-run. The ACH scores did not move. Spoliation remained at +5.20. Willful blindness remained at +6.40. Asset concealment remained at +4.70.
This stability matters. It means the hypotheses are not resting on marginal data points that could be overturned by the next batch of recovered fields. They rest on structural evidence — the architecture of entities, the pattern of compliance failures, the timing of financial flows — that does not change when individual data quality improves.
Ten Contradictions, Zero True
The synthesis engine identified 10 contradictions across the evidence base: 7 temporal misalignment and 3 coverage gap. All 10 are explained by the FedEx October 2005 cutoff — the seizure boundary that marks where shipping records end. Zero true contradictions exist: no evidence in the corpus directly conflicts with any other evidence (PAPER TRAIL Project, 2026).
This does not mean the hypotheses are proven. It means the evidence is internally consistent. The 42% of unreleased documents could contain contradictory evidence. But within what has been released, the three hypotheses face no evidentiary opposition.
References
Heuer, R. J., Jr. (1999). Psychology of intelligence analysis. Center for the Study of Intelligence, Central Intelligence Agency.
New York Department of Financial Services. (2020). Consent order under New York Banking Law: Deutsche Bank AG [Regulatory order]. https://www.dfs.ny.gov/reports_and_publications/press_releases/pr202007061
PAPER TRAIL Project. (2026). Cross-domain synthesis ACH module [Script]. app/scripts/25_cross_domain_synthesis.py.
PAPER TRAIL Project. (2026). ACH spoliation table [Data]. _exports/synthesis/ach_matrix_spoliation.csv.
PAPER TRAIL Project. (2026). ACH willful blindness table [Data]. _exports/synthesis/ach_matrix_willful_blindness.csv.
PAPER TRAIL Project. (2026). ACH asset concealment table [Data]. _exports/synthesis/ach_matrix_asset_concealment.csv.
PAPER TRAIL Project. (2026). ACH summary [Data]. _exports/synthesis/ach_summary.csv.
PAPER TRAIL Project. (2026). Contradictions export [Data]. _exports/synthesis/contradictions_all.csv.
PAPER TRAIL Project. (2026). Cross-domain synthesis methodology [Data]. research/CROSS_DOMAIN_SYNTHESIS.md.
PAPER TRAIL Project. (2026). TD Bank SAR extraction [Data]. research/td_bank_sar_extraction.md.