Apply Adversarial Verification

---
name: apply-adversarial-verification
description: Use when you have received information from a party that has motivation to mislead you — to verify claims under adversarial conditions using designed tests, inconsistency detection, and independent corroboration rather than accepting stated information at face value
source: '智囊 (Zhìnáng) 察智部 (Investigative Wisdom) — Feng Menglong (~1624 AD); CIA "Psychology of Intelligence Analysis" Richards Heuer (1999) — Analysis of Competing Hypotheses; Structured Analytic Techniques for Intelligence Analysis (2010)'
tags: [investigation, due-diligence, verification, deception-detection, decision-making]
related: [apply-devils-advocate-technique]
verified: true
---

# Apply Adversarial Verification

When information comes from a party with motivation to mislead, verify using designed tests, inconsistency detection, and independent corroboration — because in adversarial conditions, accepting stated information at face value is not due diligence, it is assumption.

## Why This Is Best Practice

**Why best:** The 察智部 (Investigative Wisdom) of 智囊 contains dozens of historical anecdotes organized around a common pattern: an official, general, or judge receives information and must determine whether it is genuine or falsified. The recurring methodology across these cases: (1) design a test whose outcome depends on the truth — if the claim is true, X will happen; if false, Y; (2) look for inconsistencies between what was stated and what the physical evidence shows; (3) obtain corroboration from sources who have not been able to coordinate their story. This methodology is explicit in the text: "察" (察智) means investigative perception — seeing through the surface to the actual state.

**CIA "Psychology of Intelligence Analysis" (Richards Heuer, 1999):** Heuer's foundational work on intelligence analysis methodology establishes that the human mind naturally seeks confirming evidence and is systematically poor at detecting deception. His proposed counter: Analysis of Competing Hypotheses (ACH) — for each competing explanation, identify what evidence each hypothesis PREDICTS and test against observed evidence. Hypotheses that make false predictions are eliminated. Standard methodology at CIA, DIA, and major intelligence services globally; taught in intelligence analysis programs at George Washington University, King's College London, and others.

**M&A due diligence practice:** Every significant acquisition involves adversarial verification because the seller has motivation to present the company favorably and conceal liabilities. Standard due diligence methodology: verify stated financials against third-party records (bank statements, tax filings); look for inconsistencies between management representations and operational data; interview customers and employees independently (without allowing management to filter access). The $30B+ due diligence industry exists precisely because self-reported information in adversarial contexts is unreliable. Required by fiduciary duty at every major investment bank and private equity firm.

**Forensic accounting:** KPMG, Deloitte, PwC, EY forensic practices specialize in detecting financial misrepresentation. Core methodology: trace physical transactions, look for inconsistencies between stated and reported figures, follow the money. Used in fraud investigations, litigation support, and regulatory enforcement globally.

**Investigative journalism:** The Society of Professional Journalists code of practice requires independent corroboration of all major claims. Standard methodology: obtain documents independently, interview sources who cannot coordinate, design requests that produce documents only the truth-holder would have. Applied at every major news organization globally.

**Why distinct from `design-intelligence-network`:** design-intelligence-network is proactive — gathering intelligence before a competitive engagement to inform a decision. apply-adversarial-verification is reactive — you have already received specific information and need to determine whether it is true. The key difference: in adversarial verification, the information source has motivation to mislead you specifically. The methodology (designed tests, inconsistency detection, isolated corroboration) is specific to adversarial conditions and differs from proactive intelligence gathering.

**Adopted by:** CIA and major intelligence services (Analysis of Competing Hypotheses standard methodology); KPMG, Deloitte, PwC, EY forensic practices; every major investment bank and private equity firm (M&A due diligence fiduciary requirement); Society of Professional Journalists (independent corroboration code of practice).

**Impact:** The $30B+ due diligence industry exists because self-reported information in adversarial contexts is systematically unreliable; structured adversarial verification in M&A prevents material mispricing of liabilities and avoids deals that collapse post-close due to undisclosed information.

## Steps

1. **Identify the adversarial condition.** Before applying this skill, confirm that the information source has motivation to mislead. The adversarial motivation may be: financial (seller hides liabilities, job candidate inflates credentials), political (ally conceals own failures), competitive (opponent feeds false intelligence), or relational (party avoids embarrassing truth). If no adversarial motivation exists, standard verification is sufficient.

2. **Map the key claims and their dependencies.** List the specific claims that, if false, would materially change your decision. For each claim, identify: what would have to be true in the world for this claim to be correct? What physical evidence, third-party records, or verifiable events would exist if this claim is true that would not exist if it is false?

3. **Design verification tests.** For each material claim, construct a test whose outcome depends on truth. Good tests have asymmetric outcomes: if the claim is true, the test produces result A; if false, result B. The party being verified should not know the test is happening or should not be able to manufacture the result. Examples:
   - Claim: "Revenue is $X." Test: request bank statements from the bank directly, not from the company. True revenue produces matching bank records; false revenue does not.
   - Claim: "Key customer X is committed long-term." Test: contact the customer independently without the counterparty's involvement. Committed customers confirm; uncommitted customers equivocate.
   - Claim: "The technology works at scale." Test: request access to production systems under load, not demo environments. Working technology performs; non-working technology fails under conditions the counterparty didn't anticipate.

4. **Map the information for internal inconsistencies.** Review all provided information for cases where two stated facts cannot both be true simultaneously. Inconsistencies do not prove deception — they may reflect error — but they identify claims requiring deeper verification. Common inconsistencies: financial figures that don't add to stated totals; timelines that contradict each other; capability claims inconsistent with known resource levels.

5. **Obtain independent corroboration from isolated sources.** For material claims, verify through sources who have not had the opportunity to coordinate their story with the primary source. Contact references before the counterparty knows you are contacting them; obtain documents from issuing institutions rather than from the party claiming them; interview multiple people with knowledge of the same events separately. When independent sources tell the same story without coordination, confidence rises; when independent sources tell different stories, investigate the discrepancy.

6. **Apply ACH: eliminate hypotheses by what they predict.** List the competing hypotheses (claim is true / claim is partially true / claim is false). For each hypothesis, identify what evidence it predicts. Collect the evidence. Eliminate hypotheses whose predictions are not borne out. The surviving hypothesis is the most defensible conclusion.

7. **Calibrate verification depth to stakes.** Full adversarial verification is costly. Apply depth proportionally: financial terms in a $10M acquisition warrant bank-level verification; a vendor's support hours claim warrants one reference call. For each material claim, estimate the cost of the claim being false and invest verification resources proportionally.

## Rules

- Absence of disconfirming evidence is not confirmation. If your verification found nothing wrong, check whether your tests were capable of detecting a problem. A verification test that only the truth would fail is evidence; a test that both truth and deception pass is not.
- Do not telegraph your verification tests to the counterparty. If they know what you're testing, they can manufacture confirming evidence. Design tests that are either invisible or whose results they cannot fabricate.
- Inconsistencies require follow-up, not immediate conclusions. An inconsistency between two stated facts indicates either deception, error, or incomplete information. Investigate which before concluding.
- Stop before full certainty — it is rarely achievable. The goal of adversarial verification is to reach a decision threshold, not absolute proof. When verification evidence is sufficient to make a decision at the confidence level required, stop. Additional verification returns diminish steeply.

## Examples

**M&A diligence:** Acquisition target reports 40% gross margin. Verification: request management accounts reconciled to tax returns (inconsistency test); compare stated COGS to supplier invoices (independent documentation); interview three customers about pricing independently (isolated corroboration). Discovery: two customers are on below-stated pricing; gross margin is 31%. Deal terms are renegotiated.

**Hiring:** Senior candidate claims to have led a team of 30 in a previous role. Verification: contact the previous employer's HR directly, not through candidate-provided references (isolated source); ask for LinkedIn-visible former reports and contact 2–3 independently without candidate's knowledge. Discovery: candidate was a team member, not the team lead. Hire decision is reconsidered.

**Partnership negotiation:** Partner claims their platform has 2M active users. Verification: request third-party traffic analytics (SimilarWeb, Sensor Tower) rather than internal reports (independent documentation); ask for access to aggregate usage dashboards during due diligence (designed test — real platforms can show this; inflated claims cannot). Discovery: analytics show 400K monthly actives. Partnership terms are adjusted accordingly.

**Intelligence/competitive:** Competitor signals through industry contacts that they are exiting a market segment (feigned retreat). Verification: check hiring patterns on LinkedIn (inconsistency test — companies exiting segments reduce, not increase, headcount); monitor patent applications in the segment (physical evidence); contact shared suppliers about orders (isolated corroboration). Discovery: hiring is accelerating and supplier orders increasing. Competitor signal was deception; real entry is planned.

## Common Mistakes

- **Using counterparty-provided references and documents as primary verification:** The entire point of adversarial verification is to obtain evidence from sources the counterparty cannot control. References selected by the counterparty have selection bias; documents provided by the counterparty can be altered. Always seek independent sources.
- **Confirming rather than disconfirming:** The natural tendency is to seek evidence that confirms your current belief. In adversarial conditions, actively seek evidence that would disconfirm the claim. If you cannot find disconfirming evidence, your verification has more weight; if you were not looking for it, your verification means little.
- **Treating verification as a checkbox rather than an inference process:** Verification is not a list of boxes to check — it is an inference process. Evidence from verification should be evaluated in terms of how much it shifts your probability assessment that the claim is true.
- **Not calibrating depth to stakes:** Verifying low-stakes claims to high-depth wastes resources. Verify proportionally to the cost if the claim is false.