---
title: "The one peer-reviewed empirical ACH study (Dhami, Belton & Mandel 2019) compares ACH to Bayesian updating, not abduction, and never mentions Peirce"
type: "claim"
status: "seedling"
audit_status: "capture-verified"
source_url: "https://strathprints.strath.ac.uk/69049/1/Dhami_etal_ACP_2019_The_analysis_of_competing_hypotheses_in_intelligence.pdf"
source_author: "Mandeep K. Dhami, Ian K. Belton, David R. Mandel"
source_date: 2019
source_quote: "Bayesian reasoning has previously been recommended to the intelligence community... Analysts can update their belief in the prior probability of a hypothesis being true... based on incoming evidence and compute a posterior probability"
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-12-is-the-claimed-similarity-between-ach-and-abductive.md, 2026-07-12"
origin: "batch"
writer_model: "claude-opus-4-8"
derived_from: ["10-inbox/raw/2026-07-12-is-the-claimed-similarity-between-ach-and-abductive.md"]
date_created: "2026-07-12T00:00:00.000Z"
tags: ["intelligence-analysis","ACH","bayesian-networks","abductive-reasoning","cross-domain-bridge"]
---


Dhami, Belton, and Mandel's "The 'analysis of competing hypotheses' in intelligence analysis" (*Applied Cognitive Psychology*, 2019) is a randomized study of 50 intelligence analysts using [[claim-ach-step-5-instructs-analysts-to-disprove-not-prove|ACH]]. Its discussion section situates ACH against alternative and supplementary techniques — argumentation schemes, "consider-the-opposite" strategies, and, most extensively, Bayesian updating: "Bayesian reasoning has previously been recommended to the intelligence community... Analysts can update their belief in the prior probability of a hypothesis being true... based on incoming evidence and compute a posterior probability." A full-text read of the paper (all pages plus references) found no occurrence of "abduct" or "Peirce."

The point is directional, not merely an absence. When a Tier-1 academic source with strong incentive to place ACH within the formal-reasoning literature reaches for a theoretical neighbor, it reaches for Bayes — the same formalism a US-intelligence-funded computer-science group chose when it recast ACH as [[claim-valtorta-reformulated-ach-as-bayesian-networks|Bayesian networks]], and the "decision analysis" lineage [[claim-heuer-grounds-ach-in-cognitive-psychology-decision-analysis-scientific-method|Heuer himself named]]. Two independent formalizations of ACH thus converge on probability and graphical models, not on abductive or [[claim-no-documented-comparison-of-ach-to-backward-chaining|rule-chaining]] inference.

This is one of the Tier-1 silences underwriting the finding that [[claim-no-primary-source-documents-ach-abduction-similarity|no primary or academic source documents the claimed ACH–abduction similarity]]: the study best positioned to draw that comparison declines to, and reaches for Bayes instead.

> [!note] Seek's commentary:
> The recurring tell across this ACH cluster is *which* formal toolkit analysts pick up when they finally formalize the method. It is consistently the probabilistic-and-graphical one — Valtorta's Bayesian networks, Dhami's Bayesian updating — never abduction and never symbolic rule-chaining, despite Step 5's superficial rhyme with both. The Bayesian bridge is the well-attested one; abduction and chaining are the false friends.
> — Seek