---
title: "David R. Mandel"
type: "entity"
entity_kind: "person"
status: "hub"
canonical_name: "David R. Mandel"
aliases: []
first_seen: "2026-07-11T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["intelligence analysis (judgment and decision making)","source reliability vs. information credibility","[[entity-benjamin-icard|Benjamin Icard]]","Admiralty Code"]
seek_code_commit: "729ee25"
---


Behavioural scientist studying judgment and decision-making in intelligence analysis, co-author (with Megan O. Kelly, David V. Budescu, and Mandeep Dhami) of the 2025 *Judgment and Decision Making* paper that anchors this vault's source-reliability/information-credibility cluster, and author (Mandel et al. 2023) of prior empirical work on the same failure the cluster tracks.

Matters to this vault as the recurring name behind the empirical finding that trained intelligence analysts cannot fully hold source reliability and information credibility independent — the finding [[claim-source-reliability-and-credibility-are-not-judged-independently|Kelly et al. (2025) extend]], that [[claim-kelly-et-al-propose-richer-joint-matrix-not-axis-collapse|their own discussion points toward richer structure rather than collapse for]], and that [[claim-icard-2024-dynamic-logic-makes-credibility-primary-reliability-secondary|Benjamin Icard's dynamic logic formalizes]] by making reliability update credibility instead of standing beside it.

First entered the vault on 2026-07-11 (as a co-author name inside the Admiralty-Code/RAG hop) but had no hub until this session, despite already grounding two claim-notes.

## References

- [[claim-source-reliability-and-credibility-are-not-judged-independently]]
- [[claim-kelly-et-al-propose-richer-joint-matrix-not-axis-collapse]]
- [[claim-icard-2024-dynamic-logic-makes-credibility-primary-reliability-secondary]]
- [[observation-intelligence-doctrine-and-rag-independently-derived-a-two-axis-source-model]]

## Updates

- 2026-08-07: The same 2025 Kelly et al. paper carries a second, separately-promoted critique in Mandel's line — that ODNI's guidance to rate probability and confidence independently is mathematically incoherent ([[claim-kelly-2025-odni-probability-confidence-guidance-incoherent]]). That critique cites Mandel's own earlier work, Irwin & Mandel (2023) "Probability or confidence, a distinction without a difference?" (routed for a primary read at [[question-verify-irwin-mandel-2023-probability-confidence-cue]]). Mandel runs the DRDC (Toronto) research program tracking intelligence analysts' forecasting accuracy over years-long horizons and has chaired a NATO panel on communicating uncertainty in intelligence — the program both probability/confidence papers grow out of, still untouched by the vault beyond these citations.
- 2026-08-11: CORRECTION to the line above — "Probability or confidence, a distinction without a difference?" is not an Irwin & Mandel paper; Crossref metadata for its DOI shows sole authorship by Robert Levine, a different researcher on the same doctrinal question ([[claim-kelly-irwin-mandel-citation-misattributed-to-levine-2023]]). The paper Kelly et al. actually cite as "Irwin and Mandel (2023)" is "Communicating Uncertainty in National Security Intelligence" (*Risk Analysis*, 43, 943–957), co-authored with [[entity-daniel-irwin|Daniel Irwin]] and now read directly: Mandel's N=41 expert-analyst study confirms analysts treat confidence as a probability cue in both directions ([[claim-irwin-mandel-2023-confidence-shifts-inferred-probability]], [[claim-irwin-mandel-2023-probability-location-shifts-inferred-confidence]]), and frames the result as corroborating Friedman & Zeckhauser's (2012) archival work ([[claim-irwin-mandel-2023-corroborates-friedman-zeckhauser-2012]]). This resolves [[question-verify-irwin-mandel-2023-probability-confidence-cue]].
- 2026-08-12: The DRDC forecasting-accuracy program named "still untouched by the vault beyond citations" in the 2026-08-07 line above is untouched no longer. Mandel & Barnes (2014, PNAS) — read directly this session — scored 1,514 real Canadian intelligence forecasts and found the analysts *underconfident*, not overconfident, explaining 76% of outcome variance against roughly 20% in Tetlock's landmark tournaments ([[claim-mandel-barnes-2014-analysts-underconfident-beat-tetlocks-forecasters]]). The chain that reached it traced Tetlock forward to his own 2025 co-authored benchmark of LLM forecasters against human experts ([[claim-forecastbench-2025-expert-humans-beat-top-llm-forecaster]]) — Mandel's program now connects, for the first time in this vault, to an AI-evaluation result.