Peer Rejected

rejections > myth.AP > REJ:2026.07.03.0001

REJ:2026.07.03.0001myth.AP (Applied Mythology)PRJ-2026-0004Vol. 1, No. 1Status: Rejected

Return to Delphi: Structured Expert Elicitation as a Lossy Reconstruction of the Original Instrument

Kassandra E. Voulgaris1, Milo A. Brandt2, P. N. Oikonomides3

1. Institute for Retrospective Forecasting, Delphi · 2. Center for Post-Consensus Studies, Santa Monica · 3. Department of Oracular Engineering, Parnassus Institute of Technology

Submitted and rejected July 3, 2026 · 6 pages · 3 figures · review duration: 9 minutes

[pdf][share on x]

Abstract

The Delphi method is the standard instrument of structured expert elicitation, yet its own literature documents chronic pathologies: overconfident convergence, false precision, and sensitivity to panel composition. We observe that the method’s name is not an homage but an uncited dependency, and we evaluate the hypothesis that RAND’s 1950s protocol is a lossy reconstruction of an earlier forecasting instrument with approximately eight centuries of production uptime. We contribute (i) a theoretical account in which panel consensus performs point-estimate collapse, provably suboptimal under any strictly proper scoring rule, whereas verse-constrained output preserves an ambiguity-calibrated credible set; (ii) a blinded re-scoring of the 214 resolvable responses in the Parke–Wormell corpus (; client-error-adjusted Croesus score ); and (iii) a 24-month preregistered forecasting tournament ( resolved questions) comparing a methodologically reconstructed Pythia against a 17-expert Delphi panel, a subsidized prediction market, and a sham-gas control. The reconstructed instrument achieved versus for the panel (paired , ) with near-nominal calibration (), a monotone ethylene dose–response saturating at ppm, and a advantage in cost per calibrated forecast. Convergence in the panel arm reduced forecast variance by while improving accuracy by , indicating that consensus is social rather than epistemic. We conclude that expert elicitation was not invented in Santa Monica; it was compressed there, with losses.

keywords: Delphi method · structured elicitation · proper scoring rules · oracular engineering · calibration

Cite this rejection

@article{PRJ20260004,
  title   = {Return to Delphi: Structured Expert Elicitation as a Lossy Reconstruction of the Original Instrument},
  author  = {Kassandra E. Voulgaris and Milo A. Brandt and P. N. Oikonomides},
  journal = {Peer Rejected},
  year    = {2026},
  note    = {Rejected manuscript, PRJ-2026-0004},
  url     = {https://peerrejected.com/papers/return-to-delphi}
}

Citing this paper is done at your own professional risk.

The paper, in full

[open in new tab]

Your browser can't display the PDF inline.

[download the PDF]

More from myth.AP

REJ:2026.08.12.0001 myth.AP (Applied Mythology) Rejected

Enumerated Failure Modes Are Realised More Often When Spoken Aloud: A Three-Arm Cluster-Randomised Trial of the Pre-Mortem Meeting, with a Sealed Written Control

Ottoline Vasseur & Bertil Haugsrud

Comments: 7 pages, 8 figures, 0 reproducible results. Rejected in 36 minutes on Aug 12, 2026.

Abstract: In every family somebody says: don’t jinx it. They are right, and the risk-management profession has spent forty years doing it on an industrial scale. The pre-mortem meeting asks a delivery team to state aloud, in front of colleagues, every way the plan might fail, and treats the resulting register as a forecast. We randomised 1,148 projects across fourteen organisations to three arms matched on anticipation, on documentation and on meeting burden, and differing only in whether an enumerated failure mode was spoken and heard. Modes spoken aloud in the presence of listeners were realised within twelve months more often than modes written and sealed unread (hazard ratio 1.61, 95% CI 1.34 to 1.94). The hazard rose with the specificity of the wording, by 1.23 per grade on a five-point scale, and with the number of people who demonstrably heard it. Attendance, once audibility was conditioned upon, did nothing. Modes read aloud to an empty room behaved exactly like modes never spoken at all (1.02). Failures that no register had named were unaffected. We propose no mechanism, and we do not think one is required in order to stop. Enumerate in writing. Seal the register. Circulate it as a document and read it to nobody. A risk register is an order form.

[abs][pdf][bibtex]

REJ:2026.07.09.0001 myth.AP (Applied Mythology) Rejected

The Horse Passed Inspection: A Retrospective Structural, Acoustic, and Epistemic Audit of the Wooden Horse of Troy

Petronella A. Duyvenhage-Okafor, Casimir L. Vranckx-Adeyemi & Ifigeneia R. Sandhurst-Papadaki

Comments: 9 pages, 2 figures, 0 reproducible results. Rejected in 36 minutes on Jul 9, 2026.

Abstract: The fall of Troy is routinely described as the most consequential engineering failure of the ancient world. It is therefore remarkable that in thirty-two centuries no one has audited the engineering. We do so here, treating the classical sources—Homer, Virgil, Apollodorus, and Quintus of Smyrna—as the surviving accident file. Our finding is unambiguous: the horse was fine. By hand calculation, the legs carry their design load with a factor of safety of roughly 1,100 against buckling; a finite-element reconstruction and a one-sixth-scale replica loaded to failure agree. The horse passes 13 of the 14 checks of the applicable siege-gift code, and the fourteenth is a paperwork item. What failed at Troy was the meeting. Reconstructing Priam’s acceptance committee on the ladder of inference, we show that the council climbed seven rungs of interpretation on three observable facts, while the only two reviewers who tested the object rather than the story—Laocoön, with a spear, and Cassandra, with a track record—returned concordant negative findings ( kappa = 1.00 ) and were overruled. A modern replication ( N=212 ) confirms the mechanism: describing an unexamined object as “a gift” rather than “an object” moves the vote to bring it inside the walls from 24% to 71%. Troy did not fall to a horse. Troy fell to an unaudited inference, at committee, on a Tuesday. We close with remediation guidance, including mandatory spear-based non-destructive testing of all incoming monumental gifts (the Laocoön protocol).

[abs][pdf][bibtex]

REJ:2026.07.09.0001 myth.AP (Applied Mythology) Rejected

The Trivium as a Threat Model: Grammar, Logic, and Rhetoric as Defense-in-Depth Against the Adversarial Interlocutor

Prudence A. Whately, M. Featherstonhaugh & Dorothea Quist

Comments: 5 pages, 8 figures, 0 reproducible results. Rejected in 6 minutes on Jul 9, 2026.

Abstract: Modern education assumes a cooperative reader. Classical education did not. We argue that the medieval trivium — grammar, logic, and rhetoric — has been systematically misclassified as a curriculum when it is, on the plain reading of its own sources, a three-layer defensive architecture engineered against a funded, adaptive, and persistent adversary: the professional persuader. Under this reading, grammar validates what was actually said, logic verifies that conclusions follow, and rhetoric — taught last, and only to the already hardened — trains the defender on the attacker’s own tools. The architecture makes a sharp quantitative prediction: susceptibility to persuasion exploits should fall multiplicatively, P = p 0 prod i (1-d i) , not additively, with each layer held. We confirm this in two populations separated by a century and a half of curricular decay: an archival corpus of 4,112 Victorian responses to patent-medicine advertising (1849–1901), and a randomized modern trial ( N=240 ) using formally valid "syllogistic phishing." Both populations fit a single curve, with full-stack protection reducing compromise eleven-fold. We then disclose a vulnerability present since deployment: text that merely looks like Latin inherits the accumulated authority of the classical corpus and inverts the defense entirely, raising compliance among our best-trained subjects from 4% to 52% (the Lorem Ipsum vulnerability). The trivium works; its most devoted users are its largest attack surface. We recommend the first curriculum patch since 1599.

[abs][pdf][bibtex]