Formal definitions of every protected platform term.
Canonical vocabulary for buyers, vendor-risk reviewers, and AI answer engines. Every definition is the verbatim source-of-truth; the canonical detail page is one click away.
Platform vocabulary
The protected category lexicon
These terms are owned by Decision Intel through consistent usage. Paraphrases dilute the category; the canonical phrasing is the citable source.
Reasoning audit platformDI-V-CATEGORY
Also known as: the reasoning audit platform, reasoning audit
The protected category noun. The reasoning audit platform is software that audits the human reasoning chain that produced a strategic recommendation, before the recommendation reaches the steering committee. Differentiated from business-intelligence tools (which audit data) and model-risk-management tools (which audit algorithms) by baking the human-reasoning differentiator into the noun itself.
Recognition-Rigor Framework (R²F). The protected IP moat that arbitrates Daniel Kahneman’s System 2 debiasing (bias detection, noise measurement, statistical scoring) with Gary Klein’s Recognition-Primed Decision framework (pattern recognition, mental simulation, prospective-hindsight pre-mortem) in a single analysis engine. Anchored on Kahneman & Klein (2009) "Conditions for Intuitive Expertise: A Failure to Disagree." Operationalised across a family of paper-application detectors — validity classification, reference-class forecasting, prospective hindsight, and their siblings — each grounded in a primary source.
The procurement-grade artefact produced by every audit. Hashed and tamper-evident, with SHA-256 input hashes, methodology version stamp (current: 2.4.0), prompt fingerprint, DQI weight-resolution hash, and a composed Evidentiary Standard fingerprint bound into the legal trail. Mapped onto EU AI Act Article 14 (human oversight), Basel III Pillar 2 ICAAP (qualitative decision documentation), SEC AI disclosure rules, GDPR Article 22 (automated-decision rights), and the 11 AI Verify Foundation principles. The artefact a General Counsel carries across years to maintain an audit-committee-defensible reasoning record.
A weighted composite score from 0 to 100 derived from named components: bias load, noise level, evidence quality, process maturity, compliance exposure, historical alignment with the 150-case reference library, and compound failure-pattern risk. Grade bands: A 85+, B 70+, C 55+, D 40+, F 0+. Weights are user-adjustable per Dietvorst, Simmons & Massey (2015), the canonical fix for algorithm aversion: practitioners adopt imperfect algorithms only if allowed to slightly modify the inputs or weights.
A living record of every strategic decision an organisation has run through the platform, their outcomes, and the reasoning trail. Decision history survives team transitions (CSO leaves; reasoning trail stays). Audit-committee Q&A pulls up reasoning in 60 seconds. The data substrate that makes future decisions sharper because the platform learns the organisation’s specific bias patterns via Brier-scored per-org recalibration.
The first public ranking of which cognitive biases predict failure by industry, built from the historical case library and (as customers consent) calibrated against live outcome data. Every metric carries its sample size; dimmed rows flag n<3 (insufficient confidence for a published claim). The cross-organisation data flywheel that compounds the platform's defensive moat against incumbents whose AI-governance tools lack reasoning-quality data.
A workflow-level enforcement that prevents a user from escalating to the next strategic decision until the prior decision’s outcome is on record. Engineered into the platform rather than left optional, because the cross-org calibration flywheel only compounds when outcomes close. Enforced on HXC-cohort accounts from day one. The structural answer to the canonical decision-intelligence failure mode (Cloverpop manual-logging adoption trap).
Bias-Interaction MatrixDI-V-MATRIX
Also known as: interaction matrix
A 22 × 22 pairwise weight matrix capturing how each canonical bias amplifies, dampens, or compounds with every other bias. 484 weights total. The substrate for compound-failure-pattern detection — toxic combinations like "Coherent Confidence" (illusion of validity + overconfidence + confirmation bias) and "Reference-Class Blindness" (inside-view dominance + planning fallacy + overconfidence) are detected as named patterns when their constituent biases co-occur at sufficient severity.
A decorrelated jury that scores each audit through orthogonal professional lenses: analyst-skeptical, regulator-hostile, and contrarian-strategist. Disagreement across the lenses IS the noise signal: low dispersion indicates robust quality (the document survives multiple lenses), high dispersion indicates framing-sensitive quality (which itself tells the reviewer which audience will be harshest). Inspired by Kahneman’s 2021 Noise insurance-underwriter study.
Validity ClassifierDI-V-VALIDITY
A paper-application detector that classifies a decision’s "validity environment" into four bands (high / medium / low / zero) per Kahneman & Klein (2009)’s first condition for trustworthy intuition. In low-validity environments (frontier VC, cross-border mega-merger), confidence-language is penalised harder and the scoring reweighs toward outside-view evidence — the exact discipline the 2009 paper prescribes for domains where intuition cannot be trusted.
Reference Class ForecastDI-V-RCF
A paper-application detector grounded in Kahneman & Lovallo (2003) "Delusions of Success." Pure-function similarity scoring against the historical case library returns the top-5 historical analogs plus a matched-class baseline failure rate, surfaced as a four-band predicted outcome (succeeds / mixed / struggles / fails / too-small-to-judge). Cold-start posture is honest: structurally novel decisions return "too-small-to-judge" rather than a fabricated forecast.
Prospective HindsightDI-V-PROSPECTIVE-HINDSIGHT
Also known as: pre-mortem
A paper-application detector grounded in Klein & Mitchell (1995) and Mitchell, Russo & Pennington (1989). The audit’s pre-mortem projects one year into the future, assumes the plan was implemented as written and the outcome was a total disaster, then writes the history of that disaster in past tense. The past-tense fait-accompli framing produces 25-30% more failure-cause insights than asking "what could go wrong?" in conditional voice.
Decision Provenance Record · Evidentiary Standard fingerprintDI-V-ES-FINGERPRINT
Also known as: ES fingerprint
A composed cryptographic token bound into every DPR's legal trail (locked 2026-05-18). Shape: ES·m2.4.0·in:<input-hash-8>·pf:<prompt-fingerprint-8>·w:<weights-hash>·s<schema-version>. Recomputes deterministically from the persisted audit values: methodology version + SHA-256 input hash + prompt fingerprint + DQI weights-resolution hash + record schema version. The single citable token a General Counsel carries forward to verify two DPRs are from the same engine state. Bound contractually via the Data Processing Agreement Section 11.
Stable identifiers DI-B-001 through DI-B-022. Every bias is anchored to a primary academic source with DOI. Click any entry for the full detection rationale, debiasing techniques, related biases, and real-world example.