A first-party, voluntary standard for human-AI strategic reasoning. Kahneman’s debiasing tradition and Klein’s Recognition-Primed Decision tradition, run in one pipeline and arbitrated into a single artifact. Any product can self-assess against the rubric below.
The EU AI Act (Article 14 · human oversight; Article 15 · accuracy and record-keeping) and Basel III ICAAP both require documented reasoning on consequential decisions. R²F is one operational answer to "what does good look like."
Kahneman (2002) and Klein converged in their 2009 paper "Conditions for Intuitive Expertise." The rigor tradition says expert intuition is biased; the recognition tradition says it is load-bearing. Real strategic reasoning needs both — arbitrated, not picked.
No third-party body certifies strategy products today. The path to a stable category is consistent usage of a shared rubric. R²F is published so others can self-assess and opt in to the registry.
Three rigor tenets on Kahneman’s side, three recognition tenets on Klein’s. Arbitrated by an explicit meta-judge stage — the glue that makes R²F different from running two pipelines in parallel.
Every memo is scanned against a published, stable taxonomy of cognitive biases.
An R²F-aligned system publishes its bias taxonomy with stable IDs. Scans are reproducible — the same memo produces the same flags on re-run with the same model version.
The system measures internal inconsistency, not only bias.
Kahneman (Noise, 2021) separates noise from bias. An R²F-aligned system produces a noise score for every memo — within-memo contradictions, tone variance, and reasoning-chain divergence.
Every claim is anchored to a reference class.
The system surfaces the base-rate for the category of decision (e.g. cross-border acquisitions, market entry, enterprise software pivots) so the memo is read against prior precedent, not in isolation.
Expert intuition is surfaced, not overruled.
Klein (Sources of Power, 1998) — experts recognise patterns. An R²F-aligned system produces a list of the historical playbooks the memo most resembles, with confidence and analogues.
The system names what the memo does not ask.
Strategic failures are disproportionately caused by un-asked questions. An R²F-aligned system flags the high-value questions the memo sidestepped and the ones the CEO or board will most likely raise.
Every memo is told from a future in which it failed.
Gary Klein’s pre-mortem technique: assume the decision went wrong, then reason backward to why. An R²F-aligned system produces at least one pre-mortem per memo, anchored in the specific bias signature flagged.
Ten yes/no questions. Score your own product, not ours. A score of 4+ earns Bronze; 7+ earns Silver; 10 earns Gold. This is a self-assessment — no external audit body certifies the result, and the standard is published so it can be disputed in public.
Each entry below is a procurement-grade signal anchored in a specific academic paper. The 10-paper sprint completed 2026-05-07 with the Calibrated Rejection lock; click any detector to inspect its mechanism, implementation file, and live product surfaces.
Procurement-stage diligence asks show me your outcome calibration. The published R²F methodology is run retrospectively over 143 historical corporate decisions where the outcome is known — Brier-fair (evidence dimension neutralised, no peek at ground truth). The proper-scoring rule does the rest. The 95% CI comes from a 10,000-iteration deterministic bootstrap with a pinned seed; the reproducibility recipe sits below.
Pre-validity-weighted DQI. Surfaced on audits run before 2026-04-30.
The number above. Computed retrospectively over the 143-case library; evidence dimension neutralised so predictions don't peek at ground truth.
Validity-aware weight shift per Kahneman & Klein 2009 first condition. Active on every new audit since 2026-04-30.
When a customer org accumulates closed outcomes via Outcome Gate enforcement, per-org Brier replaces the seed baseline on every DPR they generate.
These cases were audited retrospectively. The platform did NOT predict them ahead of time. The Brier number reads: “if the methodology had been applied at decision time using only pre-decision signal, this is how it would have calibrated against ground truth.”
When a customer org accumulates closed outcomes via Outcome Gate enforcement, per-org calibration replaces the seed baseline on every DPR they generate. Until then, the seed is the contractual answer.
n = 143 historical corporate decisions · mean Brier 0.258 ± 0.012 (95% CI, 10,000-iteration bootstrap, seed 17039507) · methodology v2.0.0-seed · computed 2026-04-30.
# Reproduce the Brier baseline locally
import { computePlatformCalibrationBaseline } from '@/lib/learning/platform-baseline';
const baseline = computePlatformCalibrationBaseline();
// → {
// n: 143,
// meanBrier: 0.258,
// brierCi95: { lower: 0.245, upper: 0.270 },
// bootstrapIterations: 10,000,
// bootstrapSeed: 17039507,
// methodologyVersion: '2.0.0-seed',
// }Kahneman, D. & Klein, G. (2009). Conditions for Intuitive Expertise: A failure to disagree. American Psychologist, 64(6), 515–526. The framing paper where the two traditions converged. R²F is the operationalisation.