Version v0.6.7-rc5 · 18 parameters · 72 open caveats · confidence 6.6/10
A Fokker–Planck equation with jump process. This describes how a probability distribution over civilization states evolves over time — with drift (predictable trends), diffusion (random fluctuations), and jumps (crises and collapses).
n
relative population (demographic pressure)
w
wage share of GDP (labor vs. capital distribution)
e
elite fraction (overproduction indicator)
d
debt-to-GDP ratio (fiscal fragility)
U
urbanization rate
π
political polarization index
T
institutional trust (0–1)
κ
network connectivity (information propagation)
Layer 1 — Micro-Foundation
Defines the 'particle parameters' Θ — probability distributions governing individual human choice. Loss aversion, temporal discounting, conformity pressure, authority deference, in-group favoritism. These parameters feed into the drift equations as behavioral constants.
Θ = {λ, β_td, γ_conf, α_auth, γ_status, ...}
Layer 2 — Meso-Structure
Defines how individual parameters aggregate into collective dynamics. Network topology G_t determines cascade thresholds and propagation speeds. The computational sociologist identifies the drift equations A_1, A_2, A_3 describing population, wage, and elite dynamics through the limit-cycle (Turchin secular cycle) framework.
G_t = (N_t, E_t, w_ij, c_ij, φ_i)
Layer 3 — Macro-Dynamics
Supplies the empirical backbone — historical patterns that the formula must reproduce, economic power laws (Pareto distributions, debt cycles), and the institutional vector I_t that modulates drift through regime type, veto players, and bureaucratic capacity.
I_t = (R_t, V_t, B_t, P_t, X_t)
Layer 4 — Formalization
Builds the mathematical backbone: the Fokker–Planck equation with jump process, the Ito-corrected PSI stress index, diffusion tensor, and the Bayesian confidence architecture that converts model outputs into honest probabilistic forecasts with calibrated intervals.
Ψ = MMP × EMP × SFD (Ito-corrected)
Cross-Cutting — Epistemology
Attacks every output for curve-fitting masquerading as theory. Enforces falsifiability requirements, detects circular validation, distinguishes causal from correlational claims, and approves or rejects every formula update. No version of CURRENT.md can be committed without philosopher approval.
Verdict ∈ {APPROVE, APPROVE_WITH_CAVEATS, REJECT}
Every parameter with its current estimate, confidence, and layer origin. Parameters marked with open disputes need further validation.
| Symbol | Name | Value | Layer | Confidence | Sources |
|---|---|---|---|---|---|
| λ | Loss aversion coefficient | LogNormal(0.55, 0.35) | micro | 72% | 8 |
| β_td | Temporal discount factor | Beta(6, 2) | micro | 68% | 6 |
| γ_conf | Conformity pressure | sigmoid(x; 0.3, 3.5) | micro | 65% | 5 |
| α_auth | Authority deference 1 open dispute(s) | 0.61 ± 0.12 (HYBRID floor+overlay) | micro | 55% | 7 |
| c_coop | Cooperation frequency invariant | 0.613 ± 0.089 | micro | 82% | 12 |
| γ_ig | In-group favoritism | 0.58 ± 0.15 | micro | 70% | 9 |
| α_pareto | Pareto wealth exponent 1 open dispute(s) | 1.5 ± 0.3 (fat tail) | macro | 75% | 14 |
| T_ineq | Inequality accumulation timescale 1 open dispute(s) | ~125 yr | macro | 52% | 6 |
| T_secular | Secular cycle period 1 open dispute(s) | [150, 265] yr (damped-spiral interpretation) | macro | 62% | 11 |
| E_threshold | Elite overproduction threshold | 0.015 (1.5% of population) | macro | 60% | 7 |
| λ_0 | Jump rate baseline 1 open dispute(s) | 0.015/yr | formalize | 40% | 8 |
| Ψ_threshold | PSI crisis threshold 1 open dispute(s) | 0.7 | formalize | 45% | 9 |
| k_jump | Jump rate exponential sensitivity 2 open dispute(s) | LogNormal(log(2.5), 0.4) | formalize | 38% | 5 |
| η_pareto | Tempered-Pareto diffusion coupling 2 open dispute(s) | LogNormal(log(0.5), 0.8) — PROVISIONAL | formalize | 25% | 3 |
| b_min | Subsistence wage regularization floor 1 open dispute(s) | 0.03 (revised from 0.05) | formalize | 50% | 4 |
| β_U | Urbanization convergence rate 1 open dispute(s) | 0.020/yr | macro | 40% | 5 |
| π_path | Institutional path dependence | 0.96 | macro | 65% | 8 |
| δ_prop | Propaganda effectiveness | Beta(2, 5) → mean 0.28 | macro | 55% | 6 |
Most consequential single-session falsification in formula history. OQ-25-D EXECUTED env-capable for first time in 26 sessions (Python 3.14.3, scipy 1.17.1) per the Session 26 frozen pre-registration (hash 27299d86...4ba6b7). Result: branch determination FAIL-A; mechanism FAIL-STRUCTURAL. Closed-form Jacobian at the analytical interior fixed point is block-triangular with all-real eigenvalues {−0.20, −0.08, −0.0145, +0.00106}/yr — no complex pair, no Hopf bifurcation possible at any agrarian-prior parameter setting. Both v0.6.6 deterministic-limit-cycle framing AND v0.6.7-rc1→rc4 noise-sustained-quasi-cycle reframing FALSIFIED at agrarian priors. The 8 Turchin secular cycles as empirical data are NOT falsified — what is falsified is one specific MATHEMATICAL MODEL of those cycles. State-vector decomposition, Bayesian framework, archetype taxonomy v1.0, Polymarket scoreboard, institutional vector I_t, Psi indicator NOT falsified. Drift-block functional forms might still be valid in modern-regime parameters; only agrarian regime conclusively falsified. Stable v0.6.7 promotion FORMALLY CANCELED at existing drift block. C12-7 falsification commitment RESOLVED-BY-FALSIFICATION. Path forward CANDIDATE only (η_n + ε bidirectional couplings; 0 new free parameters). NO SDE structural change THIS SESSION; annotation-only bump. Confidence 6.45 → 6.55 — pre-registered falsification firing against own prediction is epistemological progress, not regression. Subsequent Philosopher Mode A continuation Session 28 (2026-04-29) committed joint OQ-25-D-v2 pre-registration under SHA-256 a5940277defd7e65d89d303ead69bdb342e578de2c003e64009916ee67524e6a; no formula edit. Confidence 6.55 → 6.60.
C23-3 CLEAN attestation delivered via 4-anchor multi-method literature triangulation (Galí 2011 JEEA + Daly-Hobijn 2014 JMCB + Turchin 2003 + Wrigley-Schofield 1981). alpha_w prior tightened LogN(log 0.08, 0.30); 95% CI [0.044, 0.146]/yr. NET-ZERO retirement bundle: 1 retirement (κ_λ → derived constant h_0 = 1.219 from micro.md S1 LIVE λ̄=1.73; Tversky-Kahneman 1992 λ=2.25 explicitly DISCARDED) + 3 reclassifications (γ_K=0.69 Livi-Bacci central; e_max two-regime 0.03 agrarian / 0.10 modern; η_w=0.5 fixed-canonical macro-constant). Net free-prior change −4 (46 → 42); C15-3 NET-ZERO RESTORED conditional on Cliodynamicist OQ-23-A non-rejection of γ_K Livi-Bacci anchor. σ-noise priors PARTIAL: F.26 σ_n LogN(log 0.010, 0.5); F.27 σ_w LogN(log 0.030, 0.4); F.28 σ_e LogN(log 0.010, 0.5). OQ-25-D pre-registered (DOP853, 64-replica IC ensemble, hash 27299d86587a59dfceab07a4ebef5a7b130d00ac816564437df64c807ba4a6b7); execution env-blocked, handed to Stat Physicist Session 27. OQ-26-A: at proposed σ-noise priors, Alonso-McKane-Pascual spectrum is OVERDAMPED (refines, not invalidates, B24-3 sign prediction). Stable v0.6.7 promotion REMAINS DEFERRED. Confidence 6.20 → 6.45.
Scaffolding-only patch. OQ-24-A analytical PARTIAL via 3-way triangulation (Session 12 2D Jacobian -0.06 ± 0.02i + CROSS-054 slow eigenvalue ≈ -μ_0 + Wittmann-Kuehn 2024 DWM Hopf-slice). Predicted t_{1/2} ≈ 14.4 yr in pre-registered bracket [10, 20] yr — SUPERSEDES Session 24 [30, 70] yr per Philosopher Ruling 1 (analytically-derived, pre-numerical TIGHTENING is permitted). Three methodology pre-registrations: C23-4 Monte Carlo 95% CI (N ≥ 10,000 + Sobol); OQ-24-F hierarchical-budget cap σ(log T) ≤ 0.25; C14-1 PSIS-LOO BF 0.1/10 thresholds. C19-3 SQL infrastructure-blocker record filed. C23-3 alpha_w attestation NOT delivered by Comp Sociologist Session 25 close — DIRTY-branch confidence default. NO SDE structural change, NO new free parameters, NO retirements; p ≈ 46 unchanged. Confidence 6.35 → 6.20 (DIRTY-branch per Philosopher Ruling 7).
Literature-anchored reframing: closed A_1/A_2/A_3 system at agrarian parameters predicted to be a stable damped spiral, not a deterministic limit cycle, with oscillation sustained via the Alonso-McKane-Pascual 2007 demographic-noise amplification mechanism (Bladon-Galla-McKane 2010). Wittmann-Kuehn 2024 PLOS ONE shows Hopf bifurcation only in thin parameter slices; agrarian baseline outside the slice. Falsification tolerance widened [250, 300] → [230, 320] yr ensemble mean — Philosopher ruled PRINCIPLED (noise-induced lengthening + alpha_w sensitivity known a priori; prediction-HARDENING from 1 scalar to 4 signed predictions). macro.md §F NEW: registers γ_K, ε_loss, η_w, e_max, k_H, τ_ref, κ_λ, w_eq_0. C24-3 silent-creep audit on κ_λ; C24-6 BLOCKING-doctrine codified. C22-3 T_ineq watch CLOSED (3/3 PASS). Confidence 6.25 → 6.35 conditional.
First release-candidate version label in formula history. C12-7 BLOCKING PROPOSED-CLOSED via explicit algebraic closed forms for A_1/A_2/A_3 drift coefficients. A_1: logistic population growth with Psi-modulated K_eff plus Heaviside crisis-mortality (Turchin-Nefedov 2009; r ≈ 0.015/yr). A_2: mean-reversion -α_w(1-B_t)·[w-w_eq] with phi_loss kink and τ_ref ≈ 5yr endogenous reference wage (Galí 2010, Daly-Hobijn 2013; α_w ≈ 0.08). A_3: μ_0(1-V_t·X_t)·[(w_0-w)/w]·(1-e/e_max) (Turchin 2016; e_max ~0.10 agrarian). Linearized Hopf period T_lin ≈ 2π/√(r·α_w·η_w) ≈ 256 yr — structural match without target tuning, conditional on C23-3 alpha_w provenance attestation Session 24. A_1 label downgraded to CORRELATIONAL-AT-ALL-SCALES per Philosopher. C15-3 ESCALATION: NET +4 parameters (p ≈ 41 → 45); retirement debt owed Session 26. Confidence 6.10 → 6.25 conditional.
First NET-NEGATIVE parameter-budget session in formula history. C15-1 HIGH closed via Rosinski 2007 (Stoch. Proc. Appl. 117:677-707) tempered-stable theorem: η_pareto and α_pareto are structurally orthogonal in the Lévy-measure decomposition; no double-counting. T_ineq RETIRED to derived diagnostic (free count 42 → 41). Causal Status column added to §2 drift table (A_2 CAUSAL-AT-MICRO / CORRELATIONAL-AT-MACRO; A_3/A_5/A_7 CORRELATIONAL; A_4 CAUSAL-AT-IDENTITY). Logistic vs OU urbanization adjudication: P(logistic)=0.65, P(OU)=0.35 weights pending PSIS-LOO. LPPL crash SDs tightened. Dem 2028 first archetype firing executed: GATE_BLOCKED_NO_PREDICTION (POP_LEFT cluster ~8% vs hurdle 21.7%, gap -13.7pp; legitimate protocol behavior). Confidence 5.85 → 6.10.
C19-2 BLOCKING closed via dual-location sub-addendum pattern: full MR1-Turkey dual-coding protocol inlined into archetype_calibration_protocol.md §4.6 + forward-reference pointer at archetype_reference_classes.md §9.10 with §§1-8 byte-invariance attestation. Sub-addendum doctrine codified as a four-constraint protocol. Philosopher ruled the §§1-8 slice-hash 59b72533abc4c3ac002607bda7f48652ae591cdf9cd494007a76390c369db525 the load-bearing lock-commitment (CROSS-105 resolved). AND-gate firing rule formalized (3 scenarios must each clear ≥15% lift independently). First operational archetype prediction (Dem 2028, $1.09B) AUTHORIZED for Session 22+. C19-3 SQL execution auto-freeze extended Session 23 → Session 25 per C21-3. Confidence 5.75 → 5.85.
Mechanical-closure session. Taxonomy schema rotated v1.0.0 → v1.0.1 with new SHA-256 f9f022a19e1347032b89c7a1178a7d986f203e0a5eb3cf1632d964d000600964 (verified on disk by Philosopher). POP_RIGHT P_t/V_t numerical boundaries inlined into populist_right.md §§9-10. C18-1 closed: stale DRAFT adjective stripped from personalist_authoritarian.md §9.4. macro.md §E regime-wide vs archetype-conditional P_t clarification added. NET-ZERO parameter budget preserved. Reset cost on archetype retrodiction history: zero. Confidence 5.7 → 5.75.
C19-BS-1 closure package: 22-cell κ-propagated Beta posteriors over the 160-cell archetype × institution-vector grid; leave-Turchin-out BMA showing POP_RIGHT × LD × crisis-bulge robust at +4.6pp lift; C17-1 POP_RIGHT P_t/V_t profile boundaries LOCKED. Anchored on Tsebelis 2002, V-Dem v2mebias/v2mecenefm/v2juhcind, KESMA Hungary, Linz 1975. Conditional unfreeze granted for LD-profile POP_RIGHT predictions at horizons ≥6mo. Dem 2028 firing NOT authorized this session — deferred to Session 20+ pending CROSS-093/094/096/097 mechanical closures. 7 new C19 caveats. Confidence 5.6 → 5.7.
Patch-level bump executing Political Scientist Memo M1-M6 file-side. C16-3 / CROSS-081 CLOSED via DRAFT → LOCKED v1.0 rename of historical_tests/events/archetype_reference_classes.md. Locked file SHA-256 91a464ae72201a08d3cf36a7ba665aefb0a29117759a8e3514bd304d51942fc4; reassignment JSON SHA-256 575e93a01bb942449d1ed5bb824fbbbccf39df25127889442840eafa1c6b3395; taxonomy schema hash 1a9225f0... preserved. C17-2 RESOLVED with merged-at-§4.1 / split-at-§4.3 decision. Interwar fascism (Germany 1933 NSDAP, Italy 1922/1924, Hungary 1939/1944, Austria 1934, Romania 1937) marked UNCLASSIFIED per M4. Spain 1939 + Portugal 1933 reassigned to personalist_authoritarian (NR3 exceptions). CROSS-072 PARTIAL: T_secular tagged LEAST CIRCULAR among the three flagged Turchin parameters. No SDE parameter value changes; net-zero budget preserved. Epistemological confidence 5.5 → 5.6.
BLOCKING closure session. Patch-level bump; schema hash 1a9225f0... PRESERVED via C17-3 byte-reversion (TISZA facts migrated to a §9.8 sub-addendum outside the v1.0.0 hash base). Scope restrictions only, no SDE changes. C16-1a CLOSED: personalist_authoritarian restricted to regime-transition predictions; AUTH_PERS × LD and AUTH_PERS × ED cells NOT APPLICABLE (GWF 2014 regime-level classification is decisive). C16-1b CLOSED: symmetric mirror restriction on reformist_liberal in post-backsliding contexts (REFORM_LIB × EA, REFORM_LIB × CA NOT APPLICABLE). CROSS-086 CLOSED: populist_right two-profile split (Liberal-Democratic vs Electoral-Autocratic) grounded in Hungarian media-capture asymmetry (~80% indirect Fidesz media control via KESMA vs no equivalent in US/Italian LD contexts). Memo M1-M6 produced for ~23 NAT_AUT/MIL_REL historical case reassignments; file-side execution handed off to Cliodynamicist Session 18. TISZA VERIFIED RETRODICTED EXEMPLAR (Brier 0.0400 formula vs 0.0729 market) documented in reformist_liberal.md §9.8. Retrodiction tally unchanged (8 pass / 4 partial / 0 fail / 2 untestable). Epistemological confidence UNCHANGED at 5.5/10. Archetype operational freeze continues — sole remaining BLOCKING item is C16-3 Cliodynamicist Session 18 file-side execution.
Archetype Taxonomy v1.0 LOCKED — 8 ideological-archetype reference classes defined (populist_right, reformist_liberal, personalist_authoritarian, + 5 others) with pre-registered coding rules from CHES, PopuList, V-Dem, Manifesto Project. 8 historical baselines initialized. Archetype-level predictions remain operationally frozen pending C16 closure chain. First archetype-conditional Polymarket scoring applied to the Hungary TISZA prediction (retroactively scored as populist-right vs reformist-liberal archetype contest).
OVERFITTING REGIME FORMALLY DIAGNOSED: p≈42 free parameters, n_eff≈6–14 independent observations. Practical R² ceiling ≈0.05–0.15 at 1yr horizon. PARAMETER FREEZE enacted (net-zero budget — nothing new without retiring existing). Prior catalog expanded 20→27. Circular audit expanded 4→7 parameters; all 4 fully circular params concentrated in the jump process J[P]. b_min revised 0.05→0.03; k_jump recalibrated 3.0→2.5 (preliminary non-Turchin MLE). 3-tier out-of-sample validation protocol (LOOCV / Blocked-CV / CPCV) designed. Epistemological confidence revised down 5.8→5.5/10. No new Polymarket predictions — internal moratorium until calibration improves.
C13-1 STRUCTURALLY RESOLVED: 6 independent non-Turchin cases (Mughal, Meiji, Iran, Weimar, Rwanda, Spain) — 6 PASS / 2 PARTIAL / 0 FAIL. β_U = 0.020/yr defined: 8/8 drift coefficients now have estimates. Zone-conditional PSI validity framework.
FIRST PHILOSOPHER LEAD SESSION. Falsifiability 6/10; overfitting risk MODERATE-HIGH. 5 formula changes: scope boundary, PSI causal note, circular flags, rho_DA/chi_horm to Theoretical Scaffolding, Venezuela reclassified.
C7-1 PSI ITO RESOLVED (12-session journey). Full dΨ SDE derived, sympy-verified. Novel w=0.5 sign-change prediction. CROSS-013 downgraded CRITICAL→HIGH. Period revised [150, 265]yr.
F_pol (A_6 polarization drift) APPROVED with surgical fixes. Limit-cycle period FORMALIZED [165, 327]yr envelops empirical [160, 335]yr.
Limit-cycle reframe APPROVED: Turchin phases are quadrants of ONE limit cycle (Hopf bifurcation). Period discrepancy mechanism identified (3 factors). FINAL BOOTSTRAP SESSION.
All Section A defined: α_pareto=1.5, α_tail=3.0, m_crash=0.33, T_debt 7yr/75yr, T_ineq~125yr. 5/6 drift coefficients specified. α_tail~3 validates FP+jump architecture.
CRIT-01 RESOLVED: 10D → 8D state vector (Ψ and φ now observables). CRIT-02 PARTIALLY RESOLVED: jump kernel formalized. γ_conf sigmoid form APPROVED.
9/13 Θ_fixed defined. HYBRID floor+overlay model for evolutionary constants. Cooperation invariance (Spadaro 2022: 70 societies) — STRONGEST finding.
Institutional vector I_t = (R_t, V_t, B_t, P_t, X_t) fully defined. Per-equation drift modulations A_1–A_8 with institutional constraints. π_path=0.96; δ_prop specified.
Bayesian Confidence Architecture defined (Section 6). Prior catalog: 12 informative + 8 placeholder. First prediction: NATO exit P=2.5% [0.5%, 8%] vs market 10%.
8 macro parameters defined: λ_0, Ψ_threshold, Ψ_critical, k_jump (jump process); T_secular (secular cycles); E_threshold (elite overproduction); F_fiscal; α_war. 5 Philosopher fixes applied.
14 micro-parameters defined. 4 critical parameters: λ (loss aversion), β_td (temporal discounting), γ_conf (conformity), α_auth (authority deference). 3 parameters flagged as replication failures.
Full FP+jump formalization. 10D state vector formalized. Three order parameters (PSI, phi, Gamma) defined. Turchin-PSI connection established.
Framework skeleton: Fokker-Planck + jump process architecture. 10D state vector proposed. PSI composite index architecture sketched.