EVIDENCE SUMMARY: Moral Decay Project
EXECUTIVE SUMMARY
The project relies on a robust computational core (moral_decay_compute.py) which produces high-quality fits ($R^2 > 0.9$ for many domains). However, the textual representation of this data varies significantly across papers, leading to contradictions in dates, p-values, and mathematical models.
CRITICAL DISCREPANCIES
1. The “When” Question ($t_0$)
- Paper 1: 1967
- Paper 2: 1958.6 (Mean) / 1960.6 (Median)
- Paper 3: 1968-1973
- JSON Truth:
t0_mean= 1968.9. - Diagnosis: Paper 2 seems to be using an older or different run, or explicitly subtracting a “lag” period. The JSON data supports the “Late 60s” rupture (Paper 1/3) more than the “Late 50s” initiation (Paper 2), unless “Inflection” is defined differently.
2. The “Certainty” Question (P-Value)
- Paper 1: $p < 10^{-12}$
- Paper 2: $p < 10^{-3}$ (Abstract) / $5.5 \times 10^{-4}$ (Body)
- Paper 3: $p < 10^{-6}$
- JSON Truth:
ks_pvalue= 0.0033 ($3.3 \times 10^{-3}$). - Diagnosis: The papers drastically overstate the statistical significance. $0.003$ is strong (“One in 300”), but it is not “One in a Trillion” ($10^{-12}$). This hyperbolism endangers credibility.
3. The “How” Question (Model)
- Papers 1 & 2: Exponential Decay ($\chi \sim e^{-\lambda t}$).
- Paper 3: Phase Transition / Power Law ($\chi \sim |P-P_c|^\gamma$).
- Diagnosis: These are conflicting mathematical descriptions.
- Exponential: Gradual, constant-rate loss.
- Power Law: Stable… Stable… CRASH.
- Resolution: The data (Inflection points) supports the Phase Transition model (Paper 3). The Exponential fit might be too simple.
ACTION PLAN
- Re-Run Computation: Confirm if the
0.003p-value is the best achievable. If so, update all texts. - Standardize $t_0$: Agree on the definition. Is it the “Start of Rot” (1958) or the “Year of Breakage” (1968)? The JSON measures the Breakage.
- Unify Math: Adopt the Phase Transition model as the primary theoretical framework, with Exponential Decay describing the driving force (Constraint Removal).
DATA INVENTORY
- Processed: 23 Domains
- Excluded: 19 Domains (No variance/usable series)
- Strongest Evidence: Union Membership, Nonmarital Births, Church Attendance.
- Weakest Evidence: “Polarization” (Missing), Abortion (Data artifacts).
Generated: 2026-01-11
Canonical Hub: CANONICAL_INDEX