THEORY A, OPERATIONALIZED — the adaptation gap as a measurable claim

Version: 1.0 · Status: Ratified baseline (author, Session 4) · still living · Last updated: Session 4

Rung labels — APPLIED (Session 4k, the author’s ratifying pass; C-034 discharged, C-033 divergence closed). Under the adapted evidence ladder (docs/PRESSURE_TESTS_CAUSAL_HYPOTHESIS.md §9): Theory A’s core claim — that the adaptation gap G = R_c − R_a drives the symptom-severity composite, with G leading and surviving the wealth-pump control — is labeled theory · untested. Its measurement design (Q-001) is drafted but unrun; no claim in this document is run:survived. The decisive D3-discriminator (does G survive controlling for the wealth pump) is coded by the lag-type instrument, which was re-balanced at S4k (§8) to stop defaulting against D3 — the crux is now codable-in-principle without a built-in tilt, and untested-in-fact per the S4k concession. Flattery-audit: this doc lands nothing above theory, which is the honest ceiling for an unrun design.

The first Phase-1 work product: turning Theory A (the Adaptation Gap, CANDIDATE_THEORIES.md) from an evocative frame into a stated, falsifiable, pre-registerable claim — the flagship task (OPEN_QUESTIONS.md Q-001) and the project’s single greatest intellectual liability. Theory A has the best integrative reach and the weakest testability; this document attacks the weakness head-on. It does not claim the theory is true — it claims to make the theory checkable, which is the precondition for finding out. The honest posture throughout: a gap you cannot measure is a story, not a science (Turchin’s standing objection), and this document is the attempt to earn the second word.

1. The claim, restated as a measurable proposition

Informal (Theory A): today’s defining condition is that the systems humans build change faster than the humans and institutions inside them can adapt; most symptoms are downstream of that widening gap.

Measurable restatement:

Let R_c = the rate of change of human-built systems, and R_a = the rate at which institutions, cognition, and shared culture adapt to them. Define the adaptation gap G = R_c − R_a (or, dimensionless, the ratio R_c / R_a). Theory A predicts that G, measured independently of its symptoms, positively predicts the severity of the symptom-set (fertility collapse, anomie, populism, epistemic breakdown, coordination failure), cross-sectionally and over time, and retains explanatory power after controlling for rival drivers — chiefly the wealth pump (D3).

Three things make this a real claim rather than a restatement: G must be built from proxies chosen and weighted in advance of looking at the outcomes; the prediction must survive controls for at least the strongest rival driver; and there must exist stated conditions that would prove it wrong (§5). All three are below.

2. The two rates and their candidate proxies

The move is to measure each rate from several domains, each proxy defensible on its own, then composite them (§3). Proxies are candidates — the pre-registration (§7) fixes the final set before any outcome is examined.

R_c — rate of systemic change (faster = larger):

R_a — rate of adaptation (faster = larger, i.e. the system keeps up better):

Two design notes. (a) R_a measures adaptive throughput, not institutional quality — a fast-but-bad adaptation still counts as adaptation; the theory is about pace-matching, not goodness. (b) Several R_a proxies are inverse lags (shorter is faster), which must be sign-corrected before compositing.

3. The gap, and the commensurability problem (the hard part)

The central technical difficulty — and the most likely place the theory fails as measurement rather than as idea — is that R_c and R_a are gathered in different units across different domains. “Frontier compute doubling” and “years-to-regulate” cannot be subtracted directly. The gap G is only meaningful after normalization, and normalization is where a composite index can be quietly tuned to produce any result one likes (the near-tautology risk, C-004’s cousin).

Disciplined approach:

1. Normalize within domain, over time. Convert each proxy to a standardized series (z-scores against its own history, or an index to a base year), so each is unit-free and comparable across domains.
2. Composite by a pre-registered rule. R_c and R_a are each the (pre-declared) weighted mean of their normalized domain proxies. Default: equal weights, because any cleverer weighting invites overfitting; alternative weightings are reported as robustness checks, not as the headline.
3. G = R_c − R_a on the normalized scale (and R_c / R_a reported alongside).
4. Pre-register before outcomes. The proxy list, the sign corrections, and the weights are fixed and published before G is correlated with any symptom. This is the single most important methodological guard: it is what turns G from a curve-fit into a test.

If reasonable alternative proxy sets and weightings yield materially different G trajectories, that is not a nuisance to smooth over — it is evidence that G is an artifact of measurement choices, which is one of the falsification conditions (§5). The theory has to survive its own measurement sensitivity.

4. The prediction, and the outcome variables

Symptom-severity index (the dependent variable): a pre-registered composite of measurable symptom proxies — e.g. loneliness/anomie indices, populist vote share and affective-polarization measures, institutional-trust decline, and (where measurable) epistemic-fragmentation proxies. Fertility is included with care, since it has strong independent economic drivers.

Two forms of the prediction:

• Cross-sectional. Across comparable units (≈30 OECD countries, or sectors), units with a wider G show more severe symptoms, all else equal. Units where R_c and R_a are closer show milder symptoms.
• Longitudinal. Within a unit over time, as G widens, symptoms intensify with a lag (adaptation and its failures are slow), and where G narrows (slower adoption or faster institutional adaptation) symptoms ease. The lag structure matters: Theory A predicts G leads symptoms; if symptoms lead G, the causal story is wrong (§5).

5. Falsification conditions (stated in advance)

Theory A, so operationalized, is falsified — not merely strained — if:

1. Narrow-gap severity. Symptoms are as severe, or worse, in units/periods where G is narrow. The gap would then not be the operative variable.
2. A rival dominates. Symptom severity tracks a specific single driver — inequality/the wealth pump (D3), a particular technology, a specific policy regime — better than the composite G, and G adds no independent explanatory power once that driver is controlled (§6). Theory A would then be a less-good version of that rival.
3. Measurement artifact. Reasonable alternative proxy sets and weightings produce materially different G trajectories, so “the gap” is an artifact of construction rather than a real quantity.
4. Wrong lag direction. Symptoms systematically lead G rather than follow it, contradicting the causal ordering (overwhelm-follows-gap).

Stating these in advance is what distinguishes this from a frame that survives every outcome by reinterpretation. If none can be checked, the operationalization has failed even if the idea is appealing.

6. The discriminator: Theory A vs Turchin’s wealth pump (D3 / D-002 / D-006)

This is the decisive test, and it connects three open items — the single-root-vs-coupled-field disagreement (D-002), the accelerationist objection (D-006), and the status of D3 as a rival driver rather than a sub-case of acceleration.

Turchin’s structural-demographic theory already carries a falsifiable mechanism for several symptoms (elite overproduction + popular immiseration + the wealth pump → instability and populism). Theory A must show that the adaptation gap explains variance that the wealth pump does not. Concretely:

In a model of symptom severity, include both G and an inequality/wealth-pump measure (Gini, top-income shares, elite-overproduction proxies). Theory A earns its place only if G carries significant, independent explanatory power with the wealth pump controlled. If G collapses to noise once inequality is included, D3 is the real driver and Theory A is redundant (falsification condition 2). If both carry independent weight, A and Turchin are complementary drivers, not rivals — which would itself resolve part of D-002 empirically rather than by debate.

This is the cleanest way the project can convert a standing disagreement into a test — exactly the move the whole enterprise is supposed to make (L-005: draw the arrows; don’t just name the target).

7. A concrete first study (pre-registered, runnable)

Small, honest, shippable (the plan’s discipline). Not the whole theory — one bounded, falsifiable slice.

• Units: ≈30 OECD countries (data availability; cross-nationally comparable).
• R_c: technology-adoption speed + digital-penetration growth + firm-turnover rate (normalized, equal-weighted).
• R_a: regulatory-response lag (inverse) + institutional-trust recovery + curriculum/skill-update lag (inverse) (normalized, equal-weighted).
• G = R_c − R_a, per country, over the available panel.
• Outcome: symptom-severity composite (loneliness + populist vote share + trust decline + affective polarization).
• Control: Gini / top-income share (the D3 discriminator).
• Tests: (a) cross-sectional correlation of G with the outcome; (b) panel regression with the inequality control and country/year fixed effects; (c) lag/lead structure (does G lead?); (d) robustness across alternative proxy sets and weightings.
• Pre-registration: proxy list, signs, weights, and hypotheses fixed and published before outcomes are examined.

This study is also the heavy-Code trigger. It needs real data ingestion, cleaning, normalization, and panel modeling — the executable, data-pipeline work that DIAGRAMS.md §7 gates at Gate 2. So operationalizing Theory A is not only the top intellectual task; it is the concrete thing whose arrival would justify heavy Code. Until this study is specified and pre-registered, heavy Code has nothing to compute; once it is, it does.

8. What this changes

• Falsifiability coverage. D1 (the gap) moves from “not yet” toward a stated, testable claim — the first of the low-coverage tests to be genuinely operationalized (METRICS.md §4). It does not yet count as tested; Q-001 stays open, now with a design attached rather than only an intention.
• A disagreement becomes a test. The A-vs-Turchin question (D-002/D-006) is reframed as an empirical horse-race with a control, not a debate to be preserved indefinitely.
• The causal loop is now drawable. With G and its couplings specified, Theory A’s dynamics can be drawn as a signed causal-loop diagram — see DIAGRAMS.md §9 — the bridge from prose to any runnable model.

9. Known weaknesses of the operationalization itself (skeptical close)

• The composite could still be tuned. Pre-registration and robustness checks reduce but do not eliminate the risk that the equal-weight composite hides a fragile result. The defense is transparency: publish the sensitivity analysis, not just the headline.
• Adaptive-throughput ≠ adaptive-success. Measuring the rate of adaptation, not its adequacy, is a deliberate simplification; a society can adapt fast in the wrong direction. The theory is explicitly about pace-matching, but the simplification should be revisited if the results are weak.
• OECD-only is a narrow world. The first study inherits exactly the vantage bias the panel flagged (D-005): it measures the crises of rich democracies with good data. Generalization beyond that is a separate, harder question, not a foregone conclusion.
• Correlation is not the mechanism. Even a clean G–symptom association with controls establishes that the gap tracks the symptoms, not that it produces them. The causal claim needs the lag structure, the mechanism (the causal loop), and eventually a runnable model that catches a flaw in its own prose — the Phase-3 bar.

None of these is fatal; all are stated so the next contributor knows exactly where to push. That is the point: a claim you can attack precisely is worth more than a frame you can only admire.

Author: Hulki Okan Tabak — with Claude · License: CC BY-SA 4.0