§12 Empirical calibration

Per-result calibration hooks · three priority studies · the reflexive measurement problem

The framework's claim to be more than diagnostic critique depends on its formal results being empirically testable. This section specifies what each formal result predicts about measurable quantities, what would confirm or refute each prediction, and what the reflexive measurement problem requires.

§12.1 Per-result calibration hooks

For each of the framework's eleven formal results: the predicted quantity, how it would be measured, what would confirm the prediction, what would falsify it. The hooks are intended for empirical researchers operationalizing the framework.

Lemma 1 — Foreclosure

Predicted quantity: Survival function S(τ; θ*) of inter-stimulus intervals.
Measurement: Engagement-trace dataset → empirical CCDF → evaluate at reflective threshold θ_ref (30s to a few minutes per Kahneman 2011; Killingsworth & Gilbert 2010).
Confirmation: Ŝ(θ_ref) ≤ 10⁻³ on mature platforms.
Falsification: Ŝ(θ_ref) > 10⁻² — structural-separation hypothesis mis-calibrated or platforms have not converged.

Theorem 1' — Reclamation as learnability lower bound

Predicted quantity: Variance of any estimator Î_T(τ).
Measurement: User-state proxies (biometric + behavioral); estimate autocovariance across no-stimulus windows; track estimator variance vs. sample size.
Confirmation: Var(Î_T) · T → ∞ as T grows.
Falsification: Var(Î_T) · T stabilizing — recovery is feasible.

Proposition 4' — Dividual

Predicted quantity: Filtering posterior entropy H(u_t | y_{1:t}) → 0.
Measurement: Shadow inference — external model on engagement traces predicting user-state proxies. Track predictive entropy vs. trace length.
Confirmation: Entropy contracts at parametric rate ε_t ∼ √(d log t / t).
Falsification: Entropy plateaus above noise floor.

Proposition 5' — Metric superego

Predicted quantity: D_KL(ν_u(t) || π_R*) → 0.
Measurement: Periodic survey/behavior-based measurement of user self-evaluative distribution; independent characterization of platform reward target; compute KL across time.
Confirmation: Decreasing KL across user tenure.
Falsification: KL remains positive and bounded away from zero.

Proposition 6 — Kuramoto entrainment

Predicted quantity: Phase coherence between body and platform rhythms.
Measurement: Body-rhythm data (HRV, circadian markers, sleep timing) + platform delivery timestamps; phase-coherence statistics.
Confirmation: Sustained phase coherence in heavy platform users; coherence drops in low-use cohorts.
Falsification: No phase coherence even in heavy users.

Proposition 7' — Libidinal routing

Predicted quantity: Cross-register mutual information I_{Π_t}(u^i; u^j) > 0.
Measurement: Tag engagement events by register; compute empirical MI between cross-register engagement profiles; compare to lower-platform-use baseline.
Confirmation: Strong cross-register MI in platform-active populations.
Falsification: No cross-register coupling even in heavy users.

Proposition 8 — Cohort gradient

Predicted quantity: D_KL(ν_u^init(τ_exp) || π_R*) exponentially decreasing in τ_exp.
Measurement: Cohort-stratified attentional autocorrelation, political-engagement timescales, self-evaluative distributions; fit exponential decay; estimate K.
Confirmation: Monotone exponentially-decreasing relationship.
Falsification: No cohort effect, or non-exponential form.

Theorem 9 — Causal sensitivity

Predicted quantity: ∂θ*/∂X for each mechanism; closure amplification (1 − κ_Φ)^(−1).
Measurement: Natural-experiment data from platform policy changes, regulatory interventions, algorithm rebuilds; document policy-parameter shift; estimate sensitivities.
Confirmation: Mode A O(1), individual Mode B/C O(1/N), regulatory Mode B/C O(1).
Falsification: Sensitivities deviate substantially from predicted scaling.

Theorem 10 — Mean-field + cascades

Predicted quantity: Branching ratio ρ(W) ρ(α) ∫κ; percolation threshold ρ_c.
Measurement: Estimate W from recommendation/social graph; estimate α from cross-register Hawkes fitting; compute branching ratio.
Confirmation: Mature platforms above percolation threshold; cascades propagate platform-wide.
Falsification: High-density platforms failing to cascade or low-density platforms cascading.

Theorem 11 — Stretched-exponential foreclosure

Predicted quantity: Sub-exponential index β̂ in Class S_β classification.
Measurement: Empirical CCDF → log-log Weibull tail fit → β̂.
Confirmation: β̂ ∈ (0, 1) on contemporary platforms; β̂ → 1 as platforms mature.
Falsification: β̂ ≈ 1 uniformly (Class L) or no clear Weibull tail.

Theorem 12 — Crisis-boundary diagnostics

Predicted quantity: Bifurcation regime classification of engagement time-series; estimated platform κ.
Measurement: Time-series of platform-aggregate engagement; Lyapunov exponents, recurrence quantification, spectral analysis, return-time statistics; estimate κ from intervention sensitivities.
Confirmation: Mature platforms near κ ≈ 1⁻; documented bifurcation events during transitions.
Falsification: κ ≪ 1 uniformly (no bifurcation behavior) or κ > 1 steadily without bifurcation signature.

§12.2 Three priority empirical studies

Three studies that would maximally test the framework with tractable resource requirements. Priority by combination of: sharpness of the predicted functional form, breadth of theoretical implication, feasibility.

Priority 1 — Cohort-gradient verification (Proposition 8)

Sharpest functional form, largest implication. Mid-five-figures to execute via cross-sectional cohort study with 5,000 stratified by age-of-first-platform- exposure.

Priority 2 — Foreclosure-rate calibration (Lemma 1 + Theorem 11)

Anchors the framework's spine. Mid-five-figures via multi-platform engagement-tracking app on 1,000 users.

Priority 3 — Bifurcation diagnostics (Theorem 12)

Most ambitious; feasible on public aggregate data. Low five-figures for researcher analysis time.

§12.3 The reflexive measurement problem

The framework predicts that the cohort whose interiority has been foreclosed cannot reliably measure its own foreclosure: the apparatus the analysis would need (self-report, attentional autocorrelation, deliberative reflection) is the apparatus whose constitutive structure has been removed.

The reflexive measurement problem is specific to the closure. Older traditions mapped the geometry of self-reference — Heraclitus's eye that cannot see itself, psychoanalysis's unconscious beyond conscious report, Heisenberg's measurement that perturbs the measured. The framework names something sharper: a specific structural prediction about what closed-loop optimization does to the empirical apparatus through which a foreclosed cohort would try to test the foreclosure claim. The framework predicts that the apparatus — self-report, attentional autocorrelation, deliberative reflection — operates at the time-scales Lemma 1 has driven to operational zero. The cohort whose foreclosure the framework predicts cannot use the older apparatus to test the prediction.

Three concrete consequences:

Self-report instruments are systematically biased. Surveys of $ν_{u}$ rely on capacities the framework predicts have been formatted by platform articulation grammars. Behavioral and biometric measures are required.
Cohort comparisons depend on having a reference cohort. Proposition 8 requires comparing native to pre-platform cohorts; both are partially foreclosed on the framework's prediction. Historical data or non-platform-using controls are biased.
The researcher is in the loop. Researchers are platform users with cohort-position- shaped interiority. The framework's predictions about cohort-bounded capacity apply to researchers as well.

Mitigations: pre-registered analyses with explicit falsification criteria; replication across research teams from different cohort positions; triangulation across self-report, behavioral, biometric measures; naturalistic experimental designs; adversarial collaboration.

The reflexive problem has no in-principle solution. Explicit methodological practice can partially mitigate it. The framework's empirical credibility depends on this practice being adopted by the research community.