A Systems Theory for Today

A living, forkable attempt to build a systems theory adequate to the present — by designed disagreement.

PANEL ROSTER AMENDMENT — a measurement / causal-inference seat

Version: 1.0 · Status: RATIFIED (author, Session 4e); applied to panel/PANEL_ROSTER.md. · Last updated: Session 4e (placed by Code) Amends: panel/PANEL_ROSTER.md (the Advisory Twenty → Twenty-two). Prompted by the recommended catch in studies/study-B-optimization/PRE_REGISTRATION.md §10 (logged as C-014).

The gap. The panel was composed for philosophical deliberation — the nature, history, and meaning of total systems. The move into empirical studies (A, B, C) exposed a missing competence: causal identification, construct validity, and quasi-experimental design. Turchin partly covers it (falsifiability, retrodiction), but on the mechanics of “is this construct valid and is this effect identified from this data,” the panel had to summon a voice ad hoc during the Study-B deliberation. This amendment makes that competence standing — and, in the project’s own spirit, adds it as two voices chosen to disagree with each other, not one house methodologist.

Proposed additions to the Advisory Twenty

21. Donald T. Campbell (1916–1996) — the conscience of measurement and quasi-experiment.

  • Brings: the honest toolkit for social questions that cannot be randomized — quasi-experimental design (interrupted time series, regression discontinuity, the named catalogue of threats to validity); construct validity (does your measure measure the thing, or a convenient shadow of it?); and Campbell’s Law — “the more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures” — which is, uncannily, Theory B’s mechanism stated as a law of measurement. Also evolutionary epistemology: knowledge grows by blind variation and selective retention (a deep cousin of the project’s catches→learnings loop).
  • Key works: Experimental and Quasi-Experimental Designs for Research (with Stanley); “Reforms as Experiments”; “Assessing the Impact of Planned Social Change” (Campbell’s Law).
  • Picks fights with: Pearl (your formal DAGs assume you already know the structure; real social data is a minefield of confounds you haven’t modeled — name the threats to validity before you write the do-operator); Turchin (retrodiction is good, but construct validity comes first — are “elite overproduction” and “the gap” measured or reified?); and the project itself (are O, P, G, and “a catch” valid constructs, or are you counting shadows and calling it evidence?). Campbell is the panel’s demand that a number mean something before it is trusted.

22. Judea Pearl (b. 1936) — the formalizer of causation.

  • Brings: the machinery to say what causal claim is identifiable from what data — the ladder of causation (association → intervention → counterfactual), directed acyclic graphs, the do-calculus, confounder/collider logic. Directly relevant to the pressure-test causal hypothesis, which is a DAG: Pearl turns “draw the arrows” into “state which arrows are identifiable, and what would confound each.”
  • Key works: Causality; The Book of Why (with Mackenzie).
  • Picks fights with: Campbell (your “threats to validity” are ad hoc where a formal model would be systematic — a DAG tells you exactly which confounds matter and which are irrelevant, no catalogue needed); correlational social science (“no causes in, no causes out” — a regression is not a cause); and the project’s arrows (you drew a causal graph — now commit to the identification claims, or admit the arrows are associational). Pearl is the panel’s demand that causal language be earned by a stated identification strategy.

Why two, and why these two

Campbell and Pearl disagree productively — the pragmatist of imperfect data against the formalist of causal structure — which is exactly the designed-disagreement pattern the panel runs on. On any empirical study the project attempts, Campbell will ask “is the construct valid and what are the threats?” and Pearl will ask “what is your identification strategy and DAG?”, and the tension between those two questions is where good empirical design lives. Neither alone would have caught the full set of problems in the Study-B deliberation; together they cover the competence the philosophical roster lacked.

Scope of the amendment

  • These are advisors (on call), not core members — the core ten remain the philosophical/historical/systems spine.
  • Summon them for: any study’s pre-registration; any causal claim in the pressure-test graph; any construct-validity question about O, P, G, or the catch taxonomy; and any place a number is about to be trusted.
  • Like all panelists, they are reconstructed positions, not the persons (Ground Rule 16).

Recommended catch (logged as C-014)

Log as a catch: “the panel was composed for philosophical deliberation and lacked a measurement/causal-inference competence; the empirical turn (studies A/B/C) exposed the gap, which was first patched ad hoc in the Study-B deliberation and is now proposed as a standing amendment (Campbell + Pearl).” Class: process. Consequence: this amendment; a standing practice of summoning the measurement seat whenever a study or a causal claim is on the table.

Ratified amendment (author, S4e); applied to panel/PANEL_ROSTER.md (Advisory 20 → 22). Author: Hulki Okan Tabak — with Claude · License: CC BY-SA 4.0.