Analytical Lenses

The criteria in the evaluation pipeline do not come from nowhere. Each one is grounded in a specific intellectual tradition that developed it for a reason — usually because a field got burned by ignoring it. Falsifiability is a criterion because philosophy of science spent a century learning what happens when claims cannot be disconfirmed. Necessity-plus-sufficiency is a criterion because neuroscience spent decades discovering that necessity alone does not establish implementation. Dose-response is a criterion because pharmacology learned that single-dose measurements hide compensatory dynamics.

We call these traditions lenses. A lens is not a method or a metric — it is an analytical vocabulary, a way of thinking about evidence that shapes which questions you ask and how you interpret the answers. The framework draws on eleven lenses from five fields.

Core lenses

Five lenses map one-to-one to the five validity types. Each core lens provides the conceptual foundation for its validity type’s criteria — the “why” behind the formal specifications on the validity type pages.

Core lens	Validity type	What it contributes
Philosophy of Science	Construct	Falsifiability, convergent validity, the distinction between observables and theoretical entities
Measurement Theory	Measurement	Reliability, invariance, the MTMM matrix, the distinction between the construct and the instrument
Neuroscience	Internal	Single vs double dissociation, lesion vs stimulation, constitutive relevance, multimodal parcellation
Pharmacology	External	Dose-response curves, functional selectivity, receptor reserve, the distinction between affinity and efficacy
Mechanistic Interpretability	Interpretive	Description vs explanation, faithfulness vs understanding, Marr’s levels, the description-mode hierarchy

You do not need to read the lens pages to use the framework. The criteria stand on their own with formal pass conditions and thresholds. But when a criterion seems arbitrary — why require a double dissociation and not just a single one? why demand cross-family convergence and not just within-family agreement? — the lens page explains the intellectual history that motivates it.

Supporting lenses

Six additional lenses provide analytical tools that strengthen specific validity types without defining new ones. They are not mapped to validity types one-to-one; instead, each contributes methods or concepts that apply across multiple types.

Supporting lens	What it contributes	Validity types it supports
Control Theory	Stability margins, settling depth, observability conditions, PID-inspired steering	Internal, External
Dynamical Systems	Renormalization group flow, Koopman/DMD analysis, topological data analysis, critical phenomena	Construct, Measurement
Economics	Arbitrage search (can an alternative circuit substitute?), game-theoretic interaction models, price-impact linearity	Construct, Internal
Genetics	Knockout hierarchies, rescue experiments, Mendelian randomization, dose-response from molecular biology	Internal, External
Geometry	Fisher-Rao distances, angular steering, sheaf consistency, symmetry and equivariance	Measurement, Construct
Information Theory	Mutual information, transfer entropy, PID, information bottleneck, Granger causality	Internal, Construct

Supporting lenses are the source of many of the framework’s protocols. The genetics lens contributes 16 molecular-biology protocols (knockout hierarchies, rescue experiments, Mendelian randomization, and others adapted from experimental biology). The economics lens contributes arbitrage search — a protocol that tests construct validity by asking whether an alternative set of components can substitute for the claimed circuit. If yes, the circuit is not uniquely necessary; the construct may not carve computation at its joints.

Complete lens-to-criteria map

The table below shows every lens, the validity type it grounds (for core lenses) or supports (for supporting lenses), and the specific criteria it motivates. This is the full picture — every criterion in the framework traces back to at least one lens.

Lens	Type	Validity type	Criteria	Key question
Philosophy of Science	Core	Construct	C1 Falsifiability, C2 Structural plausibility, C3 Task specificity, C4 Minimality, C5 Convergent validity	Is the entity you named a real thing?
Measurement Theory	Core	Measurement	M1 Reliability, M2 Invariance, M3 Baseline separation, M4 Sensitivity, M5 Calibration, M6 Construct coverage	Can you trust the number?
Neuroscience	Core	Internal	I1 Necessity, I2 Sufficiency, I3 Specificity, I4 Consistency, I5 Confound control	Does the component implement the computation?
Pharmacology	Core	External	E1 Intervention reach, E2 Graded response, E3 Selectivity, E4 Effect magnitude, E5 Robustness, E6 Cross-architecture	Does it generalize?
Mechanistic Interpretability	Core	Interpretive	V1 Level declaration, V2 Level-evidence match, V3 Narrative coherence, V4 Alternative exclusion, V5 Scope honesty	Is the story right?
Control Theory	Supporting	Internal, External	Settling depth, stability margins, observability	Is the circuit steerable and stable?
Dynamical Systems	Supporting	Construct, Measurement	Koopman/DMD modes, renormalization flow, TDA	Does the structure persist across scales?
Economics	Supporting	Construct, Internal	Arbitrage search, game-theoretic interactions	Can something else substitute for it?
Genetics	Supporting	Internal, External	Knockout hierarchies, rescue, Mendelian randomization	Does it behave like a genetic pathway?
Geometry	Supporting	Measurement, Construct	Fisher-Rao distances, sheaf consistency, equivariance	Does the geometry match the computation?
Information Theory	Supporting	Internal, Construct	MI, transfer entropy, PID, Granger causality	Does information flow through it?

How lenses relate to the pipeline

Lenses sit outside the five-step evaluation pipeline. They are reference material — the intellectual context that explains why the pipeline’s criteria exist. A researcher who wants to evaluate a claim follows the pipeline (scope → evidence → criteria → verdict). A researcher who wants to understand why a criterion matters, or who wants to design a new protocol, reads the relevant lens.

The relationship is: the pipeline is how you evaluate; lenses are how you think.