Needs Verification

Empirical Adequacy & Empirical Grounding (Adäquatheit)¶

Definition & Conceptual Goal¶

The Empirical Adequacy & Grounding metric quantifies the degree of agreement between empirical observation data (\(I\)) and theoretical models (\(M\)), taking into account inevitable measurement errors, experimental approximations, and idealized boundary conditions (Balzer et al., 1987, pp. 93, 215, 217; Stegmüller, 1976, pp. 40–45, 120).

It integrates two fundamental levels of empirical validation:

Baseline Empirical Grounding (True Partial Empirical Claim): Verifying that the theory is not a vacuous mathematical exercise, but possesses a non-empty set of historically validated paradigm applications (\(I_0 \neq \emptyset\)) where the central empirical claim holds (\(I_0 \in \operatorname{Cn}(K)\)).
Approximative Empirical Adequacy: Verifying that empirical deviations across the full intended domain remain strictly within a class of admissible blurs \(\mathcal{A}\) (\(\exists X \in \operatorname{Cn}(K) \text{ s.t. } I \approx_\mathcal{A} X\)), avoiding both unachievable idealized exactness (Strong Adequacy) and ad-hoc immunization.

Relationship to Tenability

While Tenability ensures the qualitative semantic precondition that empirical observation nodes can be conceptually typed and theoretically enriched into the non-theoretical vocabulary of the theory (\(I \subseteq M_{pp}\), governed by ADR 0015), Empirical Adequacy assesses whether the resulting enriched models (\(M\)) match empirical observations within admissible blur margins (\(\mathcal{A}\)).

Theoretical Grounding & Model Formulation¶

In structuralism, every theory-element \(T = \langle K, I \rangle\) asserts the empirical claim \(I \in \operatorname{Cn}(K)\). In real empirical science, continuous laws never match observations without experimental imprecision. Balzer, Moulines, and Sneed formalized this using a uniform topology / metric space \((\mathcal{M}_{pp}, d)\) over partial potential models.

Let \(\mathcal{A} \subseteq \mathcal{P}(M_{pp} \times M_{pp})\) be a family of admissible blur relations representing acceptable error margins (surrounding a uniform space). A theory-element \(T = \langle K, I \rangle\) is approximatively empirically adequate with respect to \(\mathcal{A}\) if and only if:

\[\exists X \in \operatorname{Cn} (K) \quad \text{such that} \quad I \approx_\mathcal{A} X\]

where \(I \approx_\mathcal{A} X \iff \forall i \in I \, \exists x \in X \, ((i, x) \in \mathcal{A}) \land \forall x \in X \, \exists i \in I \, ((i, x) \in \mathcal{A})\).

Mathematical Specification & Graph Formulation¶

Let \(d (i, x)\) be the normalized distance metric in the feature space of observation \(i\) vs. theoretical model \(x\).

Distance Metric \(d(i,x)\) & LLM Proxy

The distance function \(d(i,x)\) is defined abstractly to allow for domain-specific physical or statistical residuals. To satisfy the Adequacy Score (\(AS\)) formula. In text-based graph environments where raw data is unavailable, we substitute physical residuals with an LLM-derived Plausibility Score \(p(i,x) \in [0, 1]\). Because structural distance and semantic plausibility are inversely related, the score must be inverted before calculation:

\[d(i,x) = 1 - p(i,x)\]

Grounding Ratio (\(GR\) - Baseline Existence)¶

A theory must first satisfy the True Partial Empirical Claim criterion:

\[GR (T) = \frac{|I_{\text{confirmed}}|}{|I_{\text{total}}|}\]

\[\text{ClaimStatus} (T) = \begin{cases} 1 & \text{if } |I_{\text{confirmed}}| > 0 \land I_{\text{confirmed}} \subseteq \operatorname{Cn} (K) \\ 0 & \text{otherwise} \end{cases}\]

Where \(I_{\text{confirmed}}\) is the set of :EmpiricalApplication nodes with validated evidentiary support edges (:EVIDENCED_BY / :SUPPORTED_BY).

Adequacy Score (\(AS\))¶

\[\text{Adequacy} (I, K) = \max_{X \in \operatorname{Cn} (K)} \left[ 1 - \frac{1}{|I|} \sum_{i \in I} \min_{x \in X} d (i, x) \right]\]

Blur Margin Compliance Ratio (\(BMCR\))¶

Given a maximum allowable blur threshold \(\epsilon_{\text{blur}}\):

\[BMCR (I, K) = \frac{|\{i \in I \mid \min_{x \in X} d (i, x) \le \epsilon_{\text{blur}}\}|}{|I|}\]

Measurement & Graph Implementation¶

Verify Baseline Grounding (\(GR\)):

MATCH (t:TheoryElement {id: $id})-[:INTENDED_APPLICATION]->(i:Application)
OPTIONAL MATCH (i)<-[:EVIDENCED_BY]-(d:Dataset)
RETURN count(DISTINCT i) AS total_i, count(DISTINCT d) AS evidenced_count

Model Prediction Extraction: Retrieve quantitative or qualitative assertions predicted by :TheoryCore.
Observation Edge Residuals: Compute the difference between predicted node attributes and observed empirical node values in :EmpiricalData.
Blur Evaluation: Check if residual errors fall within declared tolerance intervals stored on :ErrorMargin edges.

Diagnostic & Metascientific Value¶

\(GR(T)\)	\(BMCR\) Score	Epistemic Status	Metascientific Interpretation
\(GR = 0\)	—	Pure Formal Speculation	Mathematical formalism lacking any verified empirical instantiation.
\(GR > 0\)	\(BMCR = 1.0\)	Approximatively Adequate	All empirical applications match theoretical predictions within tolerated blur limits.
\(GR > 0\)	\(0.5 \le BMCR < 1.0\)	Partial Anomalies	Certain applications exhibit systematic deviations; candidates for specialization or anomaly tracking.
\(GR > 0\)	\(BMCR < 0.5\)	Empirical Crisis / Refuted	Major empirical failure; theory core predictions persistently fall outside acceptable error margins.

Grounding References¶

[Balzer et al., 1987] Balzer, W., Moulines, C. U., & Sneed, J. D. (1987). An Architectonic for Science. Reidel Publishing, pp. 93, 209–217.
[Stegmüller, 1976] Stegmüller, W. (1976). The Structure and Dynamics of Theories. Springer-Verlag, pp. 40–45, 120, 215–220.