Tree / Hierarchical Conformity¶

Definition & Conceptual Goal¶

The Hierarchical Conformity metric (or Theory-Tree Conformity) evaluates whether a directed specialization graph strictly conforms to a single-rooted hierarchical branching structure (Balzer et al., 1987, p. 175).

In the structuralist architecture of scientific theories, a mature and well-founded scientific theory does not have multiple conflicting foundational cores; rather, it is anchored in a unique, fundamental root theory-element (\(T_0\)) from which all further domain-specific laws branch out via successive specializations. This metric measures the macroscopic structural integrity of this deductive specialization hierarchy.

Theoretical Grounding, Graph Granularity & The Nature of \(T_0\)¶

A critical distinction must be made regarding the granularity of the graph. A theory-net is not a single atomic proposition or isolated equation. It is a macro-logical Theory-Element defined as \(T = \langle K, I \rangle\), where \(K\) is the theoretical core and \(I\) represents the intended applications.

The core (\(K = \langle M_p, M, M_{pp}, C, L \rangle\)) encapsulates a set of interdependent fundamental laws (which jointly define the actual models \(M\)). Internally, these fundamental laws are non-hierarchical; they combine via logical conjunction and mutually presuppose one another.

Consequently, the root element \(T_0\) is inherently a clustered group of cooperating axioms (e.g., Newton's laws of motion, absolute space, and universal gravitation operating together as a unified paradigmatic core). If this metric is applied to a micro-level knowledge graph where individual axioms are modeled as separate nodes, applying the hierarchy metric directly is epistemically invalid. Foundational cooperating axioms must first be aggregated into a single macro-node \(T_0\).

Prerequisites for \(T_0\): Micro-Cohesion & Validity¶

To prevent the artificial grouping of disjoint axioms into a "fake" \(T_0\) merely to achieve a perfect hierarchy score, the micro-cohesion of the foundational set must be validated before the macro-metric is applied.

For a set of individual laws or axioms to be legitimately grouped into the root core \(K_0\), they must satisfy strict internal coherence criteria evaluated by separate micro-level graph metrics:

1. Logical Consistency: The elements must not contradict one another.
2. Mutual Constraint (Cross-Binding): The axioms must share theoretical terms and jointly constrain the intended applications. In a directed dependency graph, they should form a Strongly Connected Component (SCC) of mutual presupposition.
3. High Internal Density: The propositions within \(T_0\) must exhibit dense horizontal relations (e.g., PRESUPPOSES, SHARES_VARIABLE, CONSTRAINS).

If a proposed root cluster fails these micro-cohesion checks, it cannot be legitimately grouped into a single \(T_0\). The theory is technically fragmented, and the root count \(\vert{}B(N)\vert{}\) must reflect multiple independent cores.

Mathematical Specification & Graph Formulation¶

Let \(G = (V, E)\) be the Directed Acyclic Graph (DAG) of specialization relations where directed edges \((u, v) \in E\) indicate \(u \text{ specializes } v\) (or in parent-child representation, \((v, u) \in E_{spec}\) where \(v\) is generalized into child \(u\)). It is assumed that \(G\) is a proper theory-net where valid foundational clusters have been condensed into macroscopic theory-elements.

We define the set of top-elements (or minimal elements with respect to generalization, i.e., root elements) as \(B(N)\):

A connected theory-net \(N\) is a Theory-Tree (Theorie-Baum) if and only if:

1. \(B(N)\) is a singleton: \(\vert{}B(N)\vert{} = 1\), denoted as \(B(N) = \{T_0\}\).
2. The transitive reduction of \(\alpha\) forms an out-tree (arborescence) rooted at \(T_0\), meaning every non-root node \(T_k \in T \setminus \{T_0\}\) has exactly one immediate predecessor in the specialization hierarchy.

Tree Conformity Score¶

We define the metric score \(S_{tree}(G) \in [0, 1]\) as:

Where:

• \(\vert{}B(N)\vert{}\) is the number of root macro-nodes (nodes with 0 in-degree in parent-to-child orientation).
• \(\text{in-deg}_{spec}(u)\) is the number of immediate generalizing parents for node \(u\).

Measurement & Graph Implementation¶

1. Micro-Cohesion Validation & Graph Condensation: If operating on a micro-level proposition graph, verify the mutual constraint of foundational axioms. Collapse valid, strongly connected foundational components (SCCs) into a single macro-node \(T_0\).
2. Root Identification: Compute the in-degree in parent-to-child orientation for all nodes in the theory-net subgraph. Filter nodes with in-degree equal to 0 to establish \(B(N)\).
3. Branching Factor & Multiple Inheritance: Count the number of nodes that have multiple parent specializations. In property graphs, multiple parents indicate either poly-specialization (legitimate cross-specialization from two cores) or structural ambiguity requiring consolidation.
4. Neo4j Cypher Check (Macro-Level):

   MATCH (t:TheoryElement {theory_id: $theory_id})
   WHERE NOT (t)-[:SPECIALIZES]->(:TheoryElement {theory_id: $theory_id})
   RETURN count(t) AS root_count
   

Diagnostic & Metascientific Value¶

Grounding References¶

• [Balzer et al., 1987] Balzer, W., Moulines, C. U., & Sneed, J. D. (1987). An Architectonic for Science. Reidel Publishing, pp. 172–176.
• [Stegmüller, 1976] Stegmüller, W. (1976). The Structure and Dynamics of Theories. Springer-Verlag, pp. 169–171.