Epistemic capture through specialization in post-World War II parliamentary debate

Abstract

This article investigates how parliamentary debate in the Dutch House of Representatives (“Tweede Kamer”) (1945–1995) narrowed as MPs turned into domain specialists. We call this narrowing epistemic capture: a few experts progressively bound what can be said. To detect epistemic capture, we deploy a three-layer computational pipeline. Latent-Dirichlet topic modeling converts 8.2 million sentences into 250 semantic themes; Pointwise Mutual Information networks connect themes within six-month windows; Louvain clustering traces the birth, drift and endurance of topical communities. Capture appears on every scale. Macro-level: network modularity almost doubles after 1960 while density falls, marking compartmentalized debate. Meso-level: cabinet turnovers act as “reset switches”: topic-neighborhood similarity drops in the half-year after a new coalition forms, then anneals along partisan lines. Micro-level: enduring communities – foreign policy, agriculture and education – lock topics and MPs together for decades, yet resistance to capture is visible in distinct contentious topics. These multiscale patterns show how 20th-century Dutch parliamentary debate saw a rise of technical specialism that significantly constrained the breadth of political debate. Methodologically, the study demonstrates the value of structural (network) distant reading over purely lexical counts and offers a transferable workflow for measuring how democratic discourse undergoes structural transformations.