DISCIPLINARY TRIBES AND CAUSAL STORIES: SUB-FIELD VARIATION IN THE NOVEL-ENTITY GOVERNANCE LITERATURE

The initial project aim was to synthesize the causal mechanisms of governance failure from the core scholarly canon on novel entities. However, the analysis revealed a deeper epistemic fracture: scholars from different disciplines do not merely disagree on solutions; they construct fundamentally different causal explanations of the same empirical problem. For instance, anthropologists emphasize regulatory capture by powerful industries (Moore, 2015), legal scholars highlight structural flaws in national frameworks like the U.S. Toxic Substances Control Act (Adler & Wells, 2023), chemists stress the physical persistence of these substances (Kwiatkowski, 2020), and science and technology studies (STS) researchers foreground failures to apply precautionary principles (Cordner, 2016). The governance crisis, therefore, is not only institutional but also anthropological: it reflects how knowledge is culturally produced and bounded within disciplinary silos. This finding emerged from a systematic review of 71 high-impact sources published between 1990 and 2025, representing the minimal complete canon of theoretical and empirical work on novel-entity governance failure. The corpus spans 21 fields, including anthropology, environmental law, political ecology, STS, and planetary-boundary science. From these, 420 atomic causal statements were extracted: 378 supporting claims and 42 internal counterclaims. Each statement was weighted by credibility (0.0–1.0) based on explicitness and relevance to novel entities, then netted to produce directional strengths between six governance mechanisms: regulatory capture, inadequate international coordination, national framework flaws, challenges with dispersed and persistent pollutants, short-term political influences, and failure to apply precautionary approaches. The corpus was sliced by discipline, pollutant type, publication date, and geography to test for variation. Both ChatGPT, a large language model developed by OpenAI, and Grok, an AI assistant developed by xAI, were used to rapidly identify, digest, and test the literature corpus. Further collaboration with Grok facilitated initial claim extraction and weighting, as well as synthesis of results, with pending manual verification and reconciliation to ensure accuracy. This human-AI interaction highlighted how computational tools can augment ethnographic methods in analyzing large literatures, though all final interpretations and decisions were human-led.