The Disciplinary Language Transfer Problem: How Psychological Vocabulary Produces Governance Failures in AI Agent Deployment

{'Problem': 'The paper identifies a gap in the literature regarding the impact of psychological vocabulary on AI governance, termed the disciplinary language transfer problem. This issue is particularly relevant in the context of AI agent deployment, where misinterpretations stemming from specialized language can lead to governance failures. The authors argue that existing frameworks do not adequately address these linguistic challenges, which can hinder effective communication and decision-making in AI governance. The work is presented as a preprint, indicating it has not yet undergone peer review.', 'Method': "The authors analyze several theoretical frameworks to understand the implications of disciplinary language in AI governance. These frameworks include Wittgenstein's language games, which emphasize the contextual nature of language; Kuhn's paradigm observation, which discusses how scientific paradigms shape understanding; Haraway's situated knowledge, which highlights the importance of context in knowledge production; and Star and Griesemer's boundary object theory, which focuses on how different communities can collaborate despite differing terminologies. The core output of this analysis is a Disciplinary Audit, which consists of a governance document scan accompanied by a translation taxonomy that includes thirty-seven specific terms relevant to the psychological vocabulary in question.", 'Results': 'The paper provides an abridged form of the Disciplinary Audit and the translation taxonomy, but it does not report any quantitative results or benchmarks against existing methodologies.', 'Limitations': 'The authors note that the reliance on psychological vocabulary can lead to governance failures, particularly when assumptions derived from this vocabulary do not apply to AI systems lacking developmental continuity. This limitation suggests that the findings may not be universally applicable across all AI contexts, potentially restricting the generalizability of the proposed solutions.', 'Why it matters': 'This work has significant implications for downstream research and practice in AI governance. By highlighting the challenges posed by disciplinary language, it calls for a reevaluation of communication strategies within AI development and deployment. The proposed translation taxonomy could serve as a foundational tool for improving interdisciplinary collaboration and ensuring that governance frameworks are more inclusive and effective in addressing the complexities of AI systems.'}