AI and the production of truth in accounting practices: a Foucauldian analysis
本文基于福柯治理术理论，结合作者在四大审计事务所和会计教学的经验，分析了AI如何将会计中的真理从与事实对应转向概率可信，并指出专业人士因降低职业风险而主动接受算法治理，但需制度支持以避免合规侵蚀公共责任。
This reflexive essay examines how artificial intelligence (AI) is transforming the production of truth in accounting practice. Drawing on Foucault’s analytics of governmentality and on the author’s experience as a Big Four audit director and accounting professor, the essay argues that AI-based systems are assembling a new accounting apparatus, one that shifts truth claims from representational correspondence with documented reality to probabilistic plausibility within learned distributions. Evidence, materiality, and professional judgment are being reorganized around model outputs, risk scores, and algorithmic thresholds. The essay’s distinctive contribution is to show that this transformation works through complicity rather than coercion: professionals embrace algorithmic governance because the system makes compliance the path of lower professional exposure than the exercise of contextual judgment. This willingness is itself a governmental effect. The essay identifies the institutional conditions needed to keep probabilistic truth open to contestation, arguing that individual professional courage is necessary but insufficient without supporting structures in standards, education, regulation, and organizational practice. Unless the profession builds such conditions, accounting risks narrowing into self-referential algorithmic compliance, eroding the public accountability on which its legitimacy depends.