Using AI to Identify Exogenous Shocks and Conduct Archival Accounting Research
通过实验展示了AI在会计研究中挖掘美国证券法规作为经济冲击、测试其对自愿披露的因果效应，并撰写完整论文的能力与风险。
Abstract We explore the capabilities and dangers of artificial intelligence (AI) usage in accounting research. We focus on mining U.S. securities regulations as an economic shock, testing the causal effect of these shocks on U.S. firms’ voluntary disclosure, and writing a complete academic paper, conditional upon finding statistically significant results. Overall, this research experiment demonstrates the capacity for AI to provide efficiencies in research. AI-generated papers are not ready to be submitted to top accounting journals, but they constitute a useful starting point for accounting researchers and using AI saves valuable time in identifying exogenous shocks that significantly affect an outcome of interest. When we repeat the same experiment to explore the causal effect of non-U.S. securities regulations on U.S. firms’ voluntary disclosure practices, AI writes many professional looking papers with spurious results and unsubstantiated economic arguments, highlighting the potential dangers of AI for accounting scholarship.