讲座题目： Man + machines: assessing the informativeness of critical audit matters identified by AI and human auditors （人类与机器：评估由人工智能和人类审计师识别的关键审计事项的信息性）

主讲人：郭丰 爱荷华州立大学

时间：2024年12月19日10:00

地点：学院206

主办单位：yL23411永利官网登录会计系

邀请人：潘红波

内容摘要：

We explore the potential value of generative artificial intelligence (AI) in identifying critical audit matter (CAM) topics and assess the informativeness of CAMs identified by AI and human auditors. We compare AI-generated CAMs with those reported by human auditors to evaluate their effectiveness in signaling financial reporting risks to investors. We further discern the overlapping and non-overlapping information between AI-generated CAMs and auditor-reported CAMs to investigate the (dis)advantages of AI relative to human auditors. While AI has superior information-processing capabilities and is less affected by human biases, it may be less informative than human auditors due to the lack of professional judgment and access to private client information. Our findings indicate that AI can predict approximately 37% of CAMs reported by auditors. Auditor-reported CAMs or AI-generated CAMs alone are generally not significantly associated with financial reporting risk or abnormal stock returns. However, the overlapping CAMs identified by auditors and generated by AI are significantly associated with financial reporting risk and abnormal stock returns, whereas CAMs identified exclusively by auditors or AI are not, suggesting a complementary role of generative AI and human auditors in identifying CAMs. Overall, our findings indicate that combining AI and auditor reports can help investors identify significant financial reporting risks and also highlight the information value of CAMs reported by human auditors.

我们探讨生成式人工智能（AI）在识别关键审计事项（CAM）主题中的潜在价值，并评估由AI和人类审计师识别的CAM的信息性。我们将AI生成的CAM与人类审计师报告的CAM进行比较，以评估它们在向投资者揭示财务报告风险方面的有效性。同时，我们分析AI生成的CAM与审计师报告的CAM之间的重叠信息和非重叠信息，探讨AI相较于人类审计师的优势与不足。尽管AI具有卓越的信息处理能力，且不受人类偏见的影响，但由于缺乏专业判断和对客户私密信息的访问，其信息性可能不如人类审计师。研究结果表明，AI可以预测约37%的审计师报告的CAM。仅由审计师报告的CAM或仅由AI生成的CAM通常与财务报告风险或异常股票收益无显著关联。然而，由审计师和AI同时识别的重叠CAM与财务报告风险和异常股票收益显著相关，而仅由审计师或AI单独识别的CAM则没有显著关联。这表明生成式AI与人类审计师在识别CAM方面具有互补作用。总体而言，我们的研究表明，结合AI和审计师报告可以帮助投资者更好地识别重要的财务报告风险，同时也凸显了人类审计师报告的CAM的信息价值。

主讲人简介：

郭丰，爱荷华州立大学教授。主要研究领域为公司治理，人工智能，并购，与审计，其研究成果发表在UTD24/FT50 国际顶级期刊 The Journal of Finance, The Review of Financial Studies, The Journal of Financial and Quantitative Analysis, Contemporary Accounting Research, Information Systems Research等。现担任Managerial Auditing Journal副主编及多本国际期刊包括The Accounting Review, Contemporary Accounting Research, Management Science, Journal of Financial and Quantitative Analysis, Auditing: A Journal of Practice and Theory的匿名审稿人。