Theme: Research on the Cutting-Edge Applications of Generative Artificial Intelligence in Accounting and Financial Analysis

Speaker: Professor Vernon Richardson, Walton College of Business, University of Arkansas

Host: Professor He Li, School of Accounting, Southwestern University of Finance and Economics

Time: 14:00 p.m. –16:00 p.m. , April 8, 2025 , Tuesday

Place: Room 650, Chengzheng Building, Liulin Campus

Sponsor: Academic Research Office, School of Accounting

Introduction of the speaker:

Professor Vernon Richardson is a Distinguished Professor and Chair of the Department of Accounting at the Walton College of Business, University of Arkansas, USA. His research focuses on the intersection of information technology, big data analytics, and accounting. He is one of the most outstanding scholars in the field of accounting information systems in the United States.

His research has been published in UTD 24 journals such as The Accounting Review, Journal of Accounting and Economics, MIS Quarterly, Journal of Operations Management, Journal of Marketing, and Journal of Operations Management, as well as FT50 journals such as Contemporary Accounting Research and Journal of Management Information Systems. He has served as the editor-in-chief of The Accounting Review and currently serves as the editor-in-chief of Accounting Horizons.

In terms of teaching, Professor Richardson has published textbooks such as Accounting Information Systems and Data Analytics for Accounting. Among them, the textbook Data Analytics for Accounting is an original work in the field of big data accounting, which has been widely used both domestically and internationally.

Abstract:

To encourage in-depth thinking and communication, participating students are invited to select an academic paper related to AI and prepare a 5-minute informal presentation focusing on the paper’s “incremental contribution”—i.e., what novel insights it offers compared to existing research and what unresolved issues it addresses. This session aims to foster interdisciplinary dialogue in a relaxed setting, rather than a formal presentation.

In recent years, generative artificial intelligence (Generative AI) has demonstrated significant potential in information processing and financial decision-making. We have selected two recent studies from arXiv that highlight breakthrough applications of large language models (LLMs) in accounting information interpretation and financial statement analysis. Colleagues are warmly invited to join the discussion.

Paper 1: Bloated Disclosures: Can ChatGPT Help Investors Process Information?

This study uses the stock market as an experimental platform to empirically analyze ChatGPT’s effectiveness in summarizing complex corporate disclosure documents. The findings reveal that AI-generated summaries, while shorter in length, contain higher information density and exhibit stronger explanatory power for market reactions. Text sentiment is amplified in the summaries, and “bloated disclosures” are shown to reduce price efficiency and increase information asymmetry. The paper introduces a novel quantitative metric for “bloat degree,” demonstrating generative AI’s potential to alleviate investors’ information processing burdens.

Paper 2: Financial Statement Analysis with Large Language Models

This paper examines GPT-4’s performance in standardized, anonymized financial statement analysis tasks. Even without industry context or narrative information, GPT-4 surpasses human analysts in predicting future earnings trends and matches the accuracy of state-of-the-art machine learning models. LLMs demonstrate particular advantages in scenarios where human analysts typically underperform. The study further indicates that the model’s predictions rely on analytical reasoning rather than training data memorization, and its constructed investment strategies outperform traditional models, highlighting its promise for decision support.