Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Can ChatGPT Overcome Behavioral Biases in the Financial Sector? Classify-and-Rethink: Multi-Step Zero-Shot Reasoning in the Gold Investment

Shuoling Liu, Gaoguo Jia, Yuhang Jiang, Liyuan Chen, Qiang Yang

TL;DR

The paper addresses how to mitigate framing-induced behavioral biases in financial decision-making by leveraging ChatGPT with a novel two-stage prompting framework called Classify-and-Rethink (CAR). CAR first classifies gold-news items into six categories and scores them, then prompts the model to rethink the scoring from a long-term perspective, reducing susceptibility to framing effects. Back-tested on Au9999.SGE data and dyhjw.com gold news (2018–2023), CAR outperforms baseline strategies with an 80.35% return and a Sharpe ratio of 1.071, while producing a near-normal distribution of scores that suggests more stable reasoning. The work demonstrates that structured prompt design can enhance financial reasoning in LLMs and offers a foundation for future, ethically-aware deployment of AI-driven investment analysis. Overall, CAR provides a practical, explainable approach for using LLMs to generate investment opinions while mitigating common behavioral biases in financial markets.

Abstract

Large Language Models (LLMs) have achieved remarkable success recently, displaying exceptional capabilities in creating understandable and organized text. These LLMs have been utilized in diverse fields, such as clinical research, where domain-specific models like Med-Palm have achieved human-level performance. Recently, researchers have employed advanced prompt engineering to enhance the general reasoning ability of LLMs. Despite the remarkable success of zero-shot Chain-of-Thoughts (CoT) in solving general reasoning tasks, the potential of these methods still remains paid limited attention in the financial reasoning task.To address this issue, we explore multiple prompt strategies and incorporated semantic news information to improve LLMs' performance on financial reasoning tasks.To the best of our knowledge, we are the first to explore this important issue by applying ChatGPT to the gold investment.In this work, our aim is to investigate the financial reasoning capabilities of LLMs and their capacity to generate logical and persuasive investment opinions. We will use ChatGPT, one of the most powerful LLMs recently, and prompt engineering to achieve this goal. Our research will focus on understanding the ability of LLMs in sophisticated analysis and reasoning within the context of investment decision-making. Our study finds that ChatGPT with CoT prompt can provide more explainable predictions and overcome behavioral biases, which is crucial in finance-related tasks and can achieve higher investment returns.

Can ChatGPT Overcome Behavioral Biases in the Financial Sector? Classify-and-Rethink: Multi-Step Zero-Shot Reasoning in the Gold Investment

TL;DR

The paper addresses how to mitigate framing-induced behavioral biases in financial decision-making by leveraging ChatGPT with a novel two-stage prompting framework called Classify-and-Rethink (CAR). CAR first classifies gold-news items into six categories and scores them, then prompts the model to rethink the scoring from a long-term perspective, reducing susceptibility to framing effects. Back-tested on Au9999.SGE data and dyhjw.com gold news (2018–2023), CAR outperforms baseline strategies with an 80.35% return and a Sharpe ratio of 1.071, while producing a near-normal distribution of scores that suggests more stable reasoning. The work demonstrates that structured prompt design can enhance financial reasoning in LLMs and offers a foundation for future, ethically-aware deployment of AI-driven investment analysis. Overall, CAR provides a practical, explainable approach for using LLMs to generate investment opinions while mitigating common behavioral biases in financial markets.

Abstract

Large Language Models (LLMs) have achieved remarkable success recently, displaying exceptional capabilities in creating understandable and organized text. These LLMs have been utilized in diverse fields, such as clinical research, where domain-specific models like Med-Palm have achieved human-level performance. Recently, researchers have employed advanced prompt engineering to enhance the general reasoning ability of LLMs. Despite the remarkable success of zero-shot Chain-of-Thoughts (CoT) in solving general reasoning tasks, the potential of these methods still remains paid limited attention in the financial reasoning task.To address this issue, we explore multiple prompt strategies and incorporated semantic news information to improve LLMs' performance on financial reasoning tasks.To the best of our knowledge, we are the first to explore this important issue by applying ChatGPT to the gold investment.In this work, our aim is to investigate the financial reasoning capabilities of LLMs and their capacity to generate logical and persuasive investment opinions. We will use ChatGPT, one of the most powerful LLMs recently, and prompt engineering to achieve this goal. Our research will focus on understanding the ability of LLMs in sophisticated analysis and reasoning within the context of investment decision-making. Our study finds that ChatGPT with CoT prompt can provide more explainable predictions and overcome behavioral biases, which is crucial in finance-related tasks and can achieve higher investment returns.

Paper Structure

This paper contains 17 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Motivation
  4. Datasets
  5. Prompt Design
  6. Back-testing Strategy
  7. Experiments
  8. Details
  9. Main Results
  10. Ablation Study
  11. Case Study
  12. Discussion
  13. Related Work
  14. Large Language Models
  15. Financial Reasoning
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Framework of our method CAR. There are two main components: (1) Classify. The news is categorized into one of the six classes, then the LLM will give the opinion and the score of the news. (2) Rethink. The LLM will reflect on the reasonableness of the scoring from a long-term trend perspective, and make appropriate revisions to give the final score.
  • Figure 2: Asset value of different trading strategies without considering transaction costs. Our CAR method achieves the highest revenue as it can avoid significant drops as well as keep up with the upward trend when the gold price rises rapidly
  • Figure 3: The distribution curve of scores given by different prompt strategies. The distribution of scores given by our CAR strategy is similar to the normal distribution, which is more aligned with the financial reality compared to the other two strategies.