Abstract

We investigate the capability of Artificial Intelligence (AI), specifically Large Language Models (LLMs) like ChatGPT, to exhibit gender stereotype biases against women entrepreneurs—a critical issue as AI becomes more prevalent in entrepreneurial decision-making. Despite optimism about AI’s utility, the risk of reinforcing existing gender stereotype biases poses significant ethical challenges. Our explorative research, which builds on prior studies and utilizes a regression-based method demonstrated by OpenAI, analyzes LLM responses to various scenarios. Our findings show that LLMs did not ascribe strong masculine traits to successful entrepreneurs, although some nuances still remained. Furthermore, LLMs did not rate femininity lower in general business evaluations, and explicitly identifying the entrepreneur’s sex did not alter these results. These findings seemingly suggest that AI’s perspective deviated from the traditional view that entrepreneurship is predominantly a masculine occupation. However, in investment-specific scenarios, LLMs exhibited a notable bias toward masculine traits over feminine traits, aligning with traditional stereotypical views of entrepreneurship. This variability in responses not only deepens our understanding of gender stereotype biases in AI but also underscores the complexity of addressing these biases, thereby contributing to discussions about AI’s potential to offer less biased evaluations than human judgments and paving the way for fairer AI-driven business practices. Additionally, these findings highlight the necessity for careful integration of AI-generated data into theoretical development, considering AI’s unique information processing capabilities, and the employment of suitable analytical tools to more effectively detect subtle patterns and nuances in AI-generated data.