We analyze lending by traditional as well as fintech lenders during COVID-19. Comparing samples of fintech and bank loan records across the outbreak, we find that fintech companies are more likely to expand credit access to new and financially constrained borrowers after the start of the pandemic. However, this increased credit provision may not be sustainable; the delinquency rate of fintech loans triples after the outbreak, but there is no significant change in the delinquency of bank loans. Borrowers holding both loan types prioritize the payment of bank loans. These results shed light on the benefits provided by shadow banking in a crisis and hint at the potential fragility of such institutions when delinquency rates spike.
In traditional decision-making processes, social biases of human decision makers can lead to unequal economic outcomes for underrepresented social groups, such as women and racial/ethnic minorities (1–4). Recently, the growing popularity of large language model (LLM)-based AI signals a potential shift from human to AI-based decision-making. How would this transition affect the distributional outcomes across social groups? Here, we investigate the gender and racial biases of a number of commonly used LLMs, including OpenAI's GPT-3.5 Turbo and GPT-4o, Google's Gemini 1.5 Flash, Anthropic AI's Claude 3.5 Sonnet, and Meta's Llama 3-70b, in a high-stakes decision-making setting of assessing entry-level job candidates from diverse social groups. Instructing the models to score ∼361,000 resumes with randomized social identities, we find that the LLMs award higher assessment scores for female candidates with similar work experience, education, and skills, but lower scores for black male candidates with comparable qualifications. These biases may result in ∼1–3 percentage-point differences in hiring probabilities for otherwise similar candidates at a certain threshold and are consistent across various job positions and subsamples. Meanwhile, many models are biased against black male candidates. Our results indicate that LLM-based AI systems demonstrate significant biases, varying in terms of the directions and magnitudes across different social groups. Further research is needed to comprehend the root causes of these outcomes and develop strategies to minimize the remaining biases in AI systems. As AI-based decision-making tools are increasingly employed across diverse domains, our findings underscore the necessity of understanding and addressing the potential unequal outcomes to ensure equitable outcomes across social groups.
1 February 2022 in Physica A: Statistical Mechanics and its Applications
Gender discrimination in education hinders women’s representation in various fields. How can we create a gender-neutral learning environment when teachers’ gender composition and mindset are slow to change? Recent development in artificial intelligence (AI) provides a way to achieve this goal as engineers can make AI trainers gender neutral and not take gender-related information as input. We use data from a natural experiment in which such AI trainers replace some human teachers for a male-dominated strategic board game to test the effectiveness of AI training. The introduction of AI improves teaching outcomes for boys and girls and reduces the preexisting gender gap. Survey responses indicate that AI’s information advantage, friendly appearance, and interactive features helped students to learn faster, and class recordings suggest that AI trainers’ nondiscriminatory emotional status can explain the improvement in gender equality. We demonstrate AI’s potential in improving learning outcomes and promoting diversity, equity, and inclusion in analogous settings. This paper was accepted by Elena Katok, Special Issue on the Human-Algorithm Connection. Funding: D. Huang gratefully acknowledges financial support from the National Natural Science Foundation of China [Grants 72503232, 71988101, and T2293771]. C. Lin gratefully acknowledges financial support from the National Natural Science Foundation of China [Grant 72192841] and the Research Grants Council of the Hong Kong Special Administration Region, China [Project No. T35/710/20R]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.02787 .