Financial Inclusion, Fairness and Stability in the AI Era (FinAI)

人工智能(AI)在金融行业得到了迅速的应用。它承诺为社会带来广泛的好处，包括更广泛地获得信贷，为家庭提供更好的投资建议，以及可能减少财富不平等。实现这些好处是政策制定者的优先事项，例如，通过欧盟在2021年提出的范围广泛的人工智能法案。然而，仍然存在紧迫的挑战。首先，尽管人工智能驱动的金融能够提供低成本和高质量，但它的采用受到缺乏信任和“算法厌恶”的阻碍。其次，人们担心公平和基于人工智能的对弱势群体的歧视。最后，人们对采用人工智能对全系统金融稳定的影响知之甚少。不幸的是，关于金融机构和监管机构如何应对这些挑战，几乎没有概念性的指导。我的项目是一个雄心勃勃的议程，旨在推动我们对人工智能时代金融包容性、公平和稳定的理解。我提出了新颖的理论分析和对金融行业的细粒度数据集的访问，这将为学者和政策制定者提供量化指导。该提案是一个统一的研究板块，在三个层面上开辟了新的天地：(1)个人对人工智能的采用：表征采用人工智能驱动的财务建议的限制，量化个人“算法厌恶”的维度，并为设计自动化建议提供指导。(2)金融中介机构对人工智能的采用：为“公平”部署人工智能贷款制定一个灵活的框架，包括为监管机构提供指导，以及对算法信用评分的最佳限制。(3)人工智能采用的系统效应：首次对人工智能时代的金融稳定性进行严格分析，评估人工智能驱动的投资对系统性风险的后果，并评估可能的监管应对措施的适宜性。

Artificial intelligence (AI) has been rapidly adopted in the financial industry. It promises broad benefits for society, including more inclusive access to credit, better investment advice for households, and possible reductions in wealth inequality. Realizing these benefits is a priority for policy-makers, e.g., through the wide-ranging AI Act proposed by the European Union in 2021. However, there are urgent remaining challenges. First, while AI-driven finance can deliver low cost and high quality, its adoption is encumbered by a lack of trust and "algorithm aversion". Second, there are concerns about fairness and AI-based discrimination against disadvantaged groups. Finally, little is known about the implications of AI adoption for system-wide financial stability. Unfortunately, there is very little conceptual guidance on how financial institutions and regulators can address these challenges. My project is an ambitious agenda to advance our understanding of financial inclusion, fairness and stability in the AI era. I propose novel theoretical analysis and access to granular datasets from the financial industry, which will yield quantitative guidance to academics and policy-makers. The proposal is a unified block of research that breaks new ground on three levels: (1) AI Adoption by Individuals: Characterising the limits to the adoption of AI-driven financial advice, quantifying the dimensions of individuals' "algorithm aversion", and providing guidance for designing automated advice. (2) AI Adoption by Financial Intermediaries: Developing a flexible framework for "fair" deployment of AI-powered lending, including guidance for regulators, and optimal restrictions on algorithmic credit scoring. (3) Systemic Effects of AI Adoption: Providing the first rigorous analysis of financial stability in the age of AI, evaluating the consequences of AI-driven investments for systemic risk, and assessing the suitability of possible regulatory responses.