This article addresses risks related to the use of artificial intelligence and machine learning models that were trained on and use data that overweight benign credit conditions.
Jacob Kosoff, American Banker
This paper provides a nontechnical overview of common machine learning algorithms used in underwriting credit. It provides a review – including strengths and weakness – of common machine learning techniques in credit underwriting, including tree-based models, support vector machines, and neural networks. The paper also considers the financial inclusion implications of machine learning, nontraditional data, and fintech.
Majid Bazarbash, International Monetary Fund (IMF)
This paper provides an overview of explainable machine learning as well as definitions of explainable artificial intelligence, examples of its usage, and details for responsible and human-centered use.
Patrick Hall, H20.AI
This white paper outlines some of the most important considerations for managing risk in machine learning models to create more accurate and compliant algorithms. Key recommendations include focusing on the quality of input data as well as implementing techniques to reduce and expose bias.
Andrew Burt, Immuta & Future of Privacy Forum
This paper provides an overview of the challenges and implications for the supervision and regulation of financial services with regards to the opportunities presented by BDAI technology: the phenomena of big data (BD) being used in conjunction with artificial intelligence (AI). The paper draws from market analyses and use cases to outline potential developments seen from the industry and government perspectives, and the impact on consumers.
Germany’s Federal Financial Supervisory Authority (BaFin)
This paper uses various types of machine learning models to predict credit risk using historical mortgage data. It finds gains in predictiveness that would likely lead to an increase in approvals across all demographic groups, but that average prices could increase for African-American and Hispanic borrowers due to differences in risk calculations.
Andreas Fuster, Paul Goldsmith-Pinkham, Tarun Ramadorai, and Ansgar Walther.
This paper examines AI systems and how these systems should be held accountable, in particular by one method: explanation. The paper focuses on using explanation from AI systems at the right time to improve accountability, and reviews societal, moral, and legal norms around explanation. The paper ends with advocating that at present, AI systems can and should be held responsible to a similar standard of explanation as humans are, and adapt as the future changes.
Berkman Center Research Publication Forthcoming; Harvard Public Law Working Paper No. 18-07
Finale Doshi-Velez, Mason Kortz, Ryan Budish, Christopher Bavitz, Samuel J. Gershman, David O’Brien, Stuart Shieber, Jim Waldo, David Weinberger, Alexandra Wood.
