Imagine a world where some important decisions—a judge's sentencing recommendation, a child's treatment protocol, or which individual or business should get a loan—become more trustworthy because well-designed algorithms help key decision-makers make better choices. A new MIT economics course is investigating these intriguing possibilities.
Class 14.163 (Algorithms and Behavioral Science) is a new interdisciplinary course focused on behavioral economics, the study of human cognitive abilities and limitations. The course was co-taught last spring by Ashesh Rambachan, assistant professor of economics, and Sendhil Mullainathan, visiting lecturer.
Rambachan studies economic applications of machine learning, with a focus on algorithmic tools for decision-making in the criminal justice system and consumer lending markets. He also develops methods to determine causal relationships using cross-sectional and dynamic data.
Mullainathan will soon join the faculty of MIT's Department of Electrical Engineering, Computer Science, and Economics. His research uses machine learning to understand complex problems in human behavior, social policy, and medicine. Mullainathan co-founded the Abdul Latif Jameel Poverty Alleviation Lab (J-PAL) in 2003.
The goals of the new process are both scientific (to understand people) and policy-driven (to improve society by improving decision-making). Rambachan believes that machine learning algorithms provide new tools for both the scientific and applied goals of behavioral economics.
“This course examines how computer science, artificial intelligence (AI), economics, and machine learning can be deployed to improve outcomes and reduce instances of bias in decision-making,” says Rambachan.
Rambachan believes that constantly evolving digital tools such as AI, machine learning, and large-scale language models (LLMs) have the opportunity to help reshape everything from discriminatory practices in criminal sentencing to health outcomes in underserved populations.
Students will learn how to use machine learning tools with three main goals: This means understanding what you do and how you do it, formalizing your behavioral economics insights and organizing them well within machine learning tools, and understanding the areas and topics into which behavioral economics is integrated. Algorithmic tools may be most beneficial.
Students also generate ideas, develop relevant research, and see the bigger picture. They will understand where the insights fit and see where the broader research agenda is going. Participants will be able to think critically about what a guided LLM can and cannot do, understand how to integrate these competencies with models and insights from behavioral economics, and recognize areas where applying the findings will be most beneficial. there is.
Subjectivity and the dangers of bias
According to Rambachan, behavioral economics acknowledges that biases and mistakes exist throughout our choices, even without algorithms. “The data used in our algorithms exists outside of computer science and machine learning and is instead often generated by people,” he continues. “Understanding behavioral economics is therefore essential to understanding the effectiveness of algorithms and how to build them better.”
Rambachan strived to make the course accessible regardless of the academic background of attendees. The class included advanced degree students from a variety of fields.
Rambachan provides students with an interdisciplinary, data-driven approach to investigate and discover how algorithms can improve problem solving and decision-making, laying the foundation for redesigning existing systems in law, healthcare, and consumer lending. We want to build it. , industry, etc., to name a few fields.
“Understanding how data is generated can help us understand bias,” says Rambachan. “You can ask questions about creating better outcomes than what currently exists.”
Helpful tools for reimagining your social operations
Jimmy Lin, a PhD student in economics, was skeptical about the claims made by Rambachan and Mullainathan when the class began, but changed his mind as the course continued.
“Ashesh and Sendhil started with two provocative claims: the future of behavioral science research will not exist without AI, and the future of AI research will not exist without behavioral science,” says Lin. “Throughout the semester, they deepened my understanding of both fields and guided us through numerous examples of how economics has influenced AI research.”
Lin, who previously worked in computational biology, praised the instructor's emphasis on the importance of a “producer mindset,” thinking about the next decade of research rather than the past. “This is especially important in an interdisciplinary, fast-moving area like the intersection of AI and economics. Because there is no existing, established literature, we need to ask new questions, invent new methods, and build new bridges.” he says
The speed of change that Lin mentioned also appealed to him. “We are seeing black-box AI methods fostering breakthroughs in mathematics, biology, physics and other scientific fields,” says Lin. “AI can change the way we as researchers approach intellectual discovery.”
The interdisciplinary future of economic and social systems
Studying traditional economic tools and enhancing their value through AI could lead to groundbreaking changes in the way institutions and organizations teach and empower leaders to make choices.
“We are learning how to better understand how to track shifts, align frameworks, and deploy tools that serve a common language,” Rambachan says. “We must continue to examine the intersection of human judgment, algorithms, AI, machine learning, and LLM.”
Lin enthusiastically recommended this course to students regardless of their background. “Anyone broadly interested in algorithms in society, the application of AI across disciplines, or AI as a paradigm for scientific discovery should take this class,” he says. “Every lecture felt like a goldmine of inspiration for perspectives on research, novel applications, and ways to generate new and exciting ideas.”
Rambachan argues that better-built algorithms in this process can improve decision-making across multiple fields. “By building connections between economics, computer science, and machine learning, we will be able to automate the best human choices to improve outcomes while minimizing or eliminating the worst choices,” he says.
Lin remains intrigued by the unexplored possibilities of the course. “It’s a class that gets you excited about the future of research and your role in it,” he says.