Imagine a world in which some important decision (a judge's sentencing recommendation, a child's treatment protocol, which person or company should receive a loan) became more reliable because a well-designed algorithm helped a key person make the decision. make decisions to make a better decision. A new MIT economics course is investigating these interesting possibilities.
Lecture 14.163 (Algorithms and Behavioral Sciences) is a new interdisciplinary course focused on behavioral economics, which studies the cognitive abilities and limitations of human beings. The course was co-taught last spring by assistant professor of economics Ashesh Rambachan and visiting professor Sendhil Mullainathan.
Rambachan studies the economic applications of machine learning, focusing on algorithmic tools that drive decision-making in the criminal justice system and consumer lending markets. He also develops methods to determine causality using cross-sectional and dynamic data.
Mullainathan will soon join MIT's Electrical Engineering and Computer Science and Economics departments as a professor. His research uses machine learning to understand complex problems in human behavior, social policy, and medicine. Mullainathan co-founded the Abdul Latif Jameel Poverty Action Lab (J-PAL) in 2003.
The goals of the new course are both scientific (understanding people) and policy-driven (improving society by improving decisions). Rambachan believes that machine learning algorithms provide new tools for the scientific and applied goals of behavioral economics.
“The course investigates the use of computer science, artificial intelligence (ai), economics and machine learning in the service of better outcomes and reduced instances of bias in decision-making,” says Rambachan.
Rambachan believes there are opportunities for evolving digital tools such as artificial intelligence, machine learning and large language models (LLMs) to help reshape everything from discriminatory sentencing practices to health care outcomes among underserved populations.
Students learn to use machine learning tools with three main goals: understand what they do and how they do it, formalize knowledge of behavioral economics so that it integrates well with machine learning tools, and understand areas and topics where the integration of behavioral economics and algorithmic tools could be more fruitful.
Students also produce ideas, develop associated research, and see the bigger picture. They are led to understand where an idea fits and to see where the broader research agenda leads. Participants can think critically about what supervised LLMs can (and cannot) do, to understand how to integrate those capabilities with the models and insights of behavioral economics and recognize the most fruitful areas for the application of what they discover. the investigations.
The dangers of subjectivity and bias
According to Rambachan, behavioral economics recognizes that there are biases and errors in our choices, even in the absence of algorithms. “The data used by our algorithms exists outside of computing and machine learning and is instead often produced by people,” she continues. “Therefore, understanding behavioral economics is essential to understanding the effects of algorithms and how to best build them.”
Rambachan sought to make the course accessible regardless of the academic background of the attendees. The class included advanced degree students from a variety of disciplines.
By offering students an interdisciplinary, data-driven approach to research and discover ways algorithms could improve problem-solving and decision-making, Rambachan hopes to build a foundation upon which to redesign existing systems of jurisprudence, healthcare, and consumer loans. and industry, to name a few areas.
“Understanding how data is generated can help us understand bias,” Rambachan says. “We can ask questions about how to produce a better outcome than what currently exists.”
Useful tools to reimagine social operations
Economics PhD student Jimmy Lin was skeptical of the claims Rambachan and Mullainathan made when the class began, but changed his mind as the course progressed.
“Ashesh and Sendhil began with two provocative statements: the future of behavioral science research will not exist without ai, and the future of ai research will not exist without behavioral science,” Lin says. “Throughout the semester, they deepened my understanding of both fields and walked us through numerous examples of how economics influenced ai research and vice versa.”
Lin, who had previously done research in computational biology, praised the instructors' emphasis on the importance of a “producer mentality,” thinking about the next decade of research rather than the previous decade. “That's especially important in an area as interdisciplinary and fast-moving as the intersection of ai and economics: there is no established old literature, so one is forced to ask new questions, invent new methods, and create new bridges. “. he says.
The speed of change that Lin alludes to is also an attraction for him. “We are seeing black-box ai methods facilitate advances in mathematics, biology, physics, and other scientific disciplines,” Lin says. “ai can change the way we approach intellectual discovery as researchers.”
An interdisciplinary future for economics and social systems
Studying traditional economic tools and enhancing their value with ai can lead to revolutionary changes in the way institutions and organizations teach and empower leaders to make decisions.
“We are learning to follow changes, adjust frameworks, and better understand how to implement tools in service of a common language,” Rambachan says. “We must continually interrogate the intersection of human judgment, algorithms, artificial intelligence, machine learning, and LLMs.”
Lin enthusiastically recommended the course regardless of the students' background. “Anyone interested in algorithms in society, applications of ai in academic disciplines, or ai as a paradigm for scientific discovery should take this class,” she says. “Each conference felt like a gold mine of insights into research, novel application areas, and inspiration on how to produce new and exciting ideas.”
The course, Rambachan says, argues that better-constructed algorithms can improve decision-making across disciplines. “By making connections between economics, computer science, and machine learning, perhaps we can automate the best human decisions to improve outcomes while minimizing or eliminating the worst ones,” he says.
Lin remains excited about the course's still unexplored possibilities. “It's a class that gets you excited about the future of research and your own role in it,” he says.
