Imagine a world by which essential decisions—a judge's sentencing advice, a baby's treatment protocol, which person or company should get a loan—were more reliable because a well-designed algorithm helped a key decision maker make a better option. A brand new economics course at MIT explores these interesting possibilities.
Course 14.163 (Algorithms and Behavioral Science) is a brand new interdisciplinary course focused on behavioral economics that examines 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 researches the economic applications of machine learning, specializing in algorithmic tools that guide decision-making within the criminal justice system and consumer credit markets. He also develops methods for determining causality using cross-sectional and dynamic data.
Mullainathan will soon join MIT's electrical engineering, computer science, and economics departments as a professor. His research uses machine learning to grasp 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 brand new course are each scientific (understanding people) and politically motivated (improving society through higher decisions). Rambachan believes that machine learning algorithms provide recent tools for each the scientific and applied goals of behavioral economics.
“The course explores using computer science, artificial intelligence (AI), economics and machine learning to attain higher outcomes and fewer bias in decision-making,” says Rambachan.
Rambachan believes that the constant advancement of digital tools corresponding to AI, machine learning and huge language models (LLMs) can transform many things, from discriminatory sentencing practices to healthcare for underserved populations.
Students learn to make use of machine learning tools with three foremost goals: to grasp what and the way machine learning tools work, to formalize insights from behavioral economics in order that they fit well into machine learning tools, and to grasp the areas and topics where the combination of behavioral economics tools and algorithms could possibly be most fruitful.
Students also generate ideas, develop relevant research, and see the larger picture. They are guided to grasp where a finding suits and where the broader research agenda is heading. Participants can think critically about what supervised LLMs can (and can’t) do, to grasp how these skills might be integrated with the models and insights of behavioral economics, and to discover essentially the most fruitful areas for applying the insights from research.
The dangers of subjectivity and bias
According to Rambachan, behavioral economics recognizes that our decisions are shaped by bias and error even without algorithms. “The data utilized by our algorithms exists outside of computer science and machine learning and is as an alternative often created by humans,” he continues. “Understanding behavioral economics is due to this fact essential to understanding the impact of algorithms and higher designing them.”
Rambachan desired to make the course accessible to all participants, regardless of educational background. The class included advanced students from a wide range of disciplines.
By providing students with a cross-disciplinary, data-driven approach to exploring and discovering ways by which algorithms could improve problem-solving and decision-making, Rambachan hopes to put a foundation for redesigning existing systems in law, healthcare, consumer credit, and industry, to call a number of areas.
“By understanding how data is generated, we will higher understand bias,” says Rambachan. “We can ask ourselves how we will recover results than what currently exists.”
Useful tools for redesigning social activities
Economics doctoral student Jimmy Lin was skeptical of Rambachan and Mullainathan's claims originally of the category, but modified his mind because the course progressed.
“Ashesh and Sendhil began with two provocative claims: The way forward for behavioral science won’t exist without AI, and the long run of AI research won’t exist without behavioral science,” says Lin. “Over the course of the semester, they deepened my understanding of each fields and showed us quite a few examples of how economics has influenced AI research and vice versa.”
Lin, who previously researched computational biology, praised the emphasis on the importance of a “producer mentality” that thinks concerning the next decade of research moderately than the past decade. “This is very essential in a field as interdisciplinary and fast-moving because the interface between AI and business – there isn’t any old, established literature, so you might be forced to ask recent questions, invent recent methods and construct recent bridges,” he says.
For him, too, the speed of change that Lin alludes to is an incentive. “We see black-box AI methods enabling breakthroughs in mathematics, biology, physics and other scientific disciplines,” says Lin. “AI can change the best way we as researchers approach mental discovery.”
An interdisciplinary future for economics and social systems
Studying traditional economic tools and increasing their value using artificial intelligence can result in fundamental changes in the best way institutions and organizations educate and empower leaders to make decisions.
“We are learning to trace changes, adapt frameworks, and higher understand the way to use tools within the service of a typical language,” says Rambachan. “We must always query the interface between human judgment, algorithms, AI, machine learning, and LLMs.”
Lin enthusiastically advisable the course, no matter students' backgrounds. “Anyone who is mostly considering algorithms in society, applications of AI in various academic disciplines, or AI as a paradigm for scientific discovery should take this course,” he says. “Every lecture felt like a goldmine of research perspectives, novel application areas, and inspiration for developing recent, exciting ideas.”
The course, Rambachan says, argues that better-developed algorithms can improve decision-making across all disciplines. “By making connections between economics, computer science and machine learning, we may find a way to automate the very best human decisions to enhance outcomes while minimizing or eliminating the worst,” he says.
Lin stays excited concerning the course's unexplored possibilities. “This course makes you excited concerning the way forward for research and your role in it,” he says.
