MIT Course Explores Foundations of Rationality in Artificial Intelligence

A new course at the Massachusetts Institute of Technology, titled AI and Rationality, invites students to explore the complex question of whether artificial systems can truly be rational. Rather than seeking definitive answers, the course uses philosophical inquiry to examine fundamental questions about artificial intelligence (AI), rational decision-making, and the interplay between human cognition and machine agency.

Launched as part of the Common Ground for Computing Education initiative within the MIT Schwarzman College of Computing, the course is co-taught by Leslie Kaelbling, Panasonic Professor of Computer Science and Engineering, and Brian Hedden, a professor with joint appointments in philosophy and computer science. Their collaborative approach bridges technical and philosophical perspectives, equipping students with critical tools to analyze how assumptions about rationality shape work in AI research and applications.

Rationality—typically defined as the ability to form beliefs, learn from experience, and make decisions to achieve specific goals—has long been a subject of interest for both philosophers and computer scientists. The course explores these issues by assessing topics such as the nature of rational agency, the meaning of autonomy in AI systems, and the ways beliefs and desires can be attributed to artificial agents.

According to Kaelbling, philosophical thinking has always influenced computer science, particularly in the early stages of AI. The technical aspects of philosophy merge with AI in formulating rigorous methods for what it means to think or act rationally—an intersection famously exemplified by Alan Turing, who was both a foundational thinker in computer science and a philosopher.

Hedden remarks that while philosophy and computer science may appear divergent, their underlying concerns—how to formalize and evaluate reasoning—often converge. The differences, he notes, tend to be a matter of emphasis and methodology.

The inaugural course, first offered in autumn 2025, attracted students from diverse backgrounds, including computing, engineering, philosophy, and cognitive sciences. Alongside other Common Ground offerings such as Ethics of Computing, which studies the societal impacts of AI, AI and Rationality focuses on theory-building and the ongoing debate about what rationality truly requires—both from humans and machines.

Rather than memorizing or applying standard doctrines, students engage in discussions and readings that challenge them to reconsider foundational assumptions in their own fields. By contrasting mathematical models of logic and learning with real-world human behavior—which often departs from strict rationality—participants grapple with questions about the limitations or applicable scope of rational frameworks.

One student, Amanda Paredes Rioboo, noted that the course prompted her to question whether the inconsistencies between math, logic, and human action mean that humans are irrational, or whether the problem lies in the systems or models themselves. Another student, Junior Okoroafor, appreciated the explicit comparisons between how different fields conceptualize rationality and the importance of clarifying these assumptions.

As AI becomes increasingly influential in society, the ability to critically analyze the philosophical underpinnings of machine intelligence is seen as a vital skill for future researchers, practitioners, and policymakers. Kaelbling emphasizes that students need habits of mind and intellectual flexibility to tackle challenges in a rapidly changing technological landscape—skills that may matter more than learning any particular tool or technique.

The co-taught, interdisciplinary structure of the course fosters an environment in which both students and instructors can learn from each other, demonstrating the value of integrating theoretical and applied perspectives. The popularity and continued expansion of such interdisciplinary offerings suggest a growing recognition of the need to blend technical expertise with philosophical reflection as artificial intelligence advances.

For more information, visit: news.mit.edu