Alexander “Sasha” Rakhlin stands at the vanguard of theoretical statistics and machine learning, a position reinforced by his recent appointment as the director of the MIT Statistics and Data Science Center. Since returning to his alma mater as a visiting professor in 2016 and officially joining the faculty in 2018, he has served as a bridge between the Department of Brain and Cognitive Sciences and the Institute for Data, Systems, and Society. As the inaugural holder of a distinguished professorship established through the vision of Richard Larson, his leadership marks a pivotal moment for a center tasked with translating complex mathematical frameworks into solutions for the world’s most pressing scientific challenges.
This discussion explores the evolving landscape of data science at the Institute, focusing on the integration of diverse academic perspectives from economics to physics, the rigorous mathematical foundations required for AI safety, and the center’s historical success in mentoring a new generation of interdisciplinary scholars.
Taking over the leadership of the Statistics and Data Science Center involves coordinating diverse perspectives from economics to engineering; how do you approach the task of building a unified community among such distinct fields?
The true strength of the center has always resided in its people—the students, postdocs, and faculty who arrive with sharply different ways of seeing the world. My goal is to foster an environment where a physicist and an economist can find a shared language in statistics to solve questions that neither could tackle alone. We are building upon a foundation laid by previous directors, including the work done during the 2024-25 interim period, to ensure our community remains a home for those who bring unique viewpoints to the most interesting problems of the day. By supporting this collective as it navigates the constantly shifting questions of the field, we ensure that the center remains a place where interdisciplinary friction leads to genuine intellectual breakthroughs. It is about creating a space where the rigorous science of data acts as the common denominator for innovation.
Reflecting on your two-decade journey since your doctoral work, what initially drew you to the intersections of statistics and machine learning, and how has that fascination evolved?
Looking back at my time as a student more than 20 years ago, I remember being immediately struck by the elegant, beautiful connections between probability, optimization, and game theory. Those foundational links haven’t changed, but the scale and impact of the work have shifted dramatically since I was a postdoc at Berkeley or an associate professor at the University of Pennsylvania. Today, I am surrounded by colleagues who are extending this web into the sciences, using these same principles to accelerate everything from nuclear fusion research to complex biological modeling. It is incredibly rewarding to see these theoretical tools, which I once studied in a more abstract sense, now acting as the primary engine for a modern scientific revolution. The recent leap in artificial intelligence is only extending this web further, promising to change how we think about discovery itself.
The Interdisciplinary PhD in Statistics program has seen over 75 students successfully defend their dissertations; what have you learned about the future of the field from overseeing such a diverse group of scholars?
Serving as the initial chair for this program was a deeply moving experience because it allowed me to witness the growth of an entire generation of interdisciplinary thinkers. These 75 scholars represent a vast range of departments, including our own Social and Engineering Systems program, and seeing them apply statistical rigor to such varied domains is a highlight of my career. Mentorship at this level isn’t just about teaching equations; it is about helping students find the courage to bridge the gap between their home departments and the rigorous foundations of data science. Their success proves that statistics is the vital connective tissue that allows us to understand the complexity of the modern world, from the way we manage social systems to how we engineer new technologies. We are training researchers who don’t just use tools, but who understand the mathematics deeply enough to reinvent them when necessary.
As AI becomes a staple in critical infrastructure like medicine and energy, how does your research and leadership address the fundamental mathematical questions of security and reliability?
We have to recognize that as AI enters public life, the questions of safety and security are, at their very core, statistical and mathematical ones. It is not enough to just build a powerful tool; we must be able to quantify uncertainty, provide rigorous guarantees, and understand exactly why a system might fail or how it might resist manipulation. At the center, we are focusing on building a rigorous science of the tools themselves to prevent the kind of failures that could have real-world consequences in medicine or energy sectors. This means moving beyond the “black box” approach and insisting on the same mathematical precision that we apply to any other critical engineering discipline. Our work involves a constant effort to provide the mathematical certainty that society requires before these systems are fully integrated into our daily lives.
What is your forecast for the evolution of statistical thinking as artificial intelligence continues to transform the academic and professional landscape?
I anticipate that we are entering a period where statistics will move from being a specialized tool to becoming the essential framework for all scientific inquiry. As the revolution in machine learning continues to unfold, the need for rigorous uncertainty quantification and robust algorithmic guarantees will only grow, making the work we do at the center more vital than ever. We will see a shift where the “how” of data science—the underlying math and optimization—becomes just as important as the results themselves. Ultimately, the future belongs to those who can master this language to turn raw data into reliable, actionable knowledge that benefits society as a whole. This is a moment built for statistics, and I expect our field to be the bridge that leads us toward a safer and more scientifically grounded future.
