MIT Laboratory for Information and Decision Systems

08/27/2026

When it comes to AI-assisted medical diagnosis, one size may not fit all.

A new study from LIDS PI Marzyeh Ghassemi and collaborators found that AI assistance generally helped both clinicians and non-experts more accurately diagnose skin diseases—but AI explanations affected the two groups differently.

Non-experts were more likely to trust AI recommendations, even when they were wrong. They became especially reliant on the AI when its recommendations included LLM-generated explanations—even when those explanations were incorrect, vague, or overly generic.

Clinicians were less likely to be misled by incorrect AI advice and actually performed best when they received only the AI model’s prediction, without an added explanation.

The research suggests that effective medical AI systems may need to account for who is using them—and provide different kinds of support based on users’ expertise.

Read more at MIT News: https://bit.ly/4gv0jR7

The research team includes: Xuhai ‘Orson’ Xu, Haoyu Hu, Haoran Zhang, Will Ke Wang, Reina Wang, Luis R. Soenksen, Omar Badri, Sheharbano Jafry, Elise Burger, Lotanna Nwandu, Apoorva Mehta, Erik P. Duhaime, Asif Qasim, Hause Lin, Janis Karleen Pereira, Jonathan Hershon, Paulius Mui, Alejandro A. Gru, Noémie Elhadad, Lena Mamykina, Matthew Groh, Philipp Tschandl, Roxana Daneshjou & Marzyeh Ghassemi

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

08/21/2026

Congratulations to LIDS PI Alexander “Sasha” Rakhlin, who has been named director of the MIT Statistics and Data Science Center (SDSC)!

An expert in machine learning, statistics, and computation, Rakhlin succeeds Professor Ankur Moitra and plans to deepen connections among researchers across MIT — from economics and political science to physics and engineering.

“Statistics is a shared language across MIT,” Rakhlin says. “Collaborations in areas from biology to nuclear fusion have shown how statistical thinking accelerates science itself.”

As director, he hopes to strengthen SDSC as a home for the foundations of data science and AI and a bridge to the scientific and societal challenges where those foundations can have an impact.

Read more at MIT News: https://bit.ly/4gcm08m

MIT Schwarzman College of Computing MIT Institute for Data, Systems, and Society

08/18/2026

Nature has spent millions of years developing materials that bend, twist, expand, and adapt. What if engineers could borrow those design principles to create new materials?

MIT researchers have developed a mathematical framework that breaks down natural structures — like pinecones, tree bark, and sweetgum seedpods — into building blocks that can be mixed and matched to design manufacturable, adaptive materials.

Developed by LIDS PI Gioele Zardini, graduate student Lee Marom, MIT Architecture’s Skylar Tibbits, and CEE’s Markus J. Buehler, the framework could help engineers design robotic grippers, aerospace components, and other technologies with complex behaviors inspired by nature.

Learn more at MIT News: https://bit.ly/4zr0VOK

08/05/2026

Using reason, again and again
A new MIT News profile highlights the interdisciplinary journey of LIDS PI Brian Hedden, whose work bridges philosophy, computer science, and ethics.
After returning to his alma mater in 2025, Hedden joined MIT as a shared faculty member between the MIT Schwarzman College of Computing, the department of Linguistics and Philosophy, and the MIT EECS Department. He also serves as associate dean for MIT's Social and Ethical Responsibilities of Computing (SERC) initiative.
Read more: https://bit.ly/4wadCuy

MIT School of Engineering MIT School of Humanities, Arts, and Social Sciences

07/24/2026

Remembering MIT Professor Emeritus Dimitri Bertsekas

LIDS PI Dimitri Bertsekas leaves an extraordinary legacy as a researcher, educator, and author whose work helped shape modern optimization, control theory, reinforcement learning, and artificial intelligence.

Over the course of his career, he authored more than 20 influential books and textbooks, mentored generations of students, and made lasting contributions to both academia and industry. His clear, elegant writing and passion for teaching have had a profound impact on researchers around the world.

Read a tribute to his life, work, and enduring legacy: https://bit.ly/3Rg1fip

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

07/22/2026

Following the questions where they lead

What happens when you combine computer science, economics, public health, and political science?

For MIT LIDS PI Bailey Flanigan, it leads to new ways of strengthening democracy.

A new MIT News profile explores Flanigan's interdisciplinary path—from medicine and bioengineering to her current research at MIT, where she develops computational and mathematical methods that make democratic participation more meaningful and effective.

Read more about her research journey and vision for the future of democracy: https://bit.ly/452IlyR

MIT Schwarzman College of Computing MIT EECS Department MIT School of Engineering MIT School of Humanities, Arts, and Social Sciences

A profile of MIT Assistant Professor Bailey Flanigan explores how she develops complex computational methods for helping democracy thrive.

07/17/2026

Helping AI models meet the real world
LIDS PI Devavrat Shah is working to bridge the gap between AI research and real-world decision-making.
His research focuses on developing AI methods that can make continuous decisions using limited computational resources. That work led to the creation of Ikigai Labs, where his team developed a foundation model for tabular and time-series data.
Now part of Celonis, the technology will help businesses combine with their own data and workflows to improve forecasting, planning, and everyday decision-making.
Read more about Shah's research and entrepreneurial journey: https://bit.ly/3Rnx60p
MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

07/15/2026

How can we make AI safer for children?
A collaboration between researchers at MIT LIDS, Boston University, and child safety nonprofit Thorn has introduced a new way to evaluate generative AI models for harmful capabilities—without generating illegal content during testing.
The technique could help auditors identify open-source AI models that have been modified to produce illegal material, supporting safer AI development and deployment.
The research team includes Vinith M. Suriyakumar, Ayush Sekhari, Lena Stempfle, Robertson Wang, Michael Simpson, Rebecca Portnoff, Marzyeh Ghassemi, and Ashia C. Wilson.
Read more from MIT News: https://bit.ly/4vAqCto

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

07/08/2026

In game theory, generalists sometimes win out over specialists.

A new study from MIT LIDS and collaborators suggests that AI algorithms designed for a broad range of learning tasks can sometimes outperform methods created specifically for strategic games.

The research, led by graduate student Sobhan Mohammadpour, LIDS PI Gabriele Farina, and collaborators, found that policy gradient methods can outperform specialized game-theoretic approaches in certain imperfect-information games.

The work offers a new perspective on how researchers evaluate AI systems designed for strategic decision-making and was presented at the 2026 International Conference on Learning Representations (ICLR) in Rio de Janeiro.

Read more at MIT News: https://bit.ly/4faViv1

07/01/2026

Congratulations to our 2026 L4DC Best Paper Award winners! 🏆

LIDS graduate students Andrea Goertzen and Sunbochen Tang, along with LIDS PI Navid Azizan, have received the Best Paper Award at the 2026 Learning for Dynamics and Control (L4DC) Conference.

Their paper, "ECO: Energy-Constrained Operator Learning for Chaotic Dynamics with Boundedness Guarantees," introduces a new framework for modeling chaotic systems that draw on control theory and mathematical optimization to ensure that neural network models satisfy energy-dissipation constraints, enabling more reliable and physically meaningful long-horizon forecasts.

Andrea and Sunbochen are equal-contributing student co-authors of the paper.

The award was presented at the 8th Annual L4DC Conference at the University of Southern California on June 19, 2026.

Learn more and read the paper: https://bit.ly/3RfP0Ch

Image: (L-R) Andrea Goertzen, Navid Azizan, and Sunbochen Tang

MIT Schwarzman College of Computing MIT Mechanical Engineering MIT School of Engineering MIT Aeronautics and Astronautics