Itai Ashlagi joined the MS&E faculty in 2015 to continue his intensive study of market design.
He specializes in matching markets, such as kidney exchange programs, organ allocation, school choice, and the National Resident Matching Program.
"I can't think of a better intellectual environment for my research than MS&E, where the faculty and students share a common goal of designing systems and policies that impact our lives."
What made you want to be a professor, and why Stanford?
I was inspired by the range of fascinating and deep topics I encountered at the Technion in Israel while earning my master's and PhD—especially game theory. My first research project was about the benefit that third parties, or correlation devices, can add to multi-player games. This early research and the intellectual and fun environment led me to explore further topics in game theory and mechanism design. After grad school, I continued to do postdoctoral research at Harvard, where I learned that game theory and operations research can be used to design marketplaces that affect people's lives. I became fascinated by the world of matching markets.
Why Stanford? Students here are amazing! Stanford is a hub for market design, including researchers from MS&E, economics, business, and computer science. I can't think of a better intellectual environment for my research than MS&E, where the faculty and students share a common goal of designing systems and policies that impact our lives.
Briefly describe your research.
Kidney Exchange Platforms
Kidney exchange platforms enable incompatible patient-donor pairs to swap donors through cycles or chains that are initiated by altruistic donors. Hospitals enroll pairs and donors to the platform, but participation is not mandatory. Together with Alvin Roth, we predicted that hospitals will begin free-riding platforms by organizing easier exchanges internally and enrolling their harder-to-match pairs to the platform.
In an ongoing project with colleagues at MIT and UPenn, we document that free-riding by hospitals is happening at a large scale. But to improve, we need to create a points system that rewards hospitals with transplants based on their contribution to the platform. So hospitals that enroll a greater number of easier-to-match pairs to the platform increase their chances to match their hard-to-match pairs.
A natural question we study is how to match in this dynamic environment, in which pairs arrive over time. With my colleagues and students we found that chains are crucial to increase the number of transplants and reduce waiting times, partially because of the large fraction of hard-to-match pairs in the pool. We also find that immediate-like matching leads to a good performance and there is no real benefit from waiting for the pool to thicken. These ideas are now incorporated by some national kidney exchange platforms.
The National Resident Matching Program uses an algorithm that, based on doctors' and programs' preferences, finds a stable matching—meaning that no pair of doctor and program mutually prefer each other more than their assigned matches. Theoretically there could be many stable matchings, but empirically this is not the case. With my collaborators, we resolve this longstanding puzzle by showing that two-sided matching markets typically have a unique stable matching. This means that it does not matter which stable algorithm is used, and moreover, participants have no reason to game the system by misreporting their preferences.
We also find that participants on the short side almost always choose who to match with. So this work is a step towards understanding who the realistic matches for participants in such markets are. See this blog post.
Many cities use centralized assignment mechanisms to assign students to schools based on their preferences and priorities assigned to students at different schools. Prior to 2014, the city of Boston was divided into three large zones and families could rank school within their zone. This led to a large transportation budget and students traveling a long way every morning to school. With my student Peng Shi, we used data from Boston to redesign the school choice system to balance fairness, choice, and the budget. The outcome is that every family now has a smaller, but personalized, menu of schools they can rank, including several high quality schools they can rank.
Another challenge in school choice has to do with how to ration seats at schools with excess demand. This issue has been debated in cities like New York and Amsterdam, and recently in Chile. With my student Afshin Nikzad, we compared two practical tie-breaking rules: either all schools use a common lottery, or each school uses a separate lottery. Separate lotteries seem more fair (e.g. a child with a bad lottery at one school still has chances at other schools) but may create artificial inefficiencies. Despite these trade-offs, we find that in popular schools there is no trade-off and a single lottery dominates separate lotteries.
Who has influenced your work?
My PhD Advisors Dov Monderer and Moshe Tenneholtz have inspired me with their creativity, their passion for research, and taught me all about game theory and mechanism design. This continued during my postdoc at Harvard University with Al Roth. Al spurred my interest in matching markets and my desire to engineer real marketplaces. Al, now at Stanford, helped to establish the first centralized kidney exchange platform that used a clearinghouse to enable more transplants. I joined Al in this journey to further improve the incentives and operations of such platforms.
What advice do you have for new MS&E students?
Look for questions that matter (this can take time). Follow your interests and intuition, and strive for creativity. Talk with other students and faculty—they are the best resources and some of them may turn into long-term collaborators. Develop a habit of reading papers. Five years go fast – use them for your research.
You can learn more about Professor Ashlagi’s research on his website.