About

Asu Ozdaglar is a faculty member at MIT EECS with a focus on electrical engineering and computer science. Her research areas include systems theory, control, and autonomy, as well as optimization and game theory. She is involved in developing groundbreaking systems that sense, process, and transmit energy and information, leveraging computational, theoretical, and experimental tools to address shared challenges facing humanity. Her work encompasses a broad spectrum of topics within electrical engineering and computer science, emphasizing the development of innovative algorithms, systems, and theoretical frameworks.

Research topics

• Computer science
• Computer network
• Mathematics
• Economics
• Microeconomics

Selected publications

• MisinfoEval: Generative AI in the Era of "Alternative Facts"

  arXiv (Cornell University) · 2024-10-13

  preprintOpen accessSenior author

  The spread of misinformation on social media platforms threatens democratic processes, contributes to massive economic losses, and endangers public health. Many efforts to address misinformation focus on a knowledge deficit model and propose interventions for improving users' critical thinking through access to facts. Such efforts are often hampered by challenges with scalability, and by platform users' personal biases. The emergence of generative AI presents promising opportunities for countering misinformation at scale across ideological barriers. In this paper, we introduce a framework (MisinfoEval) for generating and comprehensively evaluating large language model (LLM) based misinformation interventions. We present (1) an experiment with a simulated social media environment to measure effectiveness of misinformation interventions, and (2) a second experiment with personalized explanations tailored to the demographics and beliefs of users with the goal of countering misinformation by appealing to their pre-existing values. Our findings confirm that LLM-based interventions are highly effective at correcting user behavior (improving overall user accuracy at reliability labeling by up to 41.72%). Furthermore, we find that users favor more personalized interventions when making decisions about news reliability and users shown personalized interventions have significantly higher accuracy at identifying misinformation.

  PublisherOA PDFDOI

• Generative AI in the Era of 'Alternative Facts'

  2024-03-27 · 10 citations

  articleOpen accessSenior author

  The spread of misinformation on social media platforms threatens democratic processes, contributes to massive economic losses, and endangers public health. Many efforts to address misinformation focus on a knowledge deficit model and propose interventions for improving users’ critical thinking through access to facts. Such efforts are often hampered by challenges with scalability, and by platform users’ confirmation bias. The emergence of generative AI presents promising opportunities for countering misinformation at scale across ideological barriers. In this paper, we present (1) an experiment with a simulated social media environment to measure effectiveness of misinformation interventions generated by large language models (LLMs), (2) a second experiment with personalized explanations tailored to the demographics and beliefs of users with the goal of alleviating confirmation bias, and (3) an analysis of potential harms posed by personalized generative AI when exploited for automated creation of disinformation. Our findings confirm that LLM-based interventions are highly effective at correcting user behavior (improving overall user accuracy at reliability labeling up to 47.6%). Furthermore, we find that users favor more personalized interventions when making decisions about news reliability.

  PublisherOA PDFDOI

• Preface

  2024-03-27

  articleOpen accessSenior author

  We are pleased to share this volume of papers on generative AI and its broader impacts, written by MIT faculty and researchers and their collaborators. These papers resulted from a call issued by President Sally Kornbluth and Provost Cynthia Barnhart to provide roadmaps . . .

  PublisherOA PDFDOI

• In Memory of Paul Penfield Jr. (1933–2021) [people]

  IEEE Solid-State Circuits Magazine · 2021-01-01

  articleOpen access
  PublisherOA PDFDOI

• When is Society Susceptible to Manipulation?

  Deep Blue (University of Michigan) · 2019-01-01

  articleOpen access

  We consider a social learning model where agents learn about an underlying state of the world from individual observations as well as from exchanging information with each other. A principal (e.g. a firm or a government) interferes with the learning process in order to manipulate the beliefs of the agents. By utilizing the same forces that give rise to the ``wisdom of the crowd'' phenomenon, the principal can get the agents to take an action that is not necessarily optimal for them but is in the principal's best interest. We characterize which networks are susceptible to this kind of manipulation and derive conditions under which a social network is impervious and cannot be manipulated. In the process, we generalize some known centrality measures and describe how our model offers insights into designing networks that are resistant to manipulation.

  PublisherOA PDF

• Global Convergence Rate of Proximal Incremental Aggregated Gradient Methods

  SIAM Journal on Optimization · 2018-01-01 · 8 citations

  preprintOpen accessSenior author

  We focus on the problem of minimizing the sum of smooth component functions (where the sum is strongly convex) and a nonsmooth convex function, which arises in regularized empirical risk minimization in machine learning and distributed constrained optimization in wireless sensor networks and smart grids. We consider solving this problem using the proximal incremental aggregated gradient (PIAG) method, which at each iteration moves along an aggregated gradient (formed by incrementally updating gradients of component functions according to a deterministic order) and takes a proximal step with respect to the nonsmooth function. While the convergence properties of this method with randomized orders (in updating gradients of component functions) have been investigated, this paper, to the best of our knowledge, is the first study that establishes the convergence rate properties of the PIAG method for any deterministic order. In particular, we show that the PIAG algorithm is globally convergent with a linear rate provided that the step size is sufficiently small. We explicitly identify the rate of convergence and the corresponding step size to achieve this convergence rate. Our results improve upon the best known condition number and gradient delay bound dependence of the convergence rate of the incremental aggregated gradient methods used for minimizing a sum of smooth functions.

  PublisherOA PDFDOI

• Global Convergence Rate of Proximal Incremental Aggregated Gradient\n Methods

  arXiv (Cornell University) · 2016-08-04 · 1 citations

  preprintOpen accessSenior author

  We focus on the problem of minimizing the sum of smooth component functions\n(where the sum is strongly convex) and a non-smooth convex function, which\narises in regularized empirical risk minimization in machine learning and\ndistributed constrained optimization in wireless sensor networks and smart\ngrids. We consider solving this problem using the proximal incremental\naggregated gradient (PIAG) method, which at each iteration moves along an\naggregated gradient (formed by incrementally updating gradients of component\nfunctions according to a deterministic order) and taking a proximal step with\nrespect to the non-smooth function. While the convergence properties of this\nmethod with randomized orders (in updating gradients of component functions)\nhave been investigated, this paper, to the best of our knowledge, is the first\nstudy that establishes the convergence rate properties of the PIAG method for\nany deterministic order. In particular, we show that the PIAG algorithm is\nglobally convergent with a linear rate provided that the step size is\nsufficiently small. We explicitly identify the rate of convergence and the\ncorresponding step size to achieve this convergence rate. Our results improve\nupon the best known condition number dependence of the convergence rate of the\nincremental aggregated gradient methods used for minimizing a sum of smooth\nfunctions.\n

  PublisherOA PDFDOI

• Network security and contagion

  Journal of Economic Theory · 2016-10-10 · 159 citations

  articleOpen accessSenior author
  PublisherOA PDFDOI

• A Stronger Convergence Result on the Proximal Incremental Aggregated Gradient Method

  arXiv (Cornell University) · 2016-11-23 · 8 citations

  preprintOpen accessSenior author

  We study the convergence rate of the proximal incremental aggregated gradient (PIAG) method for minimizing the sum of a large number of smooth component functions (where the sum is strongly convex) and a non-smooth convex function. At each iteration, the PIAG method moves along an aggregated gradient formed by incrementally updating gradients of component functions at least once in the last $K$ iterations and takes a proximal step with respect to the non-smooth function. We show that the PIAG algorithm attains an iteration complexity that grows linear in the condition number of the problem and the delay parameter $K$. This improves upon the previously best known global linear convergence rate of the PIAG algorithm in the literature which has a quadratic dependence on $K$.

  PublisherOA PDFDOI

• Privacy-Constrained Network Formation

  ACM SIGMETRICS Performance Evaluation Review · 2015-11-19

  articleSenior author

  No abstract available.

  PublisherDOI

Frequent coauthors

• Daron Acemoğlu

  Massachusetts Institute of Technology

  10 shared

• Azarakhsh Malekian

  5 shared

• James Siderius

  4 shared

• Ali Makhdoumi

  3 shared

• N. Denizcan Vanli

  Massachusetts Institute of Technology

  3 shared

• Mert Gürbüzbalaban

  Rutgers, The State University of New Jersey

  3 shared

• Muriel Médard

  Massachusetts Institute of Technology

  3 shared

• Ali ParandehGheibi

  Cisco College

  2 shared

Awards & honors

• Edgerton Award winners (2026)