• Elephant

  

  

WELCOME TO TEAMCORE

The Teamcore group is focused on "AI for social impact". We are focused on advancing AI and multiagent systems research for social impact in topics such as public health, conservation and public safety and security.  We focus on fundamental research problems  in multiagent systems, machine learning, reinforcement learning, game theory, bandit algorithms driven by interdisciplinary collaborations in public health/conservation and public safety, ensuring a virtuous cycle of research and real-world applications. We simultaneously aim to achieve real-world social impact, often in domains with marginalized or endangered communities, and those that have not benefited from AI research in the past.  

The key question we often focus on is how to optimize our limited intervention resources in these domains, and often take the form of decision aids to assist Governmental or Non-Governmental (and Governmental) Organizations (NGOs). Our research is in service of the inspiring work done by these NGOs around the world, to empower them to directly use our AI  tools and services; ultimately we wish to avoid being gatekeepers to this AI technology for social impact.


Professor Tambe

 

Milind Tambe

Gordon McKay Professor of Computer Science &
Director of the Center for Research on Computation and Society (CRCS)
Harvard University

Principal Scientist & Director, "AI for Social Good" Google Research 

 

 

 

MEET THE TEAMCORE MEMBERS

Teamcore 2023 - 24

Meet our current Teamcore members.

Meet Teamcore alumni.
 

Check out our most recent events.

Soon to come, watch Teamcore in action: Seminars and other events featuring Teamcore members!

 

 

 

 

RESEARCH & PIONEERING SYSTEMS

The Teamcore group is focused on "AI for social good". We are focused on advancing AI and multiagent systems research for social impact in topics such as public health, conservation and public safety and security. We focus on fundamental research problems in multiagent systems, machine learning, reinforcement learning, game theory, bandit algorithms that are driven by these topics, ensuring a virtuous cycle of research and real-world applications. We simultaneously aim to achieve real-world social impact, often in domains with marginalized or endangered communities, and those that have not benefited from AI research in the past.

The key question we often focus on is how to optimize our limited intervention resources in these domains, and often take the form of decision aids to assist Non-Governmental (and Governmental) Organizations (NGOs). Our research is in service of the inspiring work done by these NGOs around the world, to empower them to directly use our AI tools and services; ultimately we wish to avoid being gatekeepers to this AI technology for social impact. 

Our group has a long track record of building pioneering and influential systems that have achieved social impact in practice. Read below to learn about our three large domain areas.

  

  

The systems we have
built are the first
large scale applications of
influence maximization
in social networks for
real world impactful
public health outcomes.
With respect to public health, we have large populations to serve, but limited numbers of social workers or public health resources. Concrete example is work we have done with the NGO ARMMAN; ARMMAN  focuses on maternal and child care in India, working with 40  Million beneficiaries. Our work with ARMMAN is focused on the world's two largest mobile health programs for maternal and child care. A key challenge is that beneficiaries/mothers drop out of ARMMAN's healthcare information programs despite preventable reasons. Here, we were the first to use restless bandits in this large scale public health settings to show a more than 30% reduction in drop out rate when compared to control.  Another concrete example is work we have done with youth experiencing homelessness in Los Angeles; harnessing the social networks of these youth, we show that our AI algorithms are far more effective in empowering these youth to reduce HIV risk behaviors compared to traditional approaches. The technical research areas we focus on here are social networks -- enable influence maximization in social networks under dynamism and uncertainty. To that end, we work on approaches involving reinforcement learning, graph sampling, social network influence maximization and related topics. We have extended these approaches towards topics such as TB prevention, maternal and child health care and others. The systems we have built are the first large scale applications of influence maximization in social networks for real world impactful public health outcomes. 

Learn more about our current projects facilitated by AI in Public Health.
  

With respect to conservation, a key example is the PAWS AI system developed by our team  that has been deployed in collaboration with wildlife conservation agencies to assist rangers around the world in protecting endangered wildlife. PAWS led to removals of 1000s of traps used to kill and maim endangered wildlife in national parks in countries such Cambodia and Uganda. Furthermore,  PAWS is integrated with the SMART software, making PAWS available for use at 100s of national parks around the globe. The research areas we focus on here are merging game theory and machine learning, in an approach called Green Security Games.

Learn more about our AI for Conservation research.

  



These systems demonstrate
the first use computational game theory for real-world operational security.
Finally, with respect to public safety, we have large number of targets to protect and limited security resources. Our previous work on "security games" (game theory) research was also used on a daily basis (and continues to be used) by agencies such as the US Federal Air Marshals, US Coast Guard and LA Airport, for innovative patrol strategies and for optimally allocating officers, boats, or other security for improving public safety. These systems demonstrate the first use computational game theory for real-world operational security. There are by now many new applications for security games, including in cybersecurity, protection of endangered wildlife and fisheries, protecting forests, and others. In fact, unanticipated new applications, including audit games, drug design against viruses, traffic enforcement, software code testing, adversarial machine learning, and others, have also sprung up. In addition to the research impact, this article credits our research as having produced a net benefit of over $64 million by saving costs or by increasing security at airports and sea ports.

Learn more about Teamcores work on AI for Public Safety.

 

 

 

 

TEAMCORE IN THE NEWS

   

Milind Tambe PhotoKDD Conference KEYNOTE: AI for Social Impact [2022, 1 hour]

 

     AI for maternal and child care [2022, 15 min]
 

JP Morgan Distinguished Lecture [2020, 57 minutes]

 

  


International Joint Conference on AI John McCarthy Award lecture
[2018, 45 min]

Cambridge Conservation Initiative Seminar [2021, 1 hour]

  

TEAMCORE HISTORY

The Teamcore group started in 1995. The name derives from early multiagent research we conducted in multiagent teamwork, and also inspired by the concept of the "core" in cooperative game theory.  Some milestones in terms of systems and key papers of our research group:

STEAM multiagent system, offering reusable rules for teamwork: Won the Influential paper award at AAMAS conference 2012

Elective Elves multiagent system, a pioneering system of office personal assistants

ADOPT first asynchronous complete algorithm for distributed constraint optimization

ARMOR security games deployed at LA airport first application of computational game theory for operational security

IRIS security games for optimal randomized scheduling air marshals deployed by the Federal Air Marshals

PROTECT security games systems deployed by the US Coast Guard for optimized patrolling

PAWS system for wildlife protection actually operated to remove snares from national parks in Uganda and Cambodia

CHANGE system for HIV Prevention  first large scale application of influence maximization in social network for public health