Title here
Summary here
Research
Teamwork is central to the human enterprise: our greatest achievements are the work of teams, not individuals. Increasingly, AI and robotic systems support humans in high-stakes domains that depend on teamwork. Yet, despite impressive advances, today’s intelligent systems still cannot team up effectively with people. They remain tools, not teammates.
This gap limits the value intelligent systems can deliver and, worse, can lead to adverse consequences: there is no guarantee that AI will enhance human performance rather than detract from it. To realize a future where machines complement human capability safely and effectively, we pursue research on human-AI/robot interaction, with an emphasis on teamwork in the physical world.
Computational Foundations
Effective teamwork begins with mutual understanding. We develop the computational building blocks that establish this common ground between humans and machines: algorithms that enable machines to infer human intent from behavior, and methods that explain AI decisions to human teammates. Together, these enable the two-way communication that human-AI teaming requires.
Enabling Machines to Understand Humans
Reward and policy learning algorithms to infer human intent and preferences from behavior. These algorithms let AI agents understand, anticipate, and adapt to their human teammates from limited interaction.
Explaining AI Decisions to Human Users
Explainable AI algorithms and user interfaces that summarize an agent's rewards and policies. These summaries help humans form accurate mental models of the AI and robots they work with, even before collaboration begins.
Interactive AI & Robotics
We design interactive agents as stepping stones toward human-AI teaming in the physical world. Our cognitive assistants serve as coaches and coordinators, guiding and orchestrating human teams without performing the task themselves. Our robotic assistants take the complementary role, working alongside people to carry out the task. Together, these two lines of work develop the capabilities that effective teaming demands: understanding teamwork and executing within it.
Cognitive Assistants
Effective teams require seamless coordination, from aligned decisions to smooth hand-offs. We create AI coaches and coordinators that strengthen this capability, supporting teams from the sidelines without performing the task themselves.
Robotic Assistants
Effective teammates must also carry out their share of the task alongside others. We research robots that assist people with physical tasks, adapting to their partners to work safely and fluently by their side.
Use-Inspired Research
To ground our research in real applications, we partner with collaborators who bring deep expertise in human factors, team science, and medicine. Together, we are developing an AI coach that helps strengthen teamwork and robotic tutors that assess nursing skills. We are always looking to bring our foundational research to new domains. If you see a role for human-AI teaming in your field, we would love to hear from you.
An AI Coach for Surgical Team Training
Healthcare in the operating room depends on how well the surgical team coordinates, where preventable errors put patients at risk. In collaboration with the Boston VA Research Institute, Brigham & Women's Hospital, MIT, and MET Lab, we are building the foundations of an AI coach that observes surgical teamwork and helps strengthen it, aiming to improve patient safety.
A Robotic Teaching Assistant for Nurse Training
Nurses master clinical skills through extensive hands-on practice and apprenticeship, but expert instructors' time is increasingly scarce. In collaboration with Houston Methodist Hospital and the Kavraki Lab, we are developing a robotic teaching assistant that supports instructors by assessing trainees' skills from multimodal data and generating interpretable summaries of each training session.