Interdisciplinary Team of CU Researchers to Lead New National AI Institute

Congratulations to 20 CU Boulder researchers!

As seen in the CU Boulder Today feature article titled 'New $20 million center to bring AI into the classroom', the National Science Foundation (NSF) held a press conference on August 25th announcing the funding for seven AI Institutes, including the Institute for Student-AI Teaming which will be hosted by ICS. It will explore the role that artificial intelligence may play in the future of education and workforce development—especially in providing new learning opportunities for students from historically underrepresented populations in Colorado and beyond. The aim of the Institute according to Sidney D'Mello, associate professor at ICS and the Department of Computer Science and Principal Investigator is "to advance a new science of teaming. We have a lot of knowledge of what makes effective human-human teams. The next phase is understanding what underlies effective human-agent teams. In our case, that means students, AI and teachers working together." Collaboration in the classroom improves learning effectiveness, yet teachers are challenged to facilitate all of the student interactions in the classroom. The research team will focus on how AI "partners", the pairing of students with an AI tool, may improve collaboration and effective interactions in the classroom.

Researchers will address three main challenges, referred to as 'Strands': first, work to develop new advancements in fundamental science of how machines process human language, gestures, and emotions (Strand 1). Next, the team will strive to better understand how students, AI and teachers can collaborate effectively in both classrooms and remote learning contexts (Strand 2). Last, researchers will go to classrooms in Denver Public Schools and other school partners—virtually, during the age of COVID-19—to work hand-in-hand with students and teachers to think up new technologies (Strand 3). 

The 5-year project will bring together a team of researchers from nine universities from across the country in a close collaboration with two public school districts, private companies and community leaders. Researchers from across the CU Boulder campus, including the Institute of Cognitive Science, the College of Engineering and Applied Science (CEAS), College of Arts and Sciences (CAS), and the School of Education will collaborate and contribute to the Institute.

The project will pursue three interconnected research strands.

Strand 1: Understanding and facilitating conversations

Led by Institute Co-PI Martha Palmer with Professor Ross Beveridge at Colorado State University as a Co-Lead. It is focused on building a rich dialogue capability for discussions with the students and teachers, both through language and also non-verbal communication. The CU-Boulder contributors are experts in Natural Language Processing and Human-Robot Communication, and  include Jim Martin, Katharina Kann, and Alessandro Roncone. We are teaming with Professor Lyn Walker, a world leader in discourse analysis, and Professor Jeff Flanigan, another Natural Language Processing expert, at the University of California at Santa Cruz for the language component. The non-verbal elements, including facial expression and gesture recognition, will be developed by our outstanding colleagues, Professor Jake Whitehill, Worcester Polytechnic Institute (WPI), Ross Beveridge, our Co-Lead, and Professor James Pustejovsky, Brandeis. In tracking the student conversations the system will also be able to record degrees of engagement and points of confusion or insight on the part of the students, all of which can be provided to the classroom teacher.

Martha states: “I am absolutely thrilled to have the opportunity to work with this world-class team of experts on such a very challenging issue - enabling real-time, productive communication between students and an AI Partner.  It pushes the boundaries of everything Computational Linguistics has been able to accomplish, and is only possible because of the carefully crafted team infrastructure that will provide input from the Strand 2 experts and allow beta-testing and field-testing in classrooms. There is nothing I’d rather be doing.”

Strand 2: Orchestrating interactions with AI

Co-led by ICS Research Professor Leanne Hirshfield and Professor Sadhana Puntambekar from  the University of Madison-Wisconsin. Among the group of talented researchers from multiple Universities working on this strand, Hirshfield will be assisted by fellow CU Professors Peter Foltz, McKell Carter, and Clayton Lewis, with Sidney D’Mello rounding out the team.  This strand aims to i) design student-AI teaming paradigms needed for an AI partner to scaffold 21st century learning practices, ii) investigate how AI can support teachers in orchestrating classrooms for collaboration at the individual, small group, and whole classroom level, and iii) identify socio-cognitive-affective states pertinent to student-AI team interactions (e.g., trust, joint attention) and ground them in neurophysiological signals. Invaluable team members from other Universities supporting this research strand include Professors Jamie Gorman from Georgia Tech and Mike Tissenbaum from University of Illinois Urbana-Champaign. 

Leanne shared that "One of the most exciting and impactful parts of our strand's research is how we bridge the gap between the technical AI work being done by Martha's strand 1 team with the extensive knowledge gleaned from actual teachers and students, as explored by Bill and Tammy's strand 3 team. In order for the AI to select appropriate ways to interact with the students, it’s important to not only understand what students say, and what they do, but what they’re thinking and feeling. And no one understands these delicate intricacies better than the actual teachers and students. I believe that our greatest impact to AI and education will be how we carefully and purposefully take the knowledge gleaned from teachers and students and directly build that knowledge into the architecture of our AI systems."

Strand 3: Broadening participation with co-design

Co-led by co-PI and ICS Director Tamara Sumner and Professor Bill Penuel of the Institute of Cognitive Science and the School of Education. They will be working with a team of school and community partners, including Denver Public Schools and Project VOYCE, and fellow CU Professors Kalonji Nzinga and Arturo Cortez, and Thomas Philip from the University of California Berkeley. This team will convene a set of community-level meetings that include AI researchers and community organizations, advocacy groups, and others engaged in the study of AI in society to help build a vision for how AI technologies might support more just and equitable futures. The group will also co-design and test with teachers and students a set of units in which students develop understanding of, critique, and create AI technologies that contribute to such a vision.

“Community members must be included from the very beginning when it comes to designing and developing technology that will be deployed in schools—this includes involving students, teachers, parents and other community leaders” according to Tamara. Bill adds that “students need to understand how AI functions in the world now, including its potential role in building a ‘surveillance economy,’ and how it can help communities design together for a more just future. To do that, we are going to engage a diverse group of stakeholders from the community and from schools to help us set goals for co-designing curricula for middle and high school students."

Leveraging ICS' strength in multidisciplinary and interdisciplinary collaboration, the research team hopes that this work will inspire a new generation of young people to get interested in AI and come up with new ways to use technology to help their own communities.

CU Boulder researchers involved in the AI-institute

Additional collaborators not seen here will contribute to the Institute on an ongoing basis.