H. Chad Lane
PI & Director, INVITE Institute
We are excited that the first official day of the NSF-IES INVITE AI Institute (out of 1,826) has finally arrived! As all AI institute teams know well, it was a long journey to reach this point, with many ups and downs, hundreds of hours of meetings, 1000s of hours of writing, and almost 6000 Slack posts. Our team ultimately converged on a vision of building AI-based tools that seek to model and support learners in more holistic ways. We paired this with the belief that AI systems of the future will be most effective for K-12 education if they work in concert with teachers and amplify effective, empirically supported classroom practices (this goal, incidentally, is emphasized in the Department of Education’s recent report on AI).
Recognizing this goal was vague, we consulted with teachers and we chose to focus our efforts on a trio of noncognitive skills: persistence, academic resilience, and collaboration (specifically, how children communicate in collaborative settings). We were inspired by research by Dr, Camille Farrington at the University of Chicago who found that these noncognitive skills (a wide-ranging, rich set of skills that underlie learning, described here and here) vary widely based on SES and background, and so so our educational system needs to be better prepared to support their success (for example, with educational technologies that are built to detect and support this variability). Kids can also exhibit noncognitive skills in unique ways and bring their own strengths to a learning situation. Our observation was that most AI-based learning technologies do not do a particularly good job tracking these skills, although decades of work has pointed in this direction (tracking emotional states and metacognitive skills, for example). We believe it is critical that the next generation of AI-based learning technologies both support noncognitive skill development and recognize different forms of them in action, thus more effectively approaching the goal of being more inclusive.
What would it look like for an AI-based system to support noncognitive skills? In our proposal we laid out a use case of children learning to code. Programming is unique in that it involves creation of an artifact whose behavior can subsequently be observed – kids write programs, not always understanding the resulting behavior, then run them to see what happens. To get a program to work, one must persist through working on iterations, make goal-directed changes in response to those observations, and, along the way, gain technical knowledge related to how coding works. That persistence is interesting both from Farrington’s equity-perspective, and also from an AI/ML modeling perspective (how can we detect it? how can we support it?). There has been excellent preliminary work on AI and persistence. In programming we can look at patterns of writing, running, and modifying code, as well as how children watch the results (are they taking the time to really observe behavior of code?).
We will also investigate academic resilience (how kids recover from struggle and learn when they are unsuccessful), as well collaborative communication (how they talk to each other and teachers about STEM topics). Alongside these skills, we will also collect additional assessments, like sense of belonging (in STEM), interest/identity in STEM, and self-efficacy, which are all important factors that relate to these three noncognitive skills. Pointing the power of AI at these difficult real challenges, combined with rigorous educational research and aggressive outreach plans, will collectively define the road ahead for the INVITE Institute. As PI and Director, I embrace these challenges wholeheartedly and feel extremely proud and fortunate to be able to work with such a creative and brilliant team.
To end this very first INVITE blog post, I’ll share the story of how we arrived at the name of our institute. The first part of the story goes back to some of my early work on building pedagogical agents for informal learning – in work with the Boston Museum of Science in the late 2000’s (I was at the USC Institute for Creative Technologies at the time), we built “Coach Mike” to help kids learn to program a robot at an exhibit. We were proud of our work! He was an appealing, Pixar-like character, was funny, provided helpful guidance, and museum visitors seemed to love it. When we demonstrated the system, the director of research and evaluation at MOS, Dr. Christine Reich, was very positive and liked Coach Mike. But she gently questioned his strategy to jump right in and to offer help. She explained that the philosophy at MOS was not to just provide information, but rather to invite visitors to engage with science. We completely changed how Coach Mike behaved to be more responsive and supportive (for example, we added a “help” button to the exhibit). Since learning more about the MOS approach, I have wondered how we can design inviting and welcoming educational technologies that help students feel comfortable using. When the call for the third cohort of institute proposals was released, I was drawn to track 6A which focused on the vision of “Education for all.” Institute research should “aim to reduce achievement gaps, improve access, and address the needs of all learners.” After exploring many possibilities, we came back to the idea of seeking systems that capture nuance and understand learners in new ways… the NSF AI Institute for INclusiVe and Intelligent Technologies for Education—INVITE—was born.
We “invite” you to follow us on social media (Twitter and LinkedIn) and to join our mailing list. Many opportunities for collaboration, including use of publicly available data sets and tools, lie ahead and we are eager to explore partnerships with teachers, schools, nonprofits, companies, and more to collectively advance our mission to build the next generation of educational technologies that can understand and support teaching and learning in fundamentally new ways!