²ÊÃñ±¦µä

A big part of iSAT’s mission is to create an AI Partner who helps teachers facilitate collaborative and equitable learning outcomes for small groups in the classroom. This quarter, the team science and human-computer interaction experts in Strand 2 have made important progress in understanding the framework needed to help small groups develop successful collaboration skills and how to support teachers help their students develop these skills.

What Skills Should Our AI Partner Develop?

Strand 2’s Framework Development team is dedicated to figuring out a huge overarching question for our institute: Exactly what skills should our AI Partner help students develop? 

To answer this question, the team hammered away at developing a framework of collaborative problem-solving (CPS) skills our future AI Partner will learn so it can assess and encourage small group interactions. The team’s main goals are to determine behaviors and skills that underlie successful collaborative problem-solving. The team formulates metrics to measure these skills with a focus on how skills emerge over the course of problem solving and with the goal to model metrics in real time, such as the state of CPS can be fed to the teacher and/or AI Partner.

While closely collaborating with Strand 1’s Reinforcement Learning team on the CPS skill measurement, the team will use data collected as part of the AI Collaborative Learning (AICL) environments work underway by Strand 2 to test the assumptions of the framework and further develop metrics used to asses CPS. In addition, the team is also using data collected from a previous NSF-funded project on remote CPS, in which teams collaborate within a Minecraft (a popular sandbox video game) task, to create influence metrics that will be adapted as part of this framework. This study is the first step in formulating a finalized metric for the equitable interaction skill found in the framework. The team also collaborates with Strand 3 to assess what actions should be taken in response to the measurement of a skill.

How Will Our AI Partner Talk to Teachers?

While the Framework Development team has been hard at work figuring out the most effective CPS skills to measure and implement, the Classroom Orchestration group has spent the past quarter collaborating with K–12 teachers on the best ways to present classroom data to teachers in real time. The Classroom Orchestration team conducted two participatory design studies with eight Wisconsin area teachers to determine what type of information would be most useful to them on a teacher dash- board. With the teachers’ input, the team created dashboard mockups and further iterated on the designs with a focus on specific features.

Initial analysis from the studies showed that teachers emphasized that their needs change based on the goals of each specific class. For example, some teachers preferred a general overview that directs them to where they need to focus their attention while others were more interested in diving deeper into the data to see classroom trends, to improve current and future instruction. While teachers found it helpful for group and class level information to be presented during class, most preferred information at the individual level for post-class analysis. 

Based on the dashboard designs created through the Wisconsin-area participatory design studies, the team began mock-ups for dashboards to use in the ²ÊÃñ±¦µä iSAT Lab. The iSAT Lab dashboard mockups are being designed with the Sensor Immersion work by Strand 3 in mind, but they are mostly focused on designs that can span a range of curriculum activities. In the coming quarters, the team plans to develop and use the first version of their dashboard with six teachers in Wisconsin.

iSAT Lab Experiments

Strand 2 also led the effort to create iSAT’s very own lab in ²ÊÃñ±¦µä’s Center for Creativity and Innovation (CINC) in Boulder, Colorado. The lab functions as both a research study facility and a space for all students to gather, network, and collaborate.

iSAT researchers developed studies for this lab in which students, staff, and faculty at ²ÊÃñ±¦µä work on collaborative problem solving while our researchers collect video and audio recordings of their interactions on the same equipment used by our researchers to collect classroom data. The main goal for this lab work is to test collected data on Collaborative Problem Solving skills to advance team science and to look into human-AI teaming strategies. These recordings will also further help our Strand 1 researchers with their data annotation, multi-modal, and content analysis efforts by providing them with cleaner, easier-to-process data than what is collected in noisy K–12 environments. The various teams will then use their improved data analysis processes to optimize our K–12 classroom data.