By Carol Chestnut, Laurie Speranzo, and Michael Telek

Institute for Learning
Agency is the capacity and propensity to take purposeful initiative. [Ferguson, R. F., Phillips, S. F., Rowley, F. S., & Friedlander, J. W., (October 2015)1] Making space forteachers to build their capacity requires providing evidence on which to reflect. When a teacher is alone in their classroom, actively engaged in listening and responding to students during an Accountable Talk discussion, it is near impossible to have a record of the questions asked. The Talkmoves application provides real-time quantified evidence of the questions asked, the categories into which they fall, and a student to teacher talk time comparison. The feedback from the Talkmoves platform provides an opportunity for teacher agency.

The Talkmoves Platform

Talkmoves, designed by the Institute of Cognitive Science at the University of Colorado – Boulder specifically addresses providing real-time feedback related to the use of Accountable Talk moves. Members of the project team have leveraged the IFL’s Accountable Talk Moves to create a platform that provides teachers with feedback on their classroom discussions. It encourages teachers to look at their talk data and decide how they can enhance their own practice. The Colorado team chose classroom talk for analysis because the concepts are based in research and documented classroom outcomes. Victoria Bill and Laurie Speranzo from the Institute for Learning were logical partners for the project because Accountable Talk was established as a teaching tool at the University of Pittsburgh’s Learning Research and Development Center and the IFL fellows have many years of expertise in the field.

How the APP Works: Coding Accountable Talk Moves

The team’s objective was to build an application that would automatically analyze talk from math classroom recordings and provide feedback to aid teachers’ implementation of discussions. The research identified how talk moves can be classified and coded by spending nearly a year teaching the machine talk moves through transcripts and recordings of classroom scenarios. The talk moves were then associated with words and math terms that teachers and students would typically use in a math discussion. In total, 160,000 sentences were coded so that the machine can accurately correlate what is said in classroom discussions to six teacher talk moves and four student talk moves. The moves are broken down in the table.

Institute of Cognitive Science (2019). Talk Moves: Real-time feedback helping mathematics teachers foster effective class discussion, Boulder, CO: University of Colorado Boulder. Retrieved from

Once the technology was in place, field studies gathered feedback from volunteers—21 math teachers from two separate Colorado school districts with their students—who gave TalkMoves a test drive. The field data has been accumulating since 2019. On average, teachers recorded and uploaded ten lessons per teacher for this initial use, Karla Scornavacco, Director of Field Experiences, reported. The program is designed to automatically produce an analysis of their lesson videos. The results of their confidential assessments were typically returned in about an hour and the teachers spent five to ten minutes reviewing the feedback. This is rapid, real-time, personalized feedback.

Application in Action

The process begins with a teacher using their Swivel (a cloud platform video review and collaboration program designed specifically for schools and schools of education) equipment to record and upload their video to the cloud hosting site. (See the TalkMoves website for a more complete description of the equipment.) From there the software automatically takes over, analyzing the lesson in a matter of seconds. An email alerts the teacher that the analysis is available, and they access it by logging into their TalkMoves account.

A wealth of information becomes available when they access the site. Charts and graphs outline four main questions surrounding student and teacher talk, such as:

    • Are students/teachers using talk moves?
    • Which talk moves are they using?
    • Which categories do those moves cover?
    • When are these moves taking place?

Knowing that Accountable Talk discussions need to be accountable to learning community, content knowledge, and rigorous thinking, the graph above indicates that the teacher is asking questions related to all three features of accountability. Seeing that there are significantly more moves related to content knowledge and the learning community, the teacher may recognize the need to plan more questions for rigorous thinking.

You can check out additional the sample feedback by exploring the TalkMoves website at

Beyond looking at raw data points, TalkMoves also monitors what is being said and reports the ongoing trends. It highlights how student/teacher moves change over time, the talk moves they use on average, and how the talk move categories change over time. “I think teachers seem to really like this feature of the application,” said Jennifer Jacobs, faculty member of the Institute of Cognitive Science and Associate Research Professor of science education. “They can really see what they’re doing, and when we talked to them, they usually can make some sense of this.”

For teachers unfamiliar with Talk Moves or how to integrate Accountable Talk discussion into the classroom, there are professional learning resources available on the site that are useful for sharpening Talk Moves skill sets. For those teachers who want to become more proficient at facilitating AT discussions, the experts on Accountable Talk® at the Institute of Learning are always ready to provide professional development to strengthen teacher understanding by describing how AT math discussions encourage student agency.

Potential Uses of the Application to Allow for Teacher Reflection and Agency

Talkmoves provides snapshots over time that can show evolution of practice. Those snapshots can show where one of the features of Accountable Talk is not being attended to, but more importantly, it also tells what is being done with fidelity and where teachers’ pedagogical strengths lie. If colleagues are using the app, discussions in professional learning community (PLC) meetings can center on actual practice and a rich exploration of what is working (and not working) in their math discussions and why. If a teacher using the app is working with a coach, the post-lesson conversations can be grounded in the specific evidence collected by the app. And if a teacher is working alone, the app continues to provide the opportunity to reflect on practice and set pedagogical goals. The Talkmoves team plans on extending this application and related studies, and would like to hear from school districts interested in collaborating on future research.


1Ferguson, R. F., Phillips, S. F., Rowley, F. S., & Friedlander, J. W., (October 2015). The Influence of Teaching Beyond Standardized Test Scores: Engagement, Mindsets, and Agency. Achievement Gap Initiative (AGI) at Harvard University. Seattle, USA: Raikes Foundation. Retrieved from]