A few years ago, when Ashley Yazarlou was teaching middle school, an instructional coach asked her to estimate how much time she let students talk in class. She took a guess: “Probably about five minutes.” The coach also asked her to record a lesson on her smartphone.
Her coach played the recording and counted with a stopwatch, and Yazarlou was shocked when he revealed the actual figure: Students spoke in that class for just 30 seconds.
“Seeing that and black-and-white was like: ‘Are you for real?’” she recalls. “I started being more cognizant after that about when I was giving students talk time that I was really letting them talk.”
Today, Yazarlou is the instructional coach for many of the middle and high school teachers in her new district, Hemet Unified School District in California, near Palm Springs. Taking a cue from her past, Yazarlou sometimes asks teachers to record themselves on their smartphones just as she once did to help surface the kinds of observations that proved so valuable to her.
Except, instead of breaking out the stopwatch, she’s outsourcing some of the heavy lifting to artificial intelligence—part of a growing trend of leveraging the same technological advances that make general interest tools like ChatGPT so powerful to meet the goal of enhancing teachers’ professional development and coaching.
Lately, that’s taken the form of speech recognition algorithms and chatbots, which offer teachers powerful, just-in-time tools that can be accessed anywhere, providing opportunities for reflection outside of the usual coaching cycles.
Bringing AI Into Teacher Coaching
The large language model technology that undergirds AI tools like ChatGPT can do more than carry on back-and-forth conversations. New advances have made it possible to transcribe audio files and mine them for insights. That’s the premise behind the tool Yazarlou uses from the edtech startup TeachFX, which offers a data analytics platform that provides detailed reports based on lesson recordings uploaded by teachers.
Five years ago, when CEO Jamie Poskin founded the company, it was focused on showing teachers how much time they talked during a lesson compared with their students—based on research from John Hattie and others showing that students perform better when they speak more in class.
Fast forward to today, and TeachFX’s data reports now identify more than 20 different insights, such as whether teachers are building on what students say, whether they’re asking open-ended questions that push student thinking forward, and how often they’re using academic language in lessons. The technology then analyzes the content of the files that teachers upload, picking out specific words and phrases.
Instead of providing only a staid analytics report, the tool also presents its findings conversationally, pointing out strengths and weaknesses the way ChatGPT might. It might, for example, flag the questions being asked in the lesson and ask teachers to reflect on how to reword them to better differentiate for learners. It could highlight extended periods of direct instruction and pose reflective questions, such as: “Do these stretches of teacher talk support checks for understanding, progress monitoring, or deepening of student learning?”
One important point: TeachFX does not make recommendations. Those are best left to teachers and their coaches, Poskin explains. “We’re good at providing the right information and surfacing the most relevant moments for teachers to reflect on,” he says. “But we don’t pass judgment on their teaching. We simply show them data and let them make sense of it.”
The recordings and analysis are private to teachers, who must agree to share them with coaches. For schools with enterprise licenses, a small number of anonymized, top-level insights, such as the teacher-to-student talk ratio over time, are shared with administrators.
Last year, Yazarlou used TeachFX with several middle school teachers who were looking for ways to boost academic discussions. Many of her teachers were struggling with getting students to speak up in class or to stay on topic when conversations got going.
Once Yazarlou’s teachers recorded and uploaded their audio, it took about a day for the system to produce a report, and then they reviewed the analysis together, looking at the types of questions teachers were asking and which words they used most often. (Another feature provides a word cloud visualization of the most commonly used words and their frequency). The data also revealed things like how many seconds of “think time” teachers gave students after asking questions, and how many focus questions they asked—ones that guide students to think deeply about the issue under discussion—versus directive ones like, “Did you get out your pencils?”
The goal was to use the tool as part of a true coaching inquiry cycle, where teachers set goals, review their lessons after the fact, make adjustments, and try again over the course of several weeks.
By and large, the data has led to better conversations and deeper reflections that happen faster than usual, Yazarlou explains. “When I go in to do post-observations after this whole process, I always see major growth from where they started,” she says. “I won’t say it’s from using this one tool in one lesson, but I do think seeing that data in black-and-white is a huge ‘aha’ for them.”
The Benefits of Guided Reflection
Reflecting on audio data is just one approach impacted by AI. Some teachers benefit greatly from seeing themselves on video, which can provide a stark portrait of their mannerisms and the ways they interact with students.
Long before AI technology came to teacher coaching, teachers at St. Vrain Valley Schools in Colorado would video record themselves using their smartphones or laptops and review those lessons with a coach. This was especially helpful during the remote learning of the pandemic, explains Courtney Groskin, an elementary instructional coach for the district, located north of Boulder.
Groskin’s teachers would upload videos to Edthena, a platform that allows both teachers and coaches to time-stamp and annotate clips asynchronously—that is, at different times—and discuss their observations in a process not dissimilar to the postmortem sessions they would have previously conducted in person.
Using the platform always required feedback from a coach—until now.
Edthena’s newest tool, called AI Coach, features automated interactivity, courtesy of artificial intelligence, that lets teachers design, implement, and iterate on their own professional goals via chatbot. Unlike TeachFX, the tool does not analyze video or audio, but instead acts as a sort of interactive assistant built into the video platform to help users refine and organize their thought process. It works by “guiding and mediating teacher thinking,” according to Adam Geller, the company’s CEO.
For instance, a teacher might tell the AI assistant, named Edie, that they want to focus on differentiation or student engagement. Edie will then ask them to expand on their goals and how they will be measured. When teachers upload a video of their lesson, Edie will suggest some questions to reflect on while they watch, such as: “What different types of questions am I asking?” and “How am I adjusting instruction based on student responses?”
Teachers can annotate their videos with time-stamped comments on the strengths and weaknesses of their performance. While Edie does not process the content of the video, it reads the comments teachers leave and adjusts its suggested reflections accordingly. Afterward, it will ask whether the teacher needs any help meeting their goals and provide curated online resources for those wishing to learn more about their chosen topic. And finally, it will guide them to turn their insights into a detailed action plan for improvement.
Geller likens the process a little to talk therapy in that the Edie doesn’t dictate, or even suggest, what teachers should be doing to improve, but instead helps them come to those realizations naturally. “The conversation, the guiding questions, and the particular way it flows are really about leading a teacher through that reflective process, just like an in-person coach would do,” he says. “The best coaching is what you discover about yourself.”
Sessions with AI Coach are private to teachers. While coaches or administrators can see where teachers are within a coaching cycle, and how much overall work they’ve done with the platform, they don’t have access to the specific videos, conversations, or action plans teachers add.
Benefits of Guided Reflection
A substantial body of research supports the idea that teachers can improve their practice by reviewing their lessons. Much of the existing research is centered around video, but studies have shown audio recordings are useful for focusing on specific language used in lessons and as a tool to prompt reflective analysis.
On the video side, researchers at Canada’s Université Laval conducted a meta-analysis of 89 studies and found that video is adept at making teachers more self-reflective regarding the way they teach and interact with students. A number of the cited studies showed that teachers were able to leverage recorded lessons to improve the effectiveness of their teaching.
But other studies reviewed by the researchers revealed that using video as a reflective tool is not always intuitive and that teachers do not automatically possess the ability to be self-critical or ask the right kinds of probing questions to move their practice forward.
In Groskin’s experience, the most fruitful self-analysis comes when teachers work with another person—or, now, a bot—who can provide a gentle nudge or a new perspective. “You can choose to see what you want to see when you’re viewing a video,” Groskin says. “Having a coach kind of brings everything to the surface.”
In other cases, teachers will notice on their own that they’re talking too much during lessons or not focusing enough on in-class discussion, but struggle with deciding what to change. When they share these observations with the AI coach, they might get resources (including ones from Edutopia), such as icebreaker games, clever ways to get students talking, or how to set up smooth transitions that keep students engaged in learning.
Nearly all of Groskin’s teachers who engage with AI coaching experience some improvement, such as reducing off-task behavior or increasing student engagement. “They’ve completely changed the way they deliver lessons once they have the chance to watch themselves and think, ‘Oh, I’m talking too much. I’m losing all the kids in my class.’”
But mainly, she’s seen teachers who use AI Coach—including those who were once skeptical—become more excited to engage in the coaching process.
That tracks with Yazarlou’s experience as well. The biggest gains come when teachers realize how helpful it is to think reflectively about their pedagogy and practice. Ultimately, data, and even on-demand coaching, are not what moves the needle for teachers, Yazarlou says. That comes from the sense of empowerment teachers gain when they realize how they can improve their practice on their own, using the same critical thinking skills they seek to help students develop. “It really is just a way to reflect on teacher professional growth.”