Kristen’s Corner Winter 2026 – Khan Academy Blog

By Aviv Weiss, Khan Academy Districts

Each quarter, we sit down with Dr. Kristen DiCerbo, Khan Academy’s Chief Learning Officer, to reflect on what she’s seeing in the field, in the research, and in the data. In this interview, Kristen talks about how leaders can think about tracking each student’s proficiency by skill.

Aviv Weiss: When districts pilot AI-supported instruction, what early signals should leaders look for to know it’s actually helping learning (even before test scores show up)?

Kristen DiCerbo: Before you adopt any tool to improve learning outcomes, you should have a plan for how you will determine whether it works. You should be looking for digital tools that have ways to measure learning on the platform. For example, on Khan Academy, the number of skills students get to proficient status is tracked and reported. You can then monitor these metrics to see if learning is increasing. The provider should be able to provide evidence that learning on the platform transfers to improved results on other assessments. 

The provider should be able to provide evidence that learning on the platform transfers to improved results on other assessments. 

Dr. Kristen DiCerbo

You may notice that if you adopt AI tools not specifically designed for education that there is no way to track the impact of use of the AI on learning. That is something to consider when adopting those tools. In this case, the burden is on the educator to administer measures of learning, track them over time, and correlate the gains to use of AI. 

Aviv: What kind of signals do you and your team look for to see that what you’re developing is actually helping learning? 

Kristen: We have a set of signals we use internally at Khan Academy to measure learning as we develop new features and experiences. They range from very fine-grained to more broad. At the most fine-grained, we look at students’ success on individual questions. For example, if a student answers a question with help from Khanmigo, we look at whether they are more likely to get the next question they answer on that skill correct (without using Khanmigo). If we change something about how Khanmigo interacts, we look at changes in that metric of getting the next question right.

For example, if a student answers a question with help from Khanmigo, we look at whether they are more likely to get the next question they answer on that skill correct (without using Khanmigo).

Dr. Kristen DiCerbo

On the broader side, we look at the number of skills students get to “proficient” status on the platform. As students work on skills on Khan Academy they progress from attempted to familiar to proficient to mastered. Our research shows that students who get skills to at least the proficient level are then able to successfully apply those outside our platform (including on standardized tests). We know that students who average two new skills to proficient each week see significant yearly gains on test scores. So, when we are piloting our new student experience, we are monitoring the percentage of students who get at least two skills to proficient each week and are always trying to increase the number of students reaching that goal.

Learn more about Khan Academy’s work with US districts at khanacademy.org/districts