Making AI Work for Education: An Interview with Carlo Schmidt (Part 2)

Welcome back to Formative’s four-part interview series with Director of Product Carlo Schmidt on how Formative is integrating AI into the platform.

Part 1 covered how Formative first got involved with AI, and in this segment, Carlo talks about the philosophy behind how the team approaches AI, as well as how educators and learners can get the most from this exciting new feature.

This interview has been edited and condensed for length and clarity.

Formative Team: It strikes me that you all are not looking to replace the teaching workforce at all but rather save time and use AI like an independent data analyst. That works for teachers, rather than in place of.

Carlo Schmidt: That's why we exist today: we want to save teachers time. It’s true to our mission about real-time data, about helping the teacher to synthesize the data as it's coming in. The response-frequency charts that we released: where you have the bar charts showing which responses students were choosing that updates in real time as data is coming into the system? That's another tool the teacher can use to process the incoming information. 

Formative Team: I'm really interested in what goes on under the hood. Can you give us a little perspective about how you program around AI within the Formative platform?

Carlo Schmidt: With the math keyboard or Show Your Workspace, those are all consolidated and goal-oriented experiences. Use the math keyboard because you're creating a math question. You want the students to respond in Show Your Work because you know that students are going to be drawing on their screens or uploading images. With AI, the promise and the difficulty is that it's an endless ocean. And you can't just cast your rod in the water and expect to catch something good. You’ll definitely catch something, but unless you're really explicit or directed about what you're looking for, you might not get the results you want. We're practicing, and we ourselves internally are playing and working with the AI to help us figure out how we can help it to generate meaningful content.

If you ask for a question on addition, you're going to get a question on addition, but who knows what specifically you're actually going to get? It's different if you ask it that versus if you say, “Hey, I am a fifth grade math teacher.” Now it knows “fifth grade” and “math class.” “My students are working on addition: can you generate a question for me?” The more criteria that we can feed it, the better. If I say, “Hey, I'm a fifth grade math teacher. We're covering this standard today. This is the skill we're covering. Can you generate five questions for me on that and make each [question] one of varying difficulty?” Or, “Make sure they're DOK Level One or DOK Level 2.” These criteria aren’t explicitly clear at the outset and figuring this out is the foundation for generation on meaningful content.

Formative Team: What sorts of questions guide your thinking in this regard?

Carlo Schmidt: What information do we have that we can feed to the AI to help it to make the teacher more productive or guide that experience for the teacher? We don't want to just replace Google with AI. We want [the AI] to come out with meaningful results. That's how we put those guardrails on the experience. We're looking for opportunities of what we already know about our teachers and students to put those guardrails on so that it will be even more productive and effective.

Join us next week for Part 3 of this interview: Carlo addresses some of the controversies surrounding recent developments in the generative AI field and talks about why collaborations between it and Formative help to combat bias, rather than further it.