The Essential Role of “Friction” in Student-Facing Artificial Intelligence Products

Artificial intelligence has found its way into classrooms through tools designed to support teachers and to streamline their workdays. Many proponents of the fast-evolving technology say teacher-facing applications of AI come with less cost and less risk.

But interest in student-facing AI tools has not gone away, especially as adoption of the tech and its capabilities grow. Neither have a host of accompanying questions about the true impact of artificial intelligence on student learning.

One prevailing fear is that artificial intelligence will negatively impact students’ critical thinking skills and creativity, inhibiting their ability to judge information with discernment.

In an interview with EdWeek Market Brief, Caleb Hicks and Nate Sanders, executives with SchoolAI, an AI-powered platform aimed at helping teachers and students, spoke about those tensions and how their product is trying to navigate them. They offered their vision on the balance between leveraging AI for personalized learning and doing so in a way that encourages students’ academic development.

Hicks, founder and CEO, and Sanders, chief experience officer, of the AI-powered classroom experience platform, also discuss what the evolving tech landscape looks like, and how ed-tech developers can design responsible AI tools that protect, but also encourage critical thinking across K-12.

Much of the recent focus of AI in schools has been on tapping into the tool’s power to support teachers. But SchoolAI has student-facing components. How do you mitigate for the risks involved in serving students with this technology?

Sanders: There are really clear and obvious problems to solve for teachers, like content preparation, adaptation, helping refine feedback. Everyone working in AI gets excited about its ability to be a tutor, though — it is the holy grail of education, to be able to personalize based on your interests. The complicated part is, like 15 years ago, we started thinking about social media, and how do you build that in a safe and managed way? We are all in that work right now. Critical thinking is a very important skill. Nobody wants to give AI to students in a way where students are worse off for it.

The most popular version of AI today is an assistant. It does things for you. A tutor is not there to do things for you, so you have to wrangle AI into being a productive tutor that adds friction — so that you learn, versus it doing the thing for you.

One of the first things we built was a writing coach that would not write the essay for you. It’s something that can be done to add to the student’s experience and compound their ability to do critical thinking, by having a tutor nudge them on the path of critical thinking or in productive writing.

Hicks: I want to underscore that it’s student-facing, teacher-led. Students should be experiencing deeper learning led by the teacher in the classroom.

Sanders: It’s a very important component for us to take what’s going on in those conversations, and in those experiences that the AI is working with students on, and surfacing up the important insights to the teacher. So if these six students are really struggling with this specific concept, let’s tell the teacher, so that the teacher can do something about it, and not just rely on AI to solve all the problems.

What do you say to districts that are concerned about student-facing AI tools?

Sanders: For those that are concerned, I want them to hold that high bar of doing their due diligence around, is this product safe and observable, can it be governed, because that’s important. But we are seeing a huge amount of accelerated demand for [bringing AI more directly to the educational experiences of] students the last few years. There’s a life cycle of how schools and districts have evolved and thought about how they’re going to think about AI inside the classroom, and people who used to be only focused on teacher use cases are increasingly very focused on student use cases.

So how are you trying to ensure that level of governance for your product’s student-facing AI?

Sanders: The way that we think about the safety aspect [is that] we ensure that observability and governance is at the forefront of everything, so that a teacher, a principal, is able to observe, see, understand, get critical alerts around every single conversation that happens inside of SchoolAI.

It’s also a lot of prompting, orchestration, and a lot of thoughtful work that goes into making sure that there’s clear guardrails and a lot of alignment to making sure the responses are accurate, adherent to teacher instructions, and appropriate. We grade every single AI response on how it is engaging the students based on those dimensions.

Hicks: Students have to learn how to use this tool productively and safely and responsibly as a critical 21st Century skill. We have to teach students to be adaptive because there will be new things that they have to know how to use. This is one of them. For students to get into the colleges that they’re going to want to get into, for them to get into the jobs that they want to get into, many of them are going to have to know how to use this.

As schools recognize we have a responsibility to teach students how to use AI, that is where some of that demand and interest is coming from. And that we hope, we as an industry, are looking at super-responsibly.

You mention the value of creating “friction” for students to practice critical thinking with AI. How do you do that while still serving an array of learning goals?

Hicks: You can use the same technology to NOT do the things that you ask it to do. You can say to it, “Be a writing coach, but don’t write the essay.” You have to say that — over and over again — in increasingly strong ways to make sure that it doesn’t actually do that, and that requires some nuance and complexity.

But it’s about how you use the building blocks of the technology, versus just saying, everyone gets their own AI assistant. We’ve built a lot of our platform in a way where it is consciously adding friction. It is not there to give you the answer. It is there to help you get to the answer.

What are other examples of how critical thinking can be infused in AI programs?

Hicks: Another way that we think about teaching critical thinking with AI as a tool is building literacy of how AI works. When you understand hallucination as a concept, you learn that the AI, even if you’ve given it all of the right information, can still slip up and say the wrong thing. If you teach students about how they need to use a critical eye in evaluating how the AI responds and that you can’t just take at face value what AI spits out at you, it gets better.

Can you give an example of how SchoolAI tries to accomplish that?

Hicks: One of the things that we challenge teachers to do the first time they’re using AI with their students is to play two truths and a lie with the AI. This shows the students very practically that you have to look at this with a critical eye and adding friction in the right places, so they understand what AI is great at and what it’s not great at.

It’s about telling students that AI might get something wrong, so they’re not just trusting it implicitly — they’re critically evaluating their interaction with AI.

What are the key questions product development teams across the education sector need to be thinking about, in trying to nurture critical thinking?

Sanders: Companies should be thinking about, what are the instructions for how I would want an AI [product] to respond to a student, to be able to facilitate important outcomes — and work backwards from that.

That’s a really good framework for anyone that wants to build a product successfully because if you’re saying, I want a student to be able to master a concept, to be able to understand it clearly, and to be able to have thought about it critically and have experienced that on their own, you’re going to have a lot of instructions that go into how you orchestrate and build with these models that is more Socratic in nature, which encourages critical thinking.

Hicks: Most startups are in the business of reducing friction. On our teacher side, we’re trying to create tools that reduce friction for teachers so that they can show up in a more relaxed, productive headspace with their students, so that they can provide the best experience that they’re looking for.

On the student side, you have to do the opposite. You’re creating friction. So the questions you have to start asking are, what is the job you are wanting to do with AI for this very complex education ecosystem?

How would a company convey this need for “friction” to school districts?

Hicks: You have four very different stakeholders in the school right now. You have teachers, students, school leaders, and parents that all feel very different motivations and feelings about how the school is working, and very different things about how the other three stakeholders are showing up.

So what is the job you are trying to do, and how do you apply the technology to do that? Something I would say to anyone building with AI is, it’s not just about giving it to people and hoping it does what they want. You have to be very intentional about treating the technology as a building block for a specific job to be done.

How much of the responsibility falls on districts in making sure that the tools are being implemented correctly and that students are using them properly to build critical thinking?

Hicks: We try to make it as easy as possible. The idea that a teacher who is already very busy supporting kids, coming up with curriculum, interacting with parents, doing the job of teaching, is going to be on the forefront of writing prompts for AI and orchestrating AI agents in a way that helps their students learn, it’s just not possible, and not the right expectation.

We try to give training and development for using AI in the classroom, to school leaders and to the teachers and the instructional coaches, but it’s not enough. We continue to work on the product side. How do you make it easier for students to use it in a safe and managed way and to learn how the technology works?

How is SchoolAI trying to address those ease-of-use questions?

Hicks: Districts that we see adopt really well have instructional coaches, tech directors that are focused on that. But we also provide them with all kinds of different learning modalities and training options, so online, on demand, virtual, instructor-led and in person.

We’ve also gotten multiple districts together for an onsite, full-day PD. So there’s good rigor and scaffolding built in around, how do we ensure districts and schools are AI literate? We’re able to help facilitate that so that it’s not incumbent upon them to do it.

Even before we actually sign and get going with them, a lot of our training is completely free. We see a world in which we can have a training hub — product agnostic — whether or not they use SchoolAI or another provider in the space.

How else are you preparing teachers to, in turn, prepare their students to use AI in a way that builds on existing skills?

Hicks: So much of this is about, how ready is the school and the teacher to try something new? What you see is a typical adoption curve. You have some teachers that are so excited about the technology for themselves and for their students, and we see teachers creating new types of tools and experiences on SchoolAI every day.

We try to help the district find those people first and create a plan of, what does success of using AI in your classrooms look like? Then we design their implementation, their onboarding as a district around that. Sometimes that looks like a pilot with 10 teachers.

What role does building up students’ AI literacy play in all of this?

Hicks: It’s critically important. In the same way that media literacy, digital literacy, and being able to use the internet and Google are critical skills, AI is a great tool that students have to know how to use. There’s literacy around, how do you get the most out of it? But there’s also literacy that needs to be built around, what is it good at and what is it not good at? What are the trade-offs when you use AI?

What responsibility do vendors hold in creating responsible AI products?

Sanders: Districts should invest into training for your teachers, into curriculum and training for your students around AI alignment and safety, but also make sure that they’re working with vendors that value [these things] and that want to help facilitate. AI tools should also be developed in targeted ways with very specific outcomes.

Hicks: If I were a district, I would demand that a vendor have a lot of rigor — it’s part of building trust with schools and districts in being a good steward of their time, their energy, their procurement process, in actually how we build the product. And that goes into our IP and our back end, and our deliberate nature of saying we will only ship the most important features in a way that’s usable and powerful and literate.