Let’s get one thing clear: Being more efficient with tasks isn’t the same as making schools simpler, says MIT professor Justin Reich. As director of the MIT Teaching Systems Lab, Reich and his team work to design, implement, and research the future of teacher learning. They investigate the way edtech impacts student learning and help schools implement new tools to enhance and transform classrooms—not to make teachers more efficient, but to help clear the noise so teachers can focus on what they do best.
In the October 2022 issue of Educational Leadership, Reich wrote the article “The Power of Doing Less in Schools,” which discusses the importance of subtracting rather than adding initiatives to make schools simpler. In his most recent research project, Reich explores the explosion of generative AI in schools and the way it has affected homework, assessment, cheating, and learning.
What does it take to really make schools simpler? How do schools respond to emerging technologies but also keep the foundational tenets of learning front and center? How can leaders resist the urge to take on that next “shiny new thing”? Reich’s advice to educators is cautious but optimistic, and starts with listening, analyzing, and not being afraid to admit when you’re wrong.
Why is the concept of “less is more” so important for K–12 education, especially now? And why is it so difficult to streamline priorities and initiatives?
Schools are maxed out right now. All the time, policymakers say, “This is a problem in society, we should turn to schools to fix it.” So we think, maybe the schools should take on physical fitness, or maybe the schools should take on sex ed, or new technologies, or not having kids use social media too much. And it goes on and on and on and on.
We’ve built these really complicated systems with lots of expectations in them, but it’s very difficult to do lots of different things well. To get better at educating, schools have to be more focused on the most important things.
We have to stop adding to schools, even though that’s very hard to do. Anthony Bryk, who was an education researcher, coined the phrase “Christmas tree schools.” The idea is that everyone looks at schools like a Christmas tree that they want to put an ornament on. And people are really passionate about their ornament. So, they put it on the tree, and then someone else puts on an ornament, and then someone else. There are very few who will look at the tree—the school—and say, “There are too many ornaments.”
Meaning, it’s very difficult to remove any individual initiative or program because someone says, “No, that’s my piece. That’s the piece that I really care about.”
As director of the MIT Teaching Systems Lab, you and your team have done a lot of research on how edtech tools affect learning. How have you seen technology help (or hinder) schools to do more with less?
If you interviewed a bunch of 20-year veteran teachers who have seen all kinds of new technologies enter their classrooms and asked them if they were now able to spend more time on the most important parts of their job because of technology, I would bet most of them would say no. I don’t think most teachers would say that new technologies reduce the amount of complexity in schools, as much as the developers of the tools promise to do just that. When new technologies are introduced, they almost always feel to educators like one more thing.
When I was doing consulting work helping schools integrate technology into the classroom, one of the things that I often suggested to people was that they try to not think of new technology investments as a technology initiative, but rather as a learning initiative. Teachers don’t want new technologies. They want tools that help them achieve learning objectives that they care about.
Do you think AI will help make schools more efficient?
One pitfall of efficiency is we think we’re doing something efficiently, and instead we end up doing it differently. The easiest canonical example of this is the Scantron machine. The Scantron machine made the scoring of multiple-choice items much, much more efficient. But I don’t know how many educators would say, “Wow, now that we have Scantron machines, we’re really capable of spending much more time assessing students on the most important things.” I would think there are a whole lot of educators who say, “I wish I spent a lot less of my time teaching my kids to be prepared for multiple-choice tests.”
I don’t think that was the intention of people who developed the Scantron machine, but that’s what happened. In fact, educators were probably hearing a lot of the same rhetoric then that we are hearing today about AI—that it is going to let you be more efficient at these kinds of tasks, which will let you do more important tasks.
I think there is lots of risk in using AI to make processes more efficient. We hear how using AI can make your communications to stakeholders faster. But if parents and students and community stakeholders have a sense that AI bots are writing all the messages to them, they’re not going to want to read them. You will have produced that text more quickly, but you may have turned off the people who you need to read those materials.
Similarly, there are all kinds of ways to use AI to generate curriculum materials faster. It doesn’t mean that they’re good. Now, it might be fine for educators to become more efficient at doing low-stakes tasks, if the low-stakes tasks are not important. There’s plenty of stuff that we ask teachers to do that’s maybe not a great use of their time.
But instead of wondering how machines can do a bunch of tasks that we ask educators to do, we should be asking how we can have educators do fewer tasks. Or instead of wondering how to get machines to spew out communications for people, we should ask how we make sure that we’re using communications parsimoniously so that there are fewer of them to write, there are fewer of them to read, and we’re just saying the most important things to each other.
Your latest TeachLab podcast series, The Homework Machine, involved your team interviewing nearly 100 teachers and 30 students about how AI is changing the classroom and learning. What are one or two big takeaways from these interviews that all educators should be aware of?
One great starting point for people in leadership positions is to really listen to students and teachers about their experiences in schools since ChatGPT has become more widely available. We have an amazing interview with a young man named Leandro, who spent his entire senior year doing all his homework with ChatGPT. He spent a bunch of time using multiple tools to fabricate homework and have it be not detectable. And the interviewer asked, “What do you wish your teachers would know about your AI use?” And he was like, “They really need to learn how to use detectors and to find kids who are doing this.” And the interviewer said, “But you would have been caught.” And he said, “Yeah, I wish somebody had caught me at the beginning of the year and told me not to waste my year not learning.”
Young people are incredibly wise, even though they don’t necessarily have the judgment to be able to say, “Yeah, I should have made myself not do this.” But realizing that students need guardrails definitely made me rethink my position on AI detectors and surveillance of students’ work. I recognize that it’s young people’s job to search for boundaries. And they have to find them—we have to set them—because they will keep pushing and keep searching until they do.
If there’s one characteristic that will serve us well in the years ahead, it’s humility.
What should schools be doing to prepare teachers on how to use AI tools effectively and safely?
Schools and districts need to have a collective response about AI policies. It’s not reasonable to pass AI down to teachers and say, “Figure this out on your own.” We heard over and over again from educators that they felt like they needed to be supported. And we just did a survey in partnership with RAND, with the American Teacher Panel, that shows only 25 percent of teachers get some or a lot of PD or resources about AI. Additionally, 34 percent of teachers say they have some AI policy related to academic integrity, but only 6 percent say it’s clear. The percentage of teachers who are really happy, who feel like they’re really well served by those supports, is a very, very small number (2 percent).
So school leaders need to put together policies and professional development. Here’s the dilemma, though: We do not know what those should be.
No one knows what an effective policy for managing generative AI in schools is because it’s such a new technology. No one knows what good practices are. No one knows what good teaching looks like with AI or what makes for good learning with AI. There is no rigorous research that demonstrates that anything anyone is saying right now about how to use these tools is effective.
That is a tough place for education leaders to be in. If there’s one characteristic that will serve us well in the years ahead, it’s humility. It’s saying, “OK, we don’t know, and we’re going to try some things, but they may be wrong. And we may not know they’re wrong for years.” People should have in their minds that we’re facing about a decade of uncertainty around these things. I think that’s going to be really hard for educators to navigate for a considerable period of time.
So, do you feel like this is a very exciting time or a very terrifying time for schools?
I think I find it more frightening and troubling than exciting. We have, over the last 20 years, tried a technology literacy strategy that goes like this. Silicon Valley invents a new technology. We define a set of skills that are associated with that technology. We usually do that really fast and guess a bunch of ways of teaching those skills.
And then we tell teachers, OK, now teach these skills. And there are not a lot of financial resources behind it. There’s not a lot of new staff that are put behind it. It’s just one more thing that schools are supposed to do. I just don’t think schools have done a great job at that. In part, again, because you can’t ask schools to do more things without taking away other things, and we don’t take things away from schools to make room for new things.
I will say that once you get down to the level of teachers and students in classrooms playing with AI, I think that’s super exciting. I mean, here’s a technology that the best computer scientists in the world don’t understand exactly how it works. It’s this kind of mystery in a box. It does all these really unusual things. It does things in really funny ways. It’s wrong in really funny ways. It expands our capacity in ways that let people play around.
But for these creative things to transform learning, we need to keep centered the question of, Are we giving educators and students advice that we know leads to better learning outcomes? I am in favor of that thing that makes students better writers and communicators, whether that’s AI or tablets or the web or pencils or rocks. But in the absence of evidence of efficacy of AI, I do worry that there’s too much to be rewarded by making up answers and not enough to be rewarded from taking the time to come up with good answers.
So, schools that really dive in to incorporating more AI practices, especially early ones that are untested, will serve their students best if there’s a real substantial effort to evaluate whether or not these things are working. If you’re bringing lots of AI into the English classroom, you should be grabbing stacks of old student work from 2015 and comparing it to student work in 2025. Are you really seeing improvements in whatever it is that you say you want students to be able to do when they’re in your English language art classrooms? And if you’re not seeing differences, or if you’re seeing degradations, if you’re seeing things that are getting worse, you should be ready to revisit that.
Editor’s note: This interview has been edited for space and clarity.
