We’ve used AI to enhance education for a quarter century. Here’s how newcomers fall short.
Suppose you’re a 5th grader solving a word problem by adding fractions with equal denominators.
You know how to add. You know what a numerator is and how it’s different from a denominator. Hey, you even know how to multiply fractions, to boot!
But there’s a lot of information in your brain, and, if you’re like a typical 5th grader, you’re about to make a common mistake: you’re about to mix up the procedure for adding and multiplying fractions.
Our approach to AI
How do we know this? At Carnegie Learning, we’ve been teaching math to students since our founding, and we have 25 years of data to inform us where common pitfalls occur. Our trademarked AI solution, MATHia, has been trained to recognize why students make the mistakes that they do so we can guide students through their thinking and help them get it right next time.
When a student makes the mistake of applying multiplication procedures to add fractions, we don’t just give them the right answer. Instead, we leverage our years of experience and our state-of-the-art machine learning to help students change their approach to the problem.
“AI [in] the MATHia platform provides individualized instruction based on my students’ need as they progress through mastery of each skill level,” explains Sofiya Padela, a 6th grade math teacher from Massachusetts. “MATHia provides just-in-time support and gives students hints when they are stuck.”
Research shows that having students do the work helps them learn more than when AI tools make the decisions. This results in better concept mastery over time. When students rely on AI tools for answers, they skip the productive struggle that leads to deep learning. That’s why we offer premier AI solutions that make students smarter—not reliant on technology.
How our competitors fall short
As newcomers to AI, our competitors are offering some solid problem-solving EdTech tools. But their shortcomings are demonstrated by user feedback.
“[A Teacher] resists solving a problem for the student … because the teacher knows the student is one question away from realizing the answer themselves,” writes former math teacher Dan Meyer. “AI chatbots display none of that sensitivity, immediately answering a question right when a student types it in.”
Meyer wrote of one of our competitors: “When K-12 students are in a classroom trying to learn mathematics, the majority of them do not wish to have a conversation with their computer.”
We couldn’t agree more.
That’s why our trademarked AI offers adaptive hints targeting skill gaps we can identify through advanced machine learning. By leveraging our understanding of how students think, we’re often able to recognize their reasoning before they do. This allows us to nudge them in the right direction through targeted skill practice.
Daphne Goldstein, an 8th-grade student, described her experience with the same competitor’s math tool: “If you get confused, then what? You can click on hint. But doing that comes at a cost: you can’t get ‘credit’ anymore for getting a right answer on your own. ... After a wrong answer, you can click the get help button and receive a lengthy explanation on how you should’ve done the problem.”
MATHia offers a different approach. Students don't just enter the final answer to a problem; they show their work. This allows MATHia to give them feedback step-by-step. Explanations take into account the student's strategy and specific misconceptions, so a short bit of guidance can get them back on track.
At Carnegie Learning, we don’t believe in high-stakes hints or lengthy critiques. We believe in productive struggle and learning by doing. That’s why we’re confident that our AI tools keep us ahead of the curve.
Our words of wisdom
So, to all of our competitors who learned about AI thanks to Sam Altman, we have some advice: the most sophisticated generative AI will not make literature students smarter. The quickest problem-solver won’t make math easier. The most verbose chatbot will not make kids love learning.
Instead, the best AI tools for education are those that emulate student thinking and allow for mistakes. They’re the tools that embrace productive struggle and help students master concepts instead of generating right answers.
If you only want access to smart robots, go to them. If you want to nurture smart students, come to us.
Steve Ritter is Founder and Chief Scientist at Carnegie Learning. He has been developing, analyzing and evaluating educational technology for over 20 years. He earned his Ph.D. in Cognitive Psychology at Carnegie Mellon University and was instrumental in the development and evaluation of the Cognitive Tutors for mathematics. He is the author of numerous papers on the design, architecture and evaluation of Intelligent Tutoring Systems and other advanced educational technology. He currently leads the research team at Carnegie Learning, focusing on improving the educational effectiveness of its products and services. Each year, over 500,000 students use Carnegie Learning’s mathematics curricula.
Explore more related to this authorAt Carnegie Learning, we don’t believe in high-stakes hints or lengthy critiques. We believe in productive struggle and learning by doing.
Dr. Steve Ritter
