Can AI That Lets Kids Struggle Productively Be Better?

Some research out of Stanford​ by Rose E. Wang and Megha Srivastava points out the nuances of using AI to help students learn and that sometimes the tools that promote productive struggle are the better option even if students think they are less helpful. (I’ve written about the need for productive struggle in math on Edutopia, and so much of this research resonates with me.)

One exciting direction is using AI to help human educators create better moments of productive struggle for their students. In earlier research, we observed that novice educators have difficulties helping struggling students, particularly under time pressure.
. These educators were not sure how to nudge students and come up with the right thing to say on the spot. Without guidance on how to effectively foster productive struggle, they frequently defaulted to providing the solution to the student. This meant missed opportunities to turn a student’s struggle into meaningful learning!
The Stanford AI Lab Blog “Productive Struggle: The Future of Human Learning in the Age of AI”

This research paper overviewed a tool they created called Tutor Co-Pilot, designed to help teachers and tutors. They found that beginning tutors often gave the answer instead of more nuanced approaches as done by more expert teachers when faced with student struggles. This tool prompted the tutors to learn how to help students (without giving the answer). This may be far better for helping education than just giving everyone chat GPT.

📚​Read the overview of this study.

The Evaluation Paradox

The section of the paper talking about our traditional paradigms for evaluating AI systems points out a major flaw in what we’re doing: using. user satisfaction ratings doesn’t give an accurate measure of if the tool actually helps students learn more.

Here’s what they discovered

In our work on evaluating interactions between humans and AI systems 
 —in this case, language models—for information-seeking tasks like question answering, we also observed a disconnect between users’ views of helpfulness and their task performance: the language models which users self-reported as helpful were not always the ones that led to higher task accuracy 

From looking at the research it looks like our evaluation methods need to change and the research of our colleagues in higher education is more important than ever.

“Foster Productive Struggle by Empowering Teachers”

Now, if I could take one headline out of this article and share this one, it would be it!

Teaching is an art.

Kids who struggle will often go for the tool that gives them the answer but doesn’t help them with the process of getting the answer.

Great teachers know how to help kids through the struggle in terms of motivation, encouragement, asking great questions, and helping kids celebrate when they win.

No AI can do that.

AI Can Empower Teachers But Cannot “Fix” The Systemic Shortage

I’ll never forget the moment I saw a major tech company state that their AI was going to solve the teacher shortage problem.

Solving the teacher shortage problem involves respecting teachers, honoring their artistry, empowering them with the tools they need, supporting them and their students, and understanding that kids who struggle need extra help.

This week’s radio show on Cool Cat Teacher talk is on the current state of what works in reading instruction and one of the things that came through loud and clear is that every child is different and great teachers know how to use the tools and tactics for individual children’s needs.

This is the kind of research I want to read more of. Take time to read the article and discuss it with your team!