AI In Education: Fulbright Presentation
Last month I had the pleasure of presenting at the Fulbright Association 's annual conference. As an alum of the research exchange program, I cherish the energy this community brings to tackling the most pressing issues of our day.
It's perhaps no surprise, then, that the subject of my talk was artificial intelligence and its impact on the world of education. As a humanities PhD who works in education technology, this past year has been alternately exciting and ominous, so I wanted to share some of my personal insights from working in this space with my presentation, AI in Education: Insights from the EdTech Trenches.
I began with a parable from sci-fi author Neal Stephenson's 1995 novel, The Diamond Age, Or a Young Lady's Illustrated Primer. Imagining a future in which sophisticated nanotechnology has transformed the world as we know it, the book is often cited as an example of how a fully personalized, intelligent tutor could transform human learning. The protagonist, Nell, lives in the slums northwest of Shanghai, in a world where nation-states as we know them have dissolved into self-sorting cultural associations.
When she's gifted a rare and powerful Primer equipped with AI-like technology, her mentor explains that it will help her rise above her station, if she can combine the education it confers with intelligence:
In your Primer you have a resource that will make you highly educated, but it will never make you intelligent. That comes from life. Your life up to this point has given you all the experience you need to be intelligent, but you have to think about those experiences.
Nearly three decades later, aspects of this prescient vision are materializing—though perhaps not in the ways early technologists anticipated. Drawing from years of research and references from great minds in the edtech space, this post contains highlights from my talk on artificial intelligence in education, examining both its promises and limitations.
Key Takeaways
While the conversation spanned a number of connected themes, I've condensed those I view as the most important into the sections below:
1. Disrupting Standardization
A major touchstone for edtech optimists is education professor Benjamin Bloom's 2 sigma problem. This research describes one-to-one tutoring using mastery learning techniques can improve the average student's performance by two standard deviations relative to those educated in a traditional classroom one-to-many classroom environment.
Yet the modern educational paradigm, as Massachusetts Institute of Technology Director of Open Learning Sanjay Sarma deconstructs in his book Grasp, remains tethered to Industrial Revolution-era thinking. Our systems still largely operate on a principle of standardization, sorting students through predetermined criteria much like a mechanical "winnower, an industrial machine that separates wheat from chaff. This model, while historically efficient for mass education, has systematically undervalued diverse cognitive approaches and learning styles.
Enter adaptive AI tutors like Khan Academy 's Khanmigo. Their intriguing potential is to provide each student with completely personalized instruction, realizing the gains of Bloom's 2 sigma framework without raising a new army of human educators. It's not for nothing that Sal Khan cites Stephenson's Diamond Age in the opening of his own book on AI, Brave New Words.
However, recent research has also shown the promise of such tools still lives in the realm of the hypothetical. I encourage you to read a well-researched polemic by learning scientist Jared Cooney Horvath, PhD, MEd on Jonathan Haidt 's After Babel substack titled The EdTech Revolution Has Failed for context. More recently, as Dan Meyer has noted, even optimists like Khan have moderated their rosy outlook on the impact of personalized tutors. I'll explore this issue in greater depth in another edition of this letter. For the moment, suffice to say there's so much that AI tutors could do, but the numbers show they currently fall way short of those dreams.
2. Enhancing Teaching Processes
One area that AI could have more immediate impact is in the process of teaching. Great tools like Brisk Teaching and MagicSchool AI provide teachers ways of creating personalized feedback and standards-based curricula in a fraction of the time it would take to do from scratch, reducing the weight of the many, many tasks we (unfairly) ask them to shoulder.
This is the spirit of a nice post by Eric Wang of Turnitin for the Getting Smart blog called Inverting the Teaching Pyramid. There are many thankless administrative and procedural tasks that AI-driven applications could remove from a teacher's to-do list, freeing up more time for the work that both energizes them and does the greatest good for students: live in-person classroom learning.
领英推荐
In my work at Polygence , an academic mentorship platform, we proved the impact of AI-driven feedback tools on instruction. Even before the recent gains of frontier large language models, we ran an experiment on one-on-one teaching exchanges between students and instructors with a natural language processing tool developed at the Stanford University Graduate School of Education .
In a randomized controlled trial, we found statistically significant improvement (+10%) in student engagement when we provided teachers feedback that highlighted instances of "uptake," speech in which they identify, repeat, and build upon student contributions. Developing tools to provide teachers with this hugely effective feedback in real-time could lead to even greater gains.
3. Humans in the Loop
Mixed in with these improvements to teaching interactions are also more ominous possibilities, including the replacement of educators with AI bots. Many edtech companies and developers dismiss this outcome out of hand, but I think it's important to take those concerns seriously; it's only by fully accounting for risks that we can design new tools and platforms thoughtfully with real people at the center of the teaching and learning experience.
Here, I highlight the important work of teachers and thought leaders like Ethan Mollick and his book Co-Intelligence: Living and Working with AI; fellow Fulbright scholar Andy Van Schaack, Ph.D. and his AI-based pedagogy at Vanderbilt University ; and Amanda Bickerstaff of AI for Education . They all demonstrate what it means to keep "humans in the loop" of AI education, rather than simply ceding instruction to LLMs or tools built on them wholesale.
There is reasonable concern about the ability of frontier models to beat humans at the benchmarks we have previously used to measure intelligence. For instance, the 2024 Index of AI performance from Stanford Institute for Human-Centered Artificial Intelligence (HAI) shows these models rapidly closing in on the highest degrees of success on standard tests and reasoning batteries. And even where gaps remain, a new adage reminds us how quickly they will close: the performance of large models improves with every new release.
At the same time, the human mind still demonstrates outlier-caliber creativity. A recent paper by researchers at the University of Lausanne Switzerland--where I did my Fulbright research--bears out this point. Examining 3,000 human responses to a series of creative prompts showed that while AI consistently generates high-quality outputs (with 56.8% falling within the top 10% of rated responses), humans demonstrate superior capability in generating diverse, novel ideas.
This finding reinforces an emerging paradigm of complementarity rather than replacement; AI can help us come up with lots of good ideas, but the best ideas emerge when humans use those tools to stoke their own creativity.
Looking Forward: Critical Questions and Considerations
Again, these are just highlights from the discussion, but they do serve as a bit of reflection on the advances made in 2024 and what we can look forward to in the new year. As we continue to integrate AI into educational settings in 2025, several key questions demand our attention:
Engaging in the Dialogue
Your experiences and insights in this rapidly evolving landscape are important. How is AI reshaping education in your context? What unexpected challenges or opportunities have you encountered? Let's continue this critical conversation in the comments.
This analysis draws from my recent Fulbright Conference presentation and ongoing research in educational technology. For deeper exploration of these themes or to share your perspectives, reach out: [email protected]
#AIinEducation #EdTech #FutureOfLearning #EducationInnovation #TeachingWithAI #EducationalResearch #fulbright