How can we make AI work for students? Let's ask them.

Here’s something new I learned recently: There are about 4 million software developers in the U.S. There are also about 4 million educators in the U.S. 

Our friends at Playlab hosted an education-focused AI hack-a-thon recently, and they shared this stat during the opening. They then made the crucial point: few of the AI software developers have deep knowledge of education, and virtually none of the educators build software. Grounded by this reality and presented with the hack-a-thon opportunity to build something new, I felt overwhelmed – even debilitated. 

I wanted to create something novel that would responsibly solve some of the biggest challenges we face at CPRL, all while upholding our values. But I spent the majority of the session staring at my computer and barely broke the ice on my AI tool, which I would soon find out was neither novel nor complete. 

Fast forward a couple months, and I got a second at-bat. The Robin Hood Learning + Technology Fund hosted a Playlab hack-a-thon for their grantees. Before heading to the event, I reset my thinking. My goal wasn’t to change the world. I just wanted to build something to see how easy (or hard!) it was, and to better evaluate the possibilities at hand. 

With the guidance of Playlab and fellow facilitators, we picked a problem we face at CPRL – creating icebreakers. By no means earth-shattering, but our team plans 10 ice breakers on any given day, and we vary them, so generating this content actually takes a lot of our time. Worse, we’ve exhausted Google. Every time we search for new ideas, we get the same results. 

The outcome of my second attempt building with AI? A chatbot that produces a mix of popular and creative icebreakers, with citations. To test for a number of common AI issues (like biases, cultural insensitivities), I ran some difficult queries – e.g., does the activity create a space for all voices and perspectives? The suggestions were helpful, and with tweaking I was able to produce better results and even generate decent guidance on facilitating challenging conversations. Even more, I did all this in 25 minutes from start to finish.

I share this story, in part, because it revealed to me the many opportunities we have to ensure that AI educational software is developed much, much closer to students and classrooms. Both the software and, more importantly, the students will be all the better if we do. 

When we consider ways to create public schools that work for all children, our team at CPRL looks at the core design of the education system and asks ourselves:

• Who has the power to make decisions that change student experience? 
• Whose point of view and knowledge is valued and included to shape goals and policy?
• Does the day-to-day work in districts and schools reflect a commitment to and practice of generating broad participation to make progress toward those goals and refine them over the long-term? 

For us to leverage emerging technology and build an education system that enables all children to flourish, we need to shift the role people closest to student experience are playing – moving from passive consumers of new tech and tools to co-producers and builders of resources that support exceptional learning experiences for students, teachers, and leaders across the education sector.

Before my second hack-a-thon, making that shift seemed nearly impossible. Now, it seems less so, particularly when I took advantage of the ideas we at CPRL know work when embarking on innovation in contexts of uncertainty: Experiment in collaboration with others. Work in partnership with a group with the combined expertise to take meaningful steps forward. Start small, test, iterate, improve, scale. Learn from the experiments of others, make knowledge communal, and integrate what works into daily practice, iterating to accelerate toward goals. 

Educators, families, and students must have the same experience. 

We are well-positioned to create these opportunities through our work at CPRL. Over the past several years, we’ve been practicing and developing change leadership that elevates the perspectives of those closest to the problem, aligns necessary policy changes to practices that work for students, and rewires the core operations of school systems to scale learning beyond any one program initiative or policy process. 

AI can help us realize solutions that have been on teachers’, families’, and students’ minds for years – joyful, inquiry-based learning; differentiation and personalization; effective, quick diagnostics; engaging and normed assessments, to name a few – all giving educators more time to devote to the human-to-human work essential in any educational experience.  

These breakthroughs will make way for new challenges, of course, and we’re already experiencing them: student privacy, student isolation, misinformation, biases inherent in software algorithms, hard-to-verify authenticity of student work, digital security, and increased competition from providers deploying ed tech in private, micro, and home school settings. Realizing the breakthroughs and avoiding the breakdowns in the implementation of AI in public schools will require collaboration and broad participation across many constituency groups. 

In our partnerships across the education sector, we are making sense of these changes and asking critical questions about what it means for our own work in public education and that of students, families, educators, ed leaders, and policy makers. 

We will share our learning as we go, and we hope you will do the same. Share your insights with us at [email protected], or post it with the hashtag #leadingthroughlearning. The more we learn together, the better we will do for each and every child.