社区黑料

Get stories like this delivered straight to your inbox. Sign up for 社区黑料 Newsletter

Today’s students are on a high-speed trajectory toward an 鈥渋nnovative鈥� future 鈥� one in which artificial intelligence has equal potential to enhance or undermine their learning.

Teachers are rightly concerned that AI cheats and shortcuts will erode students鈥� independent thinking and that increased screen time will the social skills and human connection kids need more than ever in a technology-powered world.

As New York City superintendents, one in the Bronx and one in Brooklyn, we decided to lean into this moment and try to develop AI-powered teaching assistants that increase student thinking, foster human connection and complement effective teaching practice.

As a guide for sorting through the many AI product pitches in our inboxes, we focused on NYC鈥檚 big goal of increasing reading achievement and decided to concentrate on improving our core English Language Arts classes. We didn鈥檛 want another supplemental solution 鈥� an extra intervention when core instruction fails to meet the needs of diverse learners. Instead, we wanted more students to receive the support and feedback they need during class, so fewer of them require additional help. 

Since the New York City Public Schools had already done a lot of to improve phonics instruction and foundational reading skills in the , we decided to focus on middle school, where rigor increases along with students’ struggles. We met with principals who wanted to be early adopters share our goals and an early demo. Eleven schools in the Bronx district and three in Brooklyn signed up.

We did not want the entire class to become tech-powered; rather, we targeted the AI toward the most challenging parts of the lessons, when students were doing close reading and writing. Teachers assign each student to a small group, and they all open their Chromebooks and log into , which takes the texts and questions from the curriculum, makes them interactive and provides more targeted support for students who need it. 

Students first collaborate with their partners, discussing their initial thinking about each question. Then, they type or speak their response into the AI. The technology confirms what the students understand through instant feedback and then pushes them to go deeper, often directing them back to a specific portion of text and asking a follow-up question that guides them from literal comprehension to inferences and author鈥檚 craft. As one student said, 鈥淚t鈥檚 like the handout is talking to me.鈥� 

While all this is going on, teachers review a live dashboard that shows every student鈥檚 level of understanding of every question. If the teachers see students are struggling, they can provide immediate assistance to get them back on track.

After about 15 minutes of students working together with each other and the AI, the teachers push a button and the AI synthesizes the two biggest misconceptions in the class in real time, suggesting a discussion question to address each one (this was a 鈥渨ow鈥� moment for our teachers!). The teachers then lead a targeted class discussion, often with a lot more student participation than usual because the kids feel more confident after working with the AI and their partner.

Finally, all students complete an exit ticket, often a short written paragraph about the final question of the lesson. They again receive up to three rounds of real-time feedback on their work and revise their writing after each round. 

Based on 2025 New York State test results, classrooms that used these tools at least twice a week for the year doubled their rate of growth compared with the rest of their district. In in the Bronx, for example, those students saw growth of between 14 and 16 percentage points over the previous year, compared with a 7-point improvement overall.

While we are still learning, we hope the knowledge we gained will help other educators actively shape this next generation of AI-powered tools. Here’s some of what we learned.

First, it was important to ensure that our AI tools worked seamlessly with the high-quality instructional materials (HQIM) we had already adopted. As Heather Peske from the has highlighted, AI tools that instantly allow teachers to create lesson plans, change assessments or dial down the level of challenge risk undermining the quality and consistent learning progression on which HQIM curricula are built. 

Second, it was important to increase student collaboration, both in small groups and during full-class discussions. Most early AI products follow the old paradigm: Students put on headsets, look at a screen,and work silently on their own. No one knows the full complement of skills that young people will need in their AI-powered futures, but will be even more critical than it is today. 

Third, the biggest decisions we made were pedagogical, not technical. We wanted the AI not just to support students or save teachers time, but to help our educators be more effective. Our teachers helped design the 鈥渕isconceptions spotlight鈥� tool so they could see and address the biggest areas of student struggle. They also asked for a 鈥渉ighlight鈥� tool so they could celebrate strong student thinking and call out exemplary work for discussion when the learning is still fresh and relevant.

Fourth, the North Star of any improvement effort must be student outcomes. Based on the 2024 NAEP results, reading achievement nationwide is at its lowest level in 30 years. In adopting any AI tool, school and district leaders must clearly define their goals at the beginning of any partnership, and then rigorously evaluate the impact. The is leading a movement to better align incentives and ensure contracts are tied to clear measures of student impact.

The decisions school leaders make today will shape tomorrow鈥檚 outcomes. When educators both embrace the transformative power of AI and hold tight to the values and knowledge of effective instruction, every school can build the future all students deserve.

Did you use this article in your work?

We鈥檇 love to hear how 社区黑料鈥檚 reporting is helping educators, researchers, and policymakers.