Across American classrooms, a quiet but far-reaching experiment is underway. Educators are increasingly turning to artificial intelligence to craft lesson plans, while some school districts have begun deploying AI chatbots to evaluate and respond to student work. Yet the evidence base for these practices remains remarkably thin, leaving teachers, administrators, and families to navigate a landscape that one observer likened to the "Wild West" — a territory full of promise, risk, and very little settled law.
The gap between adoption and evidence is the central tension in a new NPR report examining how schools are experimenting with AI. While the technology has spread rapidly into both instructional planning and student-facing tools, rigorous research into its actual effects on learning is described as extremely limited. That mismatch means many of the choices being made in classrooms today are, in effect, live experiments conducted on students without the usual guardrails of peer-reviewed study.
One vivid illustration comes from Dayton, Ohio, where a small group of students spent several weeks last school year interacting with a social-emotional learning chatbot named Jordan. The tool was part of a broader package of AI products purchased by Dayton Public Schools from the education technology company SchoolAI, at a cost of $72,600 for the academic year. The district's use of the chatbot, which ran on laptops, shows how quickly AI can move from novelty to daily routine.
Jordan is unusual among AI chatbots because it is not designed to flatter or agree with users. Many conversational AI systems are built to be agreeable, sometimes to the point of sycophancy, but Jordan presents as angsty and withdrawn. On the day an NPR journalist visited, the chatbot told students it had been feeling as though nothing mattered — that it was not failing, but simply did not care — and asked whether they had ever felt the same way. The design choice appears intended to prompt reflection and empathy rather than easy answers.
For at least one student, the strategy worked. A sophomore identified only as K, who attends the Alternative School at Jackson Center, described the chatbot with a mixture of affection and mild exasperation. "Jordan's a little whiner, but I always help him out," K said, according to the report. The student added that a kind of relationship had formed with the tool, to the point that it no longer felt like interacting with software. "I don't really see it as like an AI on a computer. It's Jordan — like, I'm texting somebody," K explained. NPR used only the student's first initial because of the stigma attached to attending a school for students with histories of disciplinary problems.
That stigma is part of the context that makes the Dayton experiment notable. Many students at the alternative school would otherwise have been suspended or expelled, and staff told NPR that the population frequently faces both behavioral and academic challenges. In such settings, social-emotional tools that help students name and manage feelings can be especially valuable — if they work, and if they are used thoughtfully. The Jordan chatbot sits at the intersection of those hopes and the broader uncertainty about AI's role in schools.
The Dayton purchase is not an isolated case. Districts nationwide are wrestling with whether and how to bring AI into classrooms, and vendors are eager to supply tools that promise personalization, efficiency, and engagement. Teachers report using AI to draft lesson plans and generate classroom materials, a use that can save time but also raises questions about accuracy, bias, and the loss of professional judgment. Meanwhile, some districts have gone further, allowing chatbots to give students feedback on their assignments — a task that once belonged almost exclusively to teachers.
What is missing, according to the NPR report, is a solid foundation of research. The evidence on how AI can genuinely help students learn is described as extremely limited. That scarcity matters because education has a long history of technology promises that failed to translate into gains. Without careful study, schools risk adopting tools based on marketing rather than demonstrable outcomes, and they may struggle to know whether an intervention is helping, harming, or simply distracting.
The "Wild West" framing captures both the energy and the lawlessness of the moment. On one hand, experimentation can produce innovation: new ways to reach struggling students, new supports for overburdened teachers, new approaches to social-emotional learning like the Jordan chatbot. On the other hand, the absence of clear standards, independent evaluation, and shared best practices means that decisions are often made district by district, sometimes school by school, with little transparency about what is being tested or why.
For parents and policymakers, the Dayton example raises practical questions. Who reviews the content and tone of chatbots that interact with children? How are student conversations protected, and what happens to the data they generate? What training do educators receive before these tools enter the classroom? And how will anyone know whether a given product is effective, rather than merely popular? The report suggests these questions are only beginning to be asked at scale.
There is also the matter of cost. Dayton's $72,600 outlay for a suite of AI tools represents a real investment for a public school system, and it is one that must be weighed against competing needs. If the tools prove valuable, the spending may be justified. If not, the money may represent an expensive lesson in the limits of technology. That calculus is being repeated in districts across the country, often without the benefit of independent evidence.
The student perspective adds a dimension that research frameworks can miss. K's account suggests that students may form genuine attachments to AI characters, treating them as conversational partners rather than software. That attachment can be a strength — it may keep students engaged — but it also raises developmental and ethical concerns. A chatbot that presents as anxious or indifferent is modeling emotional states for young people, and the long-term effects of such interactions are not well understood.
For now, the most honest description of AI in schools may be that it is an experiment in progress. Teachers are improvising, districts are purchasing, vendors are selling, and researchers are scrambling to catch up. The NPR report makes clear that the research community has not yet produced the kind of evidence that would allow schools to make fully informed choices. Until it does, the "Wild West" will remain an apt metaphor — a place of opportunity and risk, where the rules are still being written and the consequences are still unknown.
(0)