States urge schools to use AI, but leave teachers on their own to stop cheating

This article was originally published by Chalkbeat.

Paige Wyatt heard plenty from administrators about using AI to generate lesson plans at the school where she taught high school English.

Her students used AI to cheat so frequently she resorted to spending her own money on detection tools. Yet she never got help with what she really wanted: how to effectively catch and stop the cheaters, some of whom were very clearly just copying and pasting from Google’s AI summaries.

Teachers like Wyatt consistently identify cheating as a major concern in the AI era, and 70% of teens report using AI for schoolwork, a Common Sense Media survey published last month found. But that concern is getting little recognition from states. While at least 13 states have passed laws about AI in education, the laws most often direct state agencies and districts to create their own AI guidance, meaning lawmakers have frequently absolved themselves from making tough decisions.

Meanwhile, roughly three out of four states have published information to help guide AI use in education. But a Chalkbeat review found these documents tend to skirt the issue of cheating, offering little in the way of concrete steps to halt or deter students from offloading work to AI. Instead, they dedicate more real estate to AI literacy and ways teachers can use AI for efficiency.

All of that beating around the bush has left educators and districts to their own devices to come up with ways to deter and catch AI-enabled cheaters. It’s become an exhausting endeavor that hasn’t turned up any definitive solutions, educators and researchers said.

There’s no surefire way to detect AI-generated content. Nor is there an agreed-upon definition for cheating: Can students use AI to outline an essay but not write it? Can they use it to copy edit? Some states have crafted acceptable use policies which make some, but not all, of these distinctions.

Wyatt quit her teaching job last year. She didn’t leave solely due to AI, but its impact on her work was a factor.

“We really need real consequences,” she said. “We need workshops for teachers to detect cheating. We need workshops for kids.”

Most of the nation’s 10 largest school districts are developing or have developed more specific policies to try to prevent using generative AI to cheat. But it hasn’t always been easy: New York City delayed the rollout of its AI guidance for months amid fierce backlash, before announcing a new policy this week. Half of those districts explicitly discourage using AI detection tools.

Many of the official responses might underscore the urgency of the problem without helping to address it, which could only increase teachers’ anxiety.

“I think the situation teachers are in feels bad, and feels like they ought to have a solution now that they don’t have,” said Justin Reich, a professor at the Massachusetts Institute of Technology and the host of a podcast about how AI is shaping K-12 education.

Letting AI ‘Run Amok’ Could Ruin Student Learning

Historically, state policymakers haven’t rushed to create laws governing new technology and education.

States didn’t really start to adopt computer science requirements until the 2010s. And while the iPhone launched a smartphone revolution in 2007, the first state restrictions on cell phone use in schools didn’t come until 16 years later, in 2023.

States may be moving faster when it comes to mainstream generative AI, which took off with the advent of ChatGPT in late 2022.

“Folks can’t just hope this will pass. This is a major change, and students are already using it,” said John King, former U.S. education secretary and current chancellor of the State University of New York system. “This isn’t tomorrow’s problem. This is today’s problem.”

The District of Columbia and at least 37 states have issued recommendations for public schools about AI. Their documents range from short webpages to Illinois’ 400-page behemoth. Many define different types of AI, stress that any AI use should still be led by humans, and link to outside resources — some of which have been developed with the support of big AI companies.

But they often devote just a page or a few paragraphs to academic integrity.The documents also seem light on actual guidance and don’t include many citations, Reich said. That’s because there’s little evidence to show how effective AI tools are in the classroom, or what policies would effectively stop cheating.

“If you just let this run amok in secondary schools, all kinds of learning activities will stop,” he said. “We do not know what to do. There’s no evidence-based policy. There’s no basis for development and professional development. Like we’re just making stuff up.”

More than a third of states urge schools not to rely solely on tools that promise to detect the use of AI, citing studies showing they are not effective and incorrectly flag genuine student writing as AI.

But many states then offer little else other than acceptable use policies. In California’s case, the state offers a list of questions for districts to mull when making policies around integrity in addition to an acceptable use policy.

In the absence of hard and fast rules, teachers have reported requiring students to complete more graded work in class, using pencil and paper.

Christopher Harris, the director of the school library system for the Genesee Valley BOCES, a regional district in western New York, wants educators to go further than that and reconsider how they measure learning. Writing may be less of a valid tool to assess if students are learning, now that AI “can suddenly write anything.”

He said students could submit an oral assignment, or teachers could focus on the learning process more than the product by asking students to explain how they respond to feedback when developing a project or presentation.

New York is one of the 13 states that hasn’t published AI resources for educators. Harris’ library system has been developing resources about AI for the state’s educators using grant money.

“I really wanted to help teachers understand they can’t control what kids do with AI,” he said. “The only control they have is their agency in developing assignments.”

Adopting AI Policies Not Straightforward For School Districts

Sara Brescia thought she had a no-brainer policy proposal for her fellow Manassas City Public Schools board members: adding AI misuse to the district’s code of student conduct.

The policy she proposed for the district in northern Virginia — a hotbed for the data centers AI relies on — would prohibit students from trying to pass off AI-generated content as their own. It specified that AI could only be used with a teacher’s permission, and required students to cite AI use. Students would be punished, up to and including with expulsion, for violating any of those provisions.

But other board members voted her idea down in late July. Lisa Stevens, another school board member who is also a teacher according to her LinkedIn page, worried the policy would add a burden onto teachers without giving them any tools to identify AI-generated work.

Brescia is skeptical about the value of those tools.

“We do not have a comprehensive approach,’’ she said. “We have the student code of conduct that says nothing.”

Wyatt said when she left the classroom, students were only becoming more fluent in methods to cheat. They were finding tools like QuillBot to disguise AI-generated writing and make it sound more like student writing.

She’s open to going back to the classroom, if she gets more administrative support to help detect the kind of AI misconduct that felt out of control when she left.

“As a community, we need a support network,” she said.