Academic integrity is a practice, not a policy.
The students are watching.
In February, Stanford University released a study looking at the impact of AI on student cheating rates at six high schools from February to May 2024 (the second year of the ChatGPT era):
“Our findings revealed that overall cheating rates remain stable at 72.06%, consistent with historical baselines and prior studies, suggesting that AI availability has not changed overall cheating prevalence in high school.”
For the second year in a row, these researchers found that the emergence of chatbots had not increased incidents of cheating at high schools. A study of higher education in Australia comes to a similar conclusion that generative AI hasn’t had a noticeable impact on the number of plagiarism cases.
I shared some of this research at a recent speaking engagement, and a teacher came up to me afterwards and said, “That data just doesn’t feel right to me.” In her mind and in her experience, generative AI has increased how much she had to wrestle with her students’ academic integrity.
It’s possible that both the researchers and this teacher (and the many teachers who share her sentiments) are right. AI probably hasn’t dramatically increased cheating rates in schools. Those numbers were already disturbingly high, revealing deeper issues of school design, incentive structures, and motivation.
At the same time, AI has quickly become a part of the infrastructure of school, a resource that we draw on in innumerable ways, including to help us work. It’s asking us to confront core beliefs about academic integrity in new ways. And, it’s creating adversarial dynamics in classrooms that cause stress for both teachers and students.
Status, Respect, and AI
In his book 10 to 25, David Yeager explores the research on intrinsic motivation in young people. Yeager’s research shows that environments designed around high expectations, caring adults, and meaningful work are the most important factor in motivating young people to do things that are hard. The thesis of that book is deeply relevant to approaching AI in school:
“Status and respect are to a young person what food and sleep are to a baby—core needs that, when satisfied, can unlock better motivation and behavior.”
Over the course of this school year, I have spoken about AI with hundreds of students in focus groups or facilitated dialogues with teachers. Status and respect come up repeatedly.
When it comes to AI, students believe their schools see them as cheaters. The AI policies they are asked to follow spark fear and anxiety about punishment. Writing is done in class, under supervision. They are asked to install tools like browser lockdown apps on their devices or to use internet-free computers rather than their own. They are presented with percentages from AI detectors as evidence that they have cheated.
Furthermore, students’ perception is that their use of AI is being monitored and policed while their teachers have free rein to use it themselves. Students often tell me that they have teachers who use AI to 1) give feedback and/or assign grades, 2) create instructional materials with errors or telltale elements of “AI voice” that students can spot, or 3) generate assessments like worksheets, quizzes, or tests that seem disconnected from what students have been asked to study.
Especially egregious to students is when teachers do not disclose their AI use and/or teachers impose severe penalties on student use yet use it frequently themselves. For students, these teachers seem to be holding students to a different standard of integrity than they hold themselves.
Whether you agree or disagree with these students’ assessments, the important issue for me is that students do not feel status and respect on the issue of AI in school. Conditions like these are demotivating and erode trust.
In trying to gain control over cheating, what are we losing?
Defining Academic Integrity as a Practice
When dealing with a technology as ubiquitous as generative AI, the goal should be incentivizing integrity, not controlling cheating. The book The Opposite of Cheating by Tricia Bertram Gallant and David Rettinger uses AI as an opportunity to revisit the research on why students cheat and why AI makes urgent the need to design classrooms around cheating’s opposite, academic integrity.
One of the book’s core recommendations is to clearly define what we mean by “academic integrity” and to engage students in learning about that definition. Decades of research on cheating reveal that “there is no unanimous, shared definition of cheating on which one can rely.” Unsurprisingly, the lack of a shared definition can lead to confusion and conflict.
For me, this issue appears right in the abstract of the Stanford AI study (emphasis is mine):
“Additionally, more students reported using AI chatbots for support tasks like concept explanation and idea generation… [Students] still strongly supported using AI for conceptual understanding and brainstorming, and they maintained clear boundaries against using it for completing entire assignments.”
Many, many teachers I meet would have an issue with classifying those uses of AI as “support tasks” instead of as cheating or at least some form of unauthorized assistance. They see any use of AI as delegating work, and thus a potential violation of academic integrity. Their definition of cheating might be broader than the researchers’.
Students have their own definitions, too. Deeper in the study, the researchers explore the topic of “allowability.” More than half of surveyed students thought use of chatbots should be allowed at least sometimes for idea generation and concept explanation, and 83% said they should never be allowed for completing entire assignments. The researchers identify allowability as a path towards meaningful dialogue on AI and cheating:
“Additionally, if future studies of teachers’ beliefs about what should or should not be acceptable regarding AI reveal differences from students’ beliefs, these findings may inform how to help teachers and students reach agreement on AI policies.”
The key word is “agreement.” Teachers have expertise in their fields and deep investment in academic integrity. Students have a vested interest in their own success as well as increasing fluency with AI. These strengths can be combined to address AI transparently, to reveal gaps in beliefs and understanding, and to close those gaps with shared agreements.
Acknowledge What is Changing
A year ago, when I wrote about AI and the teaching of writing, I shared Sarah Elaine Eaton’s six tenets of postplagiarism. The last 12 months have only reinforced for me what Eaton has been arguing in her work for years: artificial intelligence has introduced a new category of hybrid writing where traditional notions of authorship, and therefore plagiarism, must evolve.
Generative AI has and will change the notion of what it means to “do our own work” in much the same way previous technologies like the internet, the calculator, and Google Translate did. That means AI has and will affect definitions of academic integrity.
Eaton argues, however, that the values of transparency and accountability have always and should always apply. We must acknowledge assistance and we must take responsibility for our work, regardless of what or who helped us.
Similarly, the authors of The Opposite of Cheating remind us of the International Center for Academic Integrity’s six fundamental values of academic integrity: honesty, trust, fairness, respect, responsibility, and courage. In teaching academic integrity, we should insist on these values while also acknowledging AI is changing how these values operate in practice.
In my experience, students not only share these values, they want to live them. They want to do the right thing. What they would like help on is how to put these values into practice when making decisions about schoolwork. “Allowability” is a place where teachers and students can meet and talk.
Designing Environments for Student Agency
If we want to address AI and academic integrity, I think we need to design learning environments built on transparency and accountability that prioritize giving students a sense of status and respect.
1. Ask students to defend their work. David Wiley has proposed a system of random audits, where instead of deploying technological interventions like detectors or browser lockdowns to monitor students and force them to work under supervision, students have the freedom to complete work on their own terms, and the teacher randomly selects students for follow-up conversations where they have to explain their work in order to show that they have learned. While I agree with many of Jon Dron’s critiques of this approach, I do think Wiley’s system gestures towards a more balanced agreement between educators and students that offers autonomy along with clear incentives to act with integrity.
2. Model academic integrity. Bertram Gallant and Rettinger argue that modeling is an important strategy for teaching academic integrity, and they include the below table in their book to ask educators to identify the behaviors they should prioritize in order to model integrity to their students:
Some ideas of commitments teachers might make to students:
Not using tricks like embedding white text in assignments to catch students cheating
Not using AI for feedback on important assignments (read this article on AI feedback if you want to know why)
Not allowing AI detectors to make decisions about the integrity of work
Disclosing use of AI on any materials students are asked to use for class and taking responsibility for the accuracy and quality of those materials
3. Co-construct guidelines with students. I would certainly be doing this if I were still in the classroom. Rather than refusing to acknowledge reality, I would be using my students’ knowledge of and skills with AI to our shared advantage. Using the AI Assessment Scale as a starting point, I would work with students to generate specific examples of allowability relevant to our class for each of the scale’s categories. Then, we would identify the values that undergird those uses, then articulate some shared agreements about how we should use AI, disclose our AI use, and set consequences for uses that fall outside of our guidelines. Frankly, we should be doing this for all forms of external assistance, but AI is a good place to start.
How Do Schools Respond to Challenge?
In their 1999 book The Students Are Watching, Theodore and Nancy Sizer write, “The people in a school construct its values by the way they address its challenges in ordinary and extraordinary times.”
The Sizers use their research to make the case that implicit cues in schools have a far greater impact on student behavior than explicit ones. Students watch adult behavior more than they listen to adult declarations. They find value in the kind of work they are asked to do in classrooms, not in what the teacher says the value is. They are motivated by practices, not policies.
While AI may or may not have increased cheating in schools, it has undoubtedly raised questions about how we navigate school with integrity. And, as the Sizers write, this question can only be addressed with the involvement and consent of students.
“The moral order is voluntary; the adults and the students are partners in its creation and maintenance. Both students and teachers see the point of schooling. In small and voluntary associations, shared norms emerge which make it unnecessary to devise elaborate sets of rules. The rules that last come out of environments, not books. The relationship between the needs of the community and individual freedom is not something arbitrarily imposed; it is, rather, arrived at through explanation, exploration, and persuasion.”
Upcoming Ways to Connect With Me
Speaking, Facilitation, and Consultation
If you want to learn more about my work with schools and nonprofits, take a look at my website and reach out for a conversation. I’d love to hear about what you’re working on.
In-Person Events
June 16-18. I’ll be facilitating a three-day AI program called “Learning and Leading in the Age of AI.” This intensive residential program is designed for school teams to have time and space to design classroom-based and schoolwide AI applications for the next school year. Hosted in partnership with the California Teacher Development Collaborative (CATDC) at the Midland School in Los Olivos, CA, USA.
June 23-26. I’ll be joining the Summer AI Institute at Lakefield College School (Lakefield, Ontario, Canada) as a speaker and coach. This event is for teams of educators to advance their AI work, design classroom and schoolwide AI initiatives, and learn from each other’s work. Just opened to schools from beyond Canada!
Online Workshops
May 4. I’ll be in conversation with Rob Macdonald and Tim Requarth for a free online event, “The Shifting Landscape: What AI Means for the Future of School.” Register via Trey Education.
May 13. I’ll be facilitating “AI and the Question of Assessment,” a design workshop for educators who are looking to redesign assessments to be responsive to generative AI. We’ll explore big questions about learning and motivation as well as practical design strategies that do and don’t use AI to address those questions. Offered in partnership with the Association of Independent Schools in New England (AISNE).
Links!
It’s not just educators who are debating AI. A Notre Dame undergraduate sent an email to the entire student body promoting an AI tool he built that reviews, packages, and redesigns course materials into digestible bites that “lower your cortisol.” One of his fellow students wrote an op-ed arguing that using AI in this way does just the opposite.
Ethan Mollick says that when we try to control or normalize AI, we lose its weirdness, which is its superpower. He offers an alternative strategy: Leadership, Crowd, and Lab.
A very interesting report from the Rithm Project, which surveyed 2,400 young people to find out what their relationship with AI actually looks like and how they perceive its impact on their social lives and their futures.
“The emerging picture isn’t ‘AI is good for learning’ any more than it was ever ‘AI is bad for learning.’” Phillipa Hardman on an important study of AI and cognitive offloading.
Alberto Romero has compiled and summarized as many studies as he can find on AI’s impact on the brain.
Andrew Cantarutti writes that we should stop arguing about tech integration and start building curriculum that prioritizes attention.
I was the first guest on the (amazingingly titled) Canadian podcast, “Eh, I.” I also recommend the second episode, where educator Heather Adams discusses Lakefield College School’s two-lane assessment model.
I had a fun “AI 101” conversation with Alita Guillen on the “10 Seconds to Air” podcast.
Stumbled on your Substack this morning, Eric! Terrific work thinking about integrity in the age of AI. Academic integrity and the still-developing brain. . . it's a tough sell for many students, and adults. But combine the rules that schools and society create with a robust ethics curriculum and it might be possible to cultivate thoughtful conversations about why each of us does what we do. "We" includes all of us. Of course. How many schools offer Ethics as a course? I used to teach a 12th grade course called Ethics in a Multicultural Society. Among other things, the course helped students recognize that there are different morays in every family and each culture. Surfacing these differences goes a long way towards people developing a better understanding of one another. Otherwise, we are as ships passing in the night (hopefully not in the Strait of Hormuz).
Another great essay, Eric, thank you. The emphasis on integrity emerging from the classroom environment also returns some agency back to writing teachers who feel purpose-less in the face of AI. The AI Assessment scale is a good find, too! I've been reading Leon Furze lately but hadn't seen this. Adding the "why" for each decision and talking through it with students adds structure to the class conversation, makes it less overwhelming, especially for newer teachers. I've been preaching to use AI as a site of practitioner based inquiry but for many, that is too much cognitive effort to add when they are still figuring out the classroom environment.