How Do We Maintain Academic Integrity in the ChatGPT Era?

When the artificial intelligence (AI) text generator ChatGPT was publicly released in November 2022, many in higher education had the same reaction: students will use it to cheat. Musings about the end of the college essay abounded on college and university campuses, as did calls for returning to blue book exams. These were natural reactions to the sudden availability of technology that could, in a matter of seconds, respond to essay prompts, assignment instructions, and exam questions. As someone who has specialized in academic integrity for twenty-plus years, I found it refreshing to hear so many people talking about academic integrity. As new technology emerges, we should pause to rethink how we can maintain honesty, respect, responsibility, fairness, and trustworthiness in teaching, learning, and assessment.

The question, of course, is how do we do that in the Generative Artificial Intelligence (GenAI) era? While some educators think the solution is returning to an overreliance on secure assessments (think, two proctored midterms and a proctored final), others suggest completely overhauling how we teach, learn, and assess (and a third group thinks we shouldn’t change anything at all, but a solution that is not). Those in the first camp have a point: we do have a moral obligation to graduate only those students who have demonstrated the specified knowledge and abilities. Those in the second camp also have a point: returning to the ways in which we taught and assessed before the existence of the internet, the contract cheating industry, and GenAI is not the answer. We need to move beyond this false dichotomy of protecting integrity versus teaching better; they rely on each other.

I’ve seen this misunderstanding play out on social media. One college faculty member asks how other instructors are limiting cheating opportunities, and someone responds, “You just need to accept that GenAI is here to stay, and you need to change the way you teach.” This is unhelpful advice. It is not an either/or situation. Rather, many faculty need to improve their teaching and assessment practices and many institutions need to improve their support for faculty to do so.

While the reasons a student may cheat can be complex, four causes are particularly worth noting:

1. Students are extrinsically rather than intrinsically motivated (they are focused on grades rather than learning).
2. They have low self-efficacy (they think that they can’t do the work or can’t do it at the level they want to attain).
3. The work seems meaningless (they don’t understand what they can learn from doing the work).
4. Temptations and opportunities exist (it’s easier to cheat than to do the work).

Instructors can mitigate these causes. While not all measures will work in all class contexts, here are some faculty should consider.

Increase intrinsic motivation. Give students more choice and control over how to demonstrate their learning (assign a research paper or a research presentation). Talk more about learning and less about performance: don’t talk about how to get an A in the class—talk about how to master the learning outcomes. (And if you don’t articulate learning outcomes, start doing so now. ChatGPT can help you with that!)

Enhance self-efficacy. Tell stories of previous students who struggled but succeeded. Give students in-class activities and ungraded opportunities to build their confidence. Break down a task into its many parts, and provide feedback along the way to help students build skills and raise their confidence.

Make courseworkmore meaningful. Articulate the learning outcomes for each assignment and relate how the knowledge or skills students are developing are relevant to their lives. Provide more opportunities for students to share their knowledge with people other than a single grader (with each other or the campus community, for instance).

Reduce cheating temptations and opportunities. For assignments and unproctored tests, make it harder for students to find GenAI tools useful. Ask students to create, analyze, or evaluate, rather than simply remember or understand (see Bloom’s Taxonomy), and require students to orally explain their knowledge. Proctor at least one assessment so you can be more assured that the student has the knowledge being evaluated.

Offer what GenAI can’t. GenAI can lecture, explain, answer questions, and access trillions of bytes of information. GenAI can’t, however, offer human-to-human learning experiences. So, what can you do to increase opportunities for meaningful learning, interaction, connection, and belonging? Can you include peer instruction, peer reviews, or small-group or paired discussions? Can you flip your classroom using problem-based learning, team-based learning, or process-oriented guided inquiry learning, in which students collaborate in teams and you facilitate learning? Whether you teach in STEM, the humanities, or social sciences, think about how you can help students develop human skills like critical thinking, ethical decision-making, interpersonal engagement, emotional intelligence, and oral and written communication.

Get on the GenAI train. You don’t have to fully integrate GenAI into your pedagogy (yet). But you should at least learn about it. Play with ChatGPT using your course learning outcomes, assessments, and content. What is it good and not good at doing in your course? How might it amplify or interfere with student learning? Ask students how they are already using GenAI and how they think they could ethically use the tools to maximize their learning. Then, determine and articulate a GenAI policy for your class, clarifying in what ways it may be allowed.

Institutional leaders and other administrative staff also have a role in ensuring academic integrity in the GenAI era. Here are some actions they can take.

Provide support, funding, and training for faculty to update their courses. Although teacher training is not required for higher education instructors, we evaluate faculty on their teaching. How weird is that? It’s past time that we change this; otherwise we can’t fairly expect faculty to redesign their courses in response to GenAI. You could fund course buy-outs or summer work so faculty have the time to learn, develop, and implement new approaches. You should also provide instructional design training and support, being sure not to punish anyone if a first try at a new pedagogy or assessment doesn’t work. If administrators don’t provide this level of support, we cannot honestly expect teaching and assessment to change.

Create a GenAI on Education task force. Create a task force of faculty, staff, and students that will research how GenAI affects teaching, learning, and assessment, and then develop guidelines for the appropriate implementation of GenAI tools in education. Ultimately, the task force could propose a structure for a team dedicated to providing support for GenAI guidelines implementation, staying abreast of technological developments, establishing shared practices for measuring effectiveness of changes, and advising leadership on GenAI challenges and solutions.

None of the above advice is easy to implement. However, we must do our best to facilitate and assess learning with integrity in this GenAI era. Otherwise, we’ll let down our students, potentially driving them toward the artificial learning experience that GenAI offers. We’ll also have failed to fulfill our promise that our graduates will be more knowledgeable and skilled when they leave our institutions than when they entered. Although GenAI might serve as a useful scapegoat if we fail to deliver on that promise, the blame will rest solely on our reluctance or inability to change to ensure that we are delivering a truly twenty-first-century education.