Since mid-December, news about ChatGPT, the newest AI text-generating tool, has captured the attention of those of us in higher education, both causing worry about the implications for academic integrity and intriguing us with new options for engaging students in discussions of the impacts of AI in their academic and professional careers. This post is meant to provide a brief overview of the CITL’s responses to this new tool; for a more in-depth examination, along with more detailed recommendations, see our page, How to Productively Address AI-Generated Text in Your Classroom.
So, what is our take on ChatGPT and other AI-text generation tools?
Find a balanced and learning-first response.
As with all issues related to academic integrity, we encourage instructors to take approaches that focus on student learning. Relying only on a policing approach establishes an adversarial relationship that makes teaching and learning less enjoyable and rewarding for everyone. And, ironically, policies laced with threatening language can break down your relationships with students, which can make cheating psychologically easier for them. So, put your energy into the things you love and are best at—in this case, teaching in creative and innovative ways.
Think about why students may turn to ChatGPT in the first place.
Simply seeing students who turn to ChatGPT as lazy or dishonest avoids some opportunities for learning-first approaches. Could course workload or assignment timing be an issue? Consider offering some staggered deadlines or breaking the assignment into smaller deliverables rather than one big chunk; this can help avoid the procrastinate-and-panic problem while also givings students good feedback along the way. Could students have time management issues? Turn to the Student Academic Center to help them learn to manage their work better. Could they feel a lack of academic writing skills will hurt their grade? Work with Writing Tutorial Services to create a support structure for their writing process. We don’t mean to discount student responsibility for academic integrity, but considering why students may turn to ChatGPT can help you find learning-first approaches that can help.
Beware the promises of AI detectors.
We are already hearing about online tools that claim to detect AI-generated writing, but early tests of those systems show them to be inconsistently successful, certainly not enough to rely on as the sole approach to AI-related course policies or accusations of academic misconduct. As we note in our web resource, we plugged three paragraphs of a ChatGPT-generated essay into the GPT-2 Output Detector and received a “99.92% real” report, casting more doubt on that detection tool’s accuracy. And tools like TurnItIn currently rely on comparing a submitted text to a database of other texts, so it is not very useful when the AI tools generate new content on the fly. So, this is likely going to end up being a cat-and-mouse game, with detectors always chasing the generators, and we are not likely to “catch” our way out of this situation.
Consider redesigning your assignments.
ChatGPT is very good at what it does, and it can be very creative, but it has its limitations. We go into more depth on our website, but here are a few ideas about how assignments can mitigate the impact of AI while delivering good learning opportunities for students:
- Make assignments reliant on data and contexts the AI cannot have access to—in-class activities and group work, recent publications (ChatGPT’s databases currently stop at September 2021), unique cases and scenarios, and images that must be analyzed. (Be ready with options for students who have to miss class or have visual impairments.)
- Have students demonstrate their knowledge in alternate ways—oral or poster presentations, videos, visual elements, podcasts, etc. ChatGPT could still be used to generate a script, but alternate ways of showing knowledge are more likely to demand significant student input.
- Weave in-class writing into the process. This allows you to give early feedback to students while also giving you a baseline of their writing styles.
And consider running your assignment through ChatGPT yourself to see how it does. If it can provide an acceptable response, keep working on your prompt. Articles, email lists, and blogs in own discipline are also good sources for understanding the tool’s strengths and shortcomings within your context.
Embrace the reality of ChatGPT
The genie is out of the bottle—we cannot change the reality of AI tools like ChatGPT being a part of our academic and professional worlds from here on. So, consider how you can teach students how the tools can be useful and what their shortcomings are, particularly within your disciplinary context. Could you have your students analyze ChatGPT outputs to see how what it gets wrong and how they might improve its work? Could you use it to generate thesis statements or introductions they have to expand on with class-based data? Could you talk to students about AI’s potential value and risks in their professional careers? Teaching and learning happens within real-world contexts, so find ways of addressing this new part of that context.
The sudden appearance of ChatGPT has come as a shock to many of us, and it certainly will disrupt the way we teach. But we try to see disruption as an inflection point, an opportunity to question our current practices and push us to advance our work. As always, we are here to help you with your own challenges and innovations, including developing assignments that engage students in new ways… whether those are intended to avoid AI or embrace it. Let’s keep working on this together.
For more information and assignment ideas, please see our web resource: How to Productively Address AI-Generated Text in Your Classroom
And consider joining us for a webinar series we are co-sponsoring with UITS Learning Technologies, the first session coming up this week: ChatGPT and AI in Teaching and Learning: Opportunities and Challenges (1/18/2023, 3:00-4:00 pm)