AI: A New Opportunity for Academic Dishonesty?

Author: Dr. Sarah Learman, Teaching and Learning Consultant, Office of Curriculum and Instructional Support

Artificial Intelligence (AI) in Academia?

AI text-generating technologies, a “new” method for students to dishonestly complete coursework, has raised concern among higher education faculty and staff nationwide, including those here at CMU (Marche, 2023; Nguyen, 2023a; Nguyen, 2022b; Rogers, 2022; Glazier, 2022; McMurtrie, 2022). In this blog, you’ll find a pragmatic discussion of the “AI in academia” controversy. We’ll also share suggestions and resources to help you navigate this most recent instructional detour.

In brief, in Fall 2022 reports surfaced nationwide, including here at CMU, of students utilizing AI software (such as ChatGPT or Jasper) to generate text responses to assignment prompts: essentially, to write answers, including full essays, for them, which is a clear form of plagiarism (Rogers, 2022; Glazier, 2022; McMurtrie, 2022). Rogers (2022) summarized AI technology history and functionality:

“…the gist [of how AI works] is something like this: the programmers [of AI software] made their robot read approximately 10% of the Internet, which was more than enough to give it a good idea of how people typically express themselves in a very wide variety of circumstances. Also, I think, the programmers taught the robot how to use Google and Wikipedia. So with a surprisingly small amount of input from a user, [AI software] can produce paragraphs that look a lot like something a human being might write.”

Unfortunately, current technology solutions to completely identify or mitigate AI-generated plagiarism have not quite caught up to the sophisticated text-generating software. This includes the CMU-supported technology SafeAssign software, which is not designed to (and thus does not effectively) recognize and flag AI-generated text. (Note: a possible method to utilize SafeAssign to catch AI-generated text is provided below.) However, some tools and strategies to help instructors navigate this dilemma are already being developed. Designed to check work that one might suspect as artificially created, new programs such as this Open Access AI Detector and GPTZero (which was designed by an undergraduate at Princeton University) are available. These technologies are already demonstrating effectiveness (Brown, 2022a; Brown, 2023b). Before using these or similar resources, it is important to note that they are not CMU-endorsed or supported tools, nor have they been vetted for functionality or accuracy by CMU staff.

Mitigating AI-Text-Based Cheating: Reframing our Approach

Knowing that such AI technologies are commonly used for academic dishonesty, consider how you might deter plagiarism in light of these new technologies. Further, how might you turn the tables and perhaps even use AI text generators to your instructional advantage? Review the suggestions and read the articles referenced below to learn more about deterring such AI plagiarism (and other types of cheating) and how to get started working with AI software instead of against it.

Use your tools and do your research to recognize the “enemy”

As Rogers (2022) suggests, we’re already familiar with the elements of this technology. Consider the software you encounter and use daily, all those auto listening, response generating features of our various devices and programs (e.g., text predictors, Amazon Alexa, dictation-to-text, autocorrect, autocomplete, Google, and more). Recognizing how often we encounter these technologies and understanding how they work relative to and within AI software can make the issue less intimidating.
Again, while the available recognition technology is certainly behind in this area, solutions are imminent. As mentioned, Open Access AI Detector and GPTZero represent two examples of recently discovered methods to detect AI text in written work (Brown, 2022a; Brown, 2023b).
Learn as much as you can. Go into the information wilderness of the Internet and see what else you could learn about AI-text software and how it could be either a speed bump or a smooth experience within your course. To get started, links to some of the most recent articles and resources on this topic are provided at the end of this post.
Try AI out yourself. Grab one of your current writing prompts and see what the various AI sources produce. This will give you an idea of what the AI text looks like for your course materials and help you to recognize such content in your students’ work.
SafeAssign workaround. While not foolproof, preparing SafeAssign to recognize AI-generated text from your assignments may be possible. If you use the AI software to generate a written response to a course prompt, you can upload the output into SafeAssign. The AI text will then be in the SafeAssign database and therefore more likely to be flagged as plagiarism if submitted by a student in the future.
Add Automated Feedback to your writing assignment. Use the FeedbackFruit’s Automated Feedback tool for students to submit drafts of their paper to be reviewed against the criteria you establish. Students will become more educated on the writing process and gain assurance their paper meets the standards you’ve stipulated. You will also be able to see how their writing progresses, which should help make any AI-generated text stand out.

Turn these AI lemons into… a teaching and learning garnish

Rather than viewing these applications as a negative, imagine the advantages and the possibilities that such technology might bring to our classes.

Use a pre-prepared sample of AI-Generated text in your class; show the students how the AI outputs are noticeably different from human prose. With some strategic examples, AI software can be a useful tool to help students learn what plagiarism is and how to avoid it.
Early in the semester, have students “write” an assignment using AI, demonstrating how the text is constructed. Then, ask students to review and modify the AI version as if it were their own. As their writing and revising skills improve over the semester (through other learning activities), repeat the AI exercise to demonstrate their progress and personal improvements.
Also applicable at any time during the semester, students could write their own response to an assignment prompt and compare that work to an AI-generated version of the same assignment. Providing an opportunity for students to compare the two versions and identify differences or similarities (and, in turn, your class increases your own understanding of AI-generated text!).

Assignment Structure, Expectations, Rationale, and Communication

How might you prevent using such AI technologies before they start?

Don’t ignore the existence of this software. Talk with the students about the issue openly and honestly. Tell them why they shouldn’t use this technology and get the buy-in needed for their commitment to honesty. Provide students with clear expectations (what is permitted and what is not) and the consequences when those expectations are not met (explain the disciplinary actions that will be taken if the policy is violated), and communicate this information to students often.
Share the rationale behind the writing assignment. A valuable strategy to increase academic integrity is communicating “the why” with your students. Ensuring students understand why the question matters and what they will get from it if they do the work themselves (e.g., their learning gains) (Fox, 2011). The literature unambiguously states that students are more likely to cheat when they don’t understand the purpose of an assessment or if they perceive it in a more transactional way.

“…if learners believe that the task is relevant or important for their future goals or generally useful for them, then they are more likely to be engaged in the task as well as to choose to engage in the task in the future.” (Pintrich & Zuscho, 2007, as cited by Calkins & Seidler, 2011)

Structure your assignments during the semester to ensure students have enough time to complete the work. Spacing out deadlines can decrease the pressure that students feel about the writing and thus make them less compelled to plagiarize. Consider also separating larger assignments into intermediate deadlines, where students must revise their work. Students will be less likely to use the AI generator, knowing they will have to go back, revise, and use their writing again in a larger assignment. Similarly, ensure that the assignment’s weight is not disproportionate. Students are more likely to compromise their academic integrity for assessments they perceive as higher stakes. If the stakes are perceived as “too high” (i.e., a final grade consisting of little more than a test and a paper), students are more likely to cut corners, leaning on auto-generated text to get their assignments done.
Depending on time, resources, and general sustainability, instructors may consider modifying assignments to deter the use of such AI technology or incorporating other academic integrity supporting tools. In a recent OPED piece CIS’ own Associate Director of Learning Systems Support, Marnie Roestel, spoke directly to this question:

‍“Incorporating a variety of strategies into your assessments to deter cheating is probably the best method and setup…Rely on not just one tool but rely on several things: creating tasks that are open-ended questions or essays is probably a great strategy; randomizing questions and randomizing answers so that no two students are getting presented with the same set of questions or in the same order…Consider maybe changing the activity entirely where students work together, collaborate, learn from each other. Discuss a particular question, arrive at the answer together.” (Nguyen, 2022b)

Make it personal

Incorporate reflective questions where AI couldn’t appropriately generate a quality answer, making it easier to identify. For example, ask students to communicate how they got to an answer using their own words or prompt them with some creative, personal application of the information. Such a prompt might read: “In addition to showing your work and solution to this problem, please include two sentences describing your thought process while solving it (why you did which steps)”, or, “If you were explaining this topic to your 10-year-old cousin, what would you say?” The intention is to add a prompt that requires the student to provide an individualized approach or response to the scenario. It is much easier to tell if it is your student answering or if it is some “one” else.
Build relationships with your students. Human experiences inform our perspective, priorities, and, consequently, our motivation for doing (or not doing) certain things, including academic shortcuts such as cheating. Taking the time to talk with and get to know your students can help build honesty and trust and curtail dishonest behavior before they start (Glazier, 2022).

Help is here!

The Office of Curriculum and Instructional Support is always here to help with any step in this process. If you’re unsure where to go from here, if you’d like help approaching your course, assignment, or writing prompt related to this new AI technology, or if you have any other instructional questions or needs, send us an email at TLPro@cmich.edu.
Upcoming Workshop from the CMU Writing Center
Revising Writing Assignments for the Age of AI
Hosted by Troy Hicks and Dan Lawson

The College of Liberal Arts and Social Sciences and the CMU Writing Center invite participants to join them on Thursday, February 2nd from 3:30 to 5:00 p.m. for a hyflex workshop (in person at CMU or online via WebEx) on revising writing assignments to better facilitate authentic learning goals. Please bring an assignment sheet for a current writing assignment. We will use AI writing applications to consider how best to revise those assignments and adapt our instruction for this changing context. Register here.

Additional Resources

This list contains recently published articles and links related to AI-Text Generators in academia.

References

Brown, E. (2022a). A new AI chatbot might do your homework for you. But it’s still not an A+ student. NPR. https://www.npr.org/2022/12/19/1143912956/chatgpt-ai-chatbot-homework-academia

Brown, E. (2023b). A college student created an app that can tell whether AI wrote an essay. NPR. https://www.npr.org/2023/01/09/1147549845/gptzero-ai-chatgpt-edward-tian-plagiarism

Calkins, S. & Seidler, A. (2011). Faculty perceptions of relevance in teaching and learning. International Journal of Teaching and Learning in Higher Education, 23(2).

Fox, J. (2011). Why are we doing this? Establishing relevance to enhance student learning. Faculty Focus. https://www.facultyfocus.com/articles/effective-teaching-strategies/why-are-we-doing-this-establishing-relevance-to-enhance-student-learning/

Glazier, R. (2022). How to solve the student disengagement crisis. The Chronicle of Higher Education. https://www.chronicle.com/article/how-to-solve-the-student-disengagement-crisis

March, S. (2023). The college essay is dead. The Atlantic. https://www.theatlantic.com/technology/archive/2022/12/chatgpt-ai-writing-college-student-essays/672371/

McMurtrie, B. (2022). AI and the future of undergraduate writing. The Chronicle of Higher Education. https://www.chronicle.com/article/ai-and-the-future-of-undergraduate-writing

Nguyen, N. (2023a). ChatGPT: How to adapt your courses for AI? Feedback Fruits. https://feedbackfruits.com/blog/chatgpt-how-to-adapt-your-courses-for-ai

Nguyen, N. (2022b). ChatGPT: A threat to Education? Feedback Fruits. https://feedbackfruits.com/blog/chatgpt-a-threat-to-education-opinion

Rogers, J. (2022). Writing with robots. The Habit Weekly. https://mailchi.mp/thehabit.co/writing-with-robots?e=7810232178

Office of Curriculum and Instructional Support Blog