AI and Student Cheating

As this website evolves I suspect most of what it will contain will be discussions of the benefits of AI. More and more people will begin to share what has worked for them and their students and describe how helpful AI has become to them in their teaching and learning. For this post, however, the focus is on what some in the academy see as a real threat by AI to the educational process—student cheating. While student cheating is nothing new, the advent of generative AI has made it easier and more widespread than ever before. Student self-report surveys describe the pandemic as the point where student cheating in high school began to increase dramatically, encouraged by the opportunities for cheating allowed by remote learning. With the introduction of ChatGPT in November of 2022, similar surveys indicate that 70% to 80% of high school students and 50% to 70% of college undergraduates have cheated on assignments. There is even some evidence that cheating really begins in middle school. These figures may not be as dramatic as they seem, since many students reported cheating only on rare occasions. Nevertheless it is clear that student cheating is rampant and that the new AI technologies are now the main enabler.

At Colleges and Universities, where academic integrity is one of the highest values, many faculty members, staff and administrators are deeply concerned. After all, it is our job to develop in students the knowledge and skills necessary for them to be successful in work and in life. If they use AI to do their learning, thinking, problem solving, writing and creative work, they will not be adequately educated and not be prepared for such success. As a result, many faculty members want nothing to do with AI and want things to return to a pre-AI “normal”. Unfortunately, the AI genie is out of the bottle and everywhere present in in higher education and the wider society, especially in the world of work. Instead of pretending it doesn’t exist we must find ways to use it that reduces student cheating and even enhances it use in the process of education.

Why Do Student Cheat?

If we want to reduce student cheating a good place to start is to ask why students cheat in the first place. Several answers to this question have been proposed.

• One answer is that students do not see using ChatGPT and similar AI systems as wrong. However, the vast majority of students do think it is wrong. They think it is wrong, for example, to practice plagiarism, to turn in work as their own that has been produced by others. But if they think it is wrong why do they continue to cheat?
• Some claim that students, even the best students, cheat because of the pressure from society for success and the avoidance of failure. Success today is measured in getting good grades, and not so much in how much the student learns. Student cheating in order to get good grades is even reinforced by the wider society. Parents, schools, social media, politics and business often look favorably upon those who are successful in life, even if they have used cheating, lying and ruthless behavior to get where they are.
• Students rationalize that others cheat and they will be at a disadvantage if they do not. So honesty is rationalized away for the sake of achievement. Also, the negative consequences of getting caught and punished are not much of a motivator not to cheat. Students usually don’t get caught. There are no reliable technologies for detecting plagiarism, for example, and probably will not be any time soon. So it is worth it for them to cheat.
• Some students work long hours outside of school or find their academic work too challenging. So, without the time or the resources to succeed they cheat in order to get by.
• Sometimes students think that the course material is not interesting or not important. It may be an elective and not as important as a course in their major, so they cheat instead of working hard in order to pass the course.

The conclusion is that many students justify cheating as an acceptable means to an end—educational success. Unfortunately, this utilitarian attitude that the end justifies the means is often carried forward by students to life and work, where it is commonplace today.

Ways to Reduce Student Cheating

One way to change behavior is through negative reinforcement. If students get caught and punished appropriately, cheating would subside. Faculty members should keep an eye out for student cheating, but doing so excessively has its own problems. We know that any attempt to construct software that is good at detecting AI plagiarism has been a failure. Too often using what is currently on the market results in false positives. The student didn’t cheat but the software said she did. This makes for damaging adversarial relationships between faculty members and their students. Instructors can play “cop” in other ways as well, such as requiring students to submit a writing sample at the beginning of the course and comparing it to future submissions. The trouble with this otherwise reasonable strategy is that ChatGPT can be prompted to write on various levels of sophistication, thereby undermining the effort to catch student cheating by comparing the level on which they usually write with a ChatGPT submission.

A more promising direction is to use various types of positive reinforcement to convince students that it is not in their best interest to cheat. There are many examples of how to do this that are easily found with a quick search of the Internet. Some of the recurring suggestions include:

• Explaining what AI systems are in general and what they can and cannot do. When used appropriately they can be very helpful tools; when used inappropriately they are often inaccurate, biased and discriminatory.
• Being very clear what the rules are in your class for using AI is also very important. A general discussion and explanation of your AI expectations at the beginning of the course, as well as a specific statement on your syllabus about when AI can be used and how to cite its use, is very helpful both as a guide to students and an additional form of positive reinforcement.
• Being very clear about learning goals is also an important element in educating students about the point of the course and how AI can help or hinder the achievement of these goals. Some goals may be enhanced by AI while others may be best achieved with no AI help. If one of my learning goals in an introduction to philosophy class, for example, is for students to be able to construct and analyze arguments, using AI to construct arguments that students then evaluate may be a very helpful way for students to practice this skill
• Constructing methods of assessment that test the level of student achievement in ways that minimize cheating possibilities is also an important way to mitigate against cheating. For example, in-class written quizzes, exams and brief writing assignments, as well as classroom presentations where the desired knowledge and skills are demonstrated, are some ways to accomplish this.
• Constructing assignments that are resistant to cheating is another important way to reduce the amount of student cheating. A large project, such as writing a research paper, for example, can be divided into smaller units—brainstorming for a topic, clarifying basic concepts, gathering information, an outline, a first draft, various levels of revision—each of which items can be allowed to be aided by AI in specific ways or not. Other assignments may focus less on gathering information and more on students’ analyzing and evaluating the information.
• Explaining why plagiarism and other forms of cheating are harmful to the student is yet another way to reduce the incentive to cheat. Explaining that writing, for example, is not just putting one word after another, as generative AI systems do. Writing is the way we create and express our thinking, our problem solving skills, our imaginations, our emotions, and even ourselves. Just as an athlete who doesn’t practice much will not excel at her sport, if students do not develop the knowledge and skills that are produced only by the hard work demanded by acquiring a good education, they will not be successful in work and in life. While it may not seem so in the moment, the long term consequences of cheating are ultimately harmful to the student.
• One of the best ways to mitigate the urge to cheat is by showing students how the appropriate use of AI can be in their interests. One of the most helpful uses of AI, for example, is as a personal tutor. This is an AI system that can somewhat play the role of an instructor. New technologies of this type have been developed for this purpose, with more certainly to follow. One such AI system is based on ChatGPT-4 but more restricted, so it does not so much provide answers to assignments as to suggest improvements of the student’s own work--much as an instructor might do during office hours. This type of feedback does not substitute for student work, but does lead to its critical assessment and improvement. Having a personal tutor also creates more equal opportunities among students who are less prepared for college success than some of their peers.
•  

Conclusion

As educators we have the challenging task of teaching our students how to use this powerful new technology appropriately and beneficially. In the end, as has always been the case, it is up to the student whether or not he or she adheres to the principles of honesty and academic integrity. In the meantime we can take steps to encourage them to do so, both for their own sake and for the sake of the wider society that they will soon enter.