Why Preventing AI Plagiarism Leads to Skilled AI Users
To evaluate the quality of an AI tool's output, students need the skills that they develop when unable to use AI.
[image created with Midjourney]
Welcome to AutomatED: the newsletter on how to teach better with tech.
Each week, I share what I have learned — and am learning — about AI and tech in the university classroom. What works, what doesn't, and why.
Let’s discuss why, counterintuitively, AI-immune assignments make students better at using AI tools.
When discussing plagiarism and artificial intelligence with professors, I often hear some say that we should not worry about AI plagiarism because the future of inquiry, business, creativity, and many other domains just is artificial intelligence. AI tools are — or soon will be — our interface with everything we want to know and everything we want to do.
On this line of thinking, the goal of all of us should be to understand how to use AI tools in the most effective and efficient ways. So, in preparing students for their future lives and careers, we, the professors, should be encouraging the skillful use of AI tools to the maximum degree, not stymying their use with worries about ownership, attribution, or authenticity. Our only goal should be to teach students how to best use AI tools to do the tasks characteristic of our fields.
There is something to this line of thought. It is overblown, to be sure, but there is a nugget of truth here.
This week, I discuss the sense in which this view is correct. Interestingly, the net result is a reason for professors to develop assignments that cannot be completed with the help of AI tools — that is, to seek AI-immunity in their course design.
🪨 ⛏️ The Nugget of Truth
It is not certain whether AI is the future for all or many of the important domains of our lives and those of our students. It is not certain that AI will be our interface with everything we want to know and everything we want to do. The same goes for many similar claims predicting the AI revolution to have exceptionally grand scope.
The lack of certainty is owed, in part, to the difficulty of predicting how complex systems like economies will evolve.
It also is owed to the fact that we can be certain that some jobs and activities will be largely unaffected by the AI revolution. For instance, AI will not significantly alter the jobs of the massage therapist, the chef, or the politician; the sport of recreational soccer; or the hobby of knitting. The list of the fundamentally unchanged is quite long.
If we understand claims like “AI is the future” in superlative terms, then it is very likely they are false.
The dynamic with artificial intelligence is parallel to how personal computers changed how significant parts of how many domains are structured, or how search engines and the internet are crucial interfaces for many things. These technologies were game-changers, but few games were made entirely obsolete by them. Further, none of them changed all or even the majority of games significantly.
But we should not get carried away by this negative point about scope. After all, there is no doubt that AI will change how significant parts of many domains are structured; it will be a very important interface for learning and action; and so on.
There are two upshots for the role of university education relative to AI:
We, the professors, should identify those bits of information and knowledge, as well as those skills and methods, that students need to acquire or develop regardless of the role of AI in their future careers and lives; and
We should identify those bits of information and knowledge, as well as skills distinctive of our professions, that students can and should use AI tools to acquire and develop.
Professors are in the business of inculcating what students need regardless of the technological landscape, as well as how to best navigate this landscape.
Since AI is and will be a significant part of the technological landscape, professors must understand it and how to best use it — an understanding which enables them to see where it will be and where it will not be a game-changer.
🛡️ ❓ The Role of AI-Immunity
Last week, I announced that we at AutomatED are hosting a challenge where professors submit take-home writing assignments that are intended to be AI-immune. These are assignments that the submitting professors believe are impossible or tough to crack using AI tools — if a student tried to complete them satisfactorily with AI tools, they would fail.
Below, I will provide an update on the challenge, but first I want to relate this challenge to the preceding.
Those who assert claims like “AI is the future” also tend to downplay the idea that AI plagiarism should be a serious concern of professors.
They argue that we should be focused almost entirely on the second upshot noted above. Our focus as professors should entirely be on how we can best help students use AI tools to acquire knowledge and develop their skills. On this view, there is much less information and knowledge, or skills and methods, that students will need to acquire or develop that does not relate to the capabilities of AI. Worrying about AI plagiarism is barking up the wrong tree, they say. Indeed, we should award those students who effectively use AI to complete their assignments, whatever our views on citation and attribution.
For the reasons already given, I reject this lopsided framing. In short, there will be many domains largely unaffected by the AI revolution.
Yet, even if we grant that I am wrong about that, I still reject the view because AI-immune assignments help students develop the skills and acquire the knowledge that make them better at using AI tools, as counterintuitive as it sounds.
A professor who has pedagogically appropriate assignments that prevent AI plagiarism is one who has isolated what students need to successfully use AI tools. The reason for this is simple: in most cases, one’s successful use of AI tools requires judgment about what makes their outputs good relative to what one seeks.
Take the case of a student who copies all of their classmate’s work for a class. If they copy this work for weeks and weeks, and they do not develop the knowledge or skills essential to the subject, then they are not in any position to know when their classmate has given them bad work. They lack any judgment because they are entirely dependent on their classmate.
Swap out the classmate for an AI tool and you get the same result. By limiting students to learning the essentials of a subject, AI-immune assignments put students in a position to better evaluate products and performances in the domain of that subject, whether they are produced or performed by the student themselves, someone else, or an AI tool.
Next week, my colleague Caleb Ontiveros will explore this subject further when he presents some recent work on how to teach students to use AI well.
🏆 An Update on the AI-Immunity Challenge
So far, we have received many excellent submissions for our AI-immunity challenge. We are very excited to see the creativity of our community in coming up with AI-immune assignments for us to try to crack with the latest AI tools.
We plan to continue to receive and review assignment submissions for the next several weeks, so please check out the challenge parameters and submit an assignment if you believe you have a good one.
We hope to start posting pieces on the results of the challenge by early May.