AI in college
This story is going to end with you getting your ass kicked. (In the story, I mean, not in real life. I hope.)
Imagine that, instead of going to college, you enroll in a four-year program at a sports training institute. It doesnât matter what your sport is. Mixed martial arts will be a vivid example, but it can be anythingâtennis, basketball, running, or whatever you like. The purpose of this four-year program is to teach, train, and prepare you for a career competing professionally. Once you finish your program, you willâif everything goes according to planâbe hired by a team or a league. They will pay you and arrange for you to compete.
You see the analogy with going to college and then getting a job.
While you are at this sports training institute, you have to summit records of the training and practice that you do on your own, and sometimes you have to practice, train, and compete in front of trainers and coaches. You, however, thinking that you are clever and because you have an identical twin, donât actually do any training, practicing, or competing during your four years. You make up the records of the training that you were supposed to do on your own, and you send your twin for all of the in-person activities.
Because of your deceit, as the end of your four years approaches, you are the highest ranked mixed martial arts fighterâor whatever your sport isâat the training institute. The top four MMA leagues in the world desperately want you. (And the other leagues know that they have no chance of getting you.) You accept an offer, and your twin, who has no interest in competing professionally, goes back to live with your parents. A week later, you step into the cage for your first professional MMA fight, and you get your ass kicked.
I could add more details to the story to address this or that thought about extending the deceit, but you get the idea. You spent four years supposedly preparing to compete in some sport, but when you finish, youâre completely unprepared to do so. And so it goes with using A.I. tools in college.1
You have to be prepared for what comes next
The moral of the story about you getting your ass kicked is that you have to be prepared when you finish college. To one degree or another, this has always been true, but, ironically, the rise of A.I. is making it even more pressing. If weâre being optimistic, we can hope that A.I. will help us create a wealthier and healthier society, one in which people will work less, have more time to pursue their interests and hobbies, and spend more time with friends and family.2 Before that happens, though, there will have to be a ârealignmentâ to our economy. Because A.I. will do the work more cheaply and efficiently, a lot of jobs are going to disappear and a lot of workers are going to be laid off or never hired in the first place (also).
So, what does this mean for undergraduates today? It means that you need to do everything possible to develop and improve your academic and intellectual skills, especially (although not only) those that are not domain specific: critical thinking and reasoning, problem solving, reading comprehension, clear and grammatical writing (with, if possible, a compelling writing style), and confidence speaking publicly. Such skills will give you the best chance to stay ahead of others in the race for dwindling economic rewards. They will also put you in the best position to adapt to new challenges and job opportunities in a changing economic landscape. And they will distinguish you from what, for the time being, A.I. can do. Maybe in the future, A.I. will be able to mimic some of these skills, but even (or especially) in that future world, you want to be a person who has them, not one who doesnât.
How do you develop these skills? This question deserves a long answer, which Iâm not going to give here. (But if youâre one of my students, you can ask me if you like.) This is the short answer: take challenging courses, especially ones where you have access to the professor and will receive (or can solicit) constructive feedback on your work. AND!!! donât cheat or take shortcuts with A.I. The only way to develop these academic and intellectual skills is to go through the process yourself. You wonât always have success right awayâin fact, you shouldnâtâbut over time youâll improve. Along the way, donât get too discouraged by mistakes and failures. Theyâre part of the process.
Detecting cheating
The final point that I am going to make is that, while it may seem that using A.I. is an undetectable way of cheating, itâs not. Like cheating in general, sometimes students will get away with it but sometimes they wonât.
Itâs well known (by professors and, I think, by students) that ChatGPT and other large language models have a recognizable style. This, perhaps, can be adjusted to fool the reader, but my preferred ways of detecting cheating is different anyway.
One concerns the content of what students submit. Generally, I provide all of the materials that the students in my courses use (i.e., the articles, chapters, books, notes, and videos). I pretty much have all of these materials memorized, and so I have a very good idea about what I can expect my students to know. For instance, if I assign an introductory chapter on the problem of free will and have my students write an essay using it (and only it) as a resource, then I know the range of issues and terminology that (I hope) they understandâdetermined, caused, free, arbitrary, and the like.
Therein lies the catch. When doing your work for you, ChatGPT is liable to slip in terms that, perhaps, make sense but shouldnât be thereâsay, ââFrankfurt-style casesââ or ââcategorical analysisâââand a student who doesnât know what is in the assigned chapter isnât going to realize it. So, when an essay refers to advanced concepts or ideas with which I wouldnât expect a student to be familiar, then the student will have to explain the essay to me.
My second method involves citations. I used to be rather lax about them. I would tell students to provide the page number for any quotes and for any specific ideas (but not direct quotes) from the assigned texts. Most of them could sometimes manage the first and struggled with the second. This wasnât too big of a deal since, as I said, I have the assigned readings memorized. I knew what ideas were coming from where, and I knew what the students were trying to convey.
That said, I think that itâs fair to expect perfectly accurate citations from any student who is beyond the 8th grade. So, now in the age of A.I., I require an in-text citations (i.e., the page number) for everything: every quote, every piece of paraphrased content, and every idea that is not the studentâs own or common knowledge. If the citation is missing, I penalize the submission. Is that harsh? Maybe. Are rigorous citations too much to expect from college students? No. Does it catch and penalize students who try to submit A.I. generated content? Yes.
-
Of course, the sophisticated use of A.I. in business, medicine, science, and other fields is a skill that many people will need and want to develop. My philosophy courses (and many others) are not designed to impart those skills, however, and students wonât develop them by putting âwrite a six paragraph essay on the pros and cons of the categorical imperativeâ into ChatGPT. â©
-
The optimistic scenario is that, after an economic adjustment, A.I. produces significant benefits for society. But this isnât guaranteed. Consider the lesson of globalization. In principle, a globalized economy is superior to one in which every countryâs economy is isolated from all others, but globalization can still inflict harms without enough offsetting benefits. Speaking about globalization, the Nobel Prize winning economist Joseph Stiglitz makes a point that might very well apply to the A.I. driven economy: âThe argument was always that the winners could compensate the losers. But the winners never do.â â©