Brain not needed. AI like ChatGPT sneaking into our jobs and lives

[steve ridgway]

My wife has recently become very keen on the Grok AI on X Twitter to speed up her in depth enquiries into health and pharmaceutical topics. It will find relevant research material much faster than she can although is not completely reliable. Only today she asked it for a link to unconvincing research findings it gave her and it had to admit it couldn't actually find them. She has also challenged and argued with it to the point it has admitted being wrong and has changed its opinion. When one of her followers told her Grok had rated her 65% on a particular topic she took it on and debated it up to 95% in the end. So it can be very useful but could also be quite dangerous to believe everything it says.

[krummholz]

@krummholz and @DavidW So, and this is going back a ways, don't students still have to list sources and references (bottom of page, etc.)?

K

For term papers/projects and essays, yes. For homework problems... well, there aren't SUPPOSED to be any external sources, right?

[DavidW]

That's not to say that I don't assign homework - I do, in fact I have been experimenting with a colleague's idea to give them challenging problems with the additional information that the weekly quiz - or next week's exam - will have a very similar problem on it. To get a good grade they have to at least learn how to solve problems very much like THAT one. The problem with the method was that many students just didn't care enough to do the work, even when their grade was on the line.

I don't have a good answer, and none of my colleagues do either.

That is what one of my colleagues did before he retired. He would have a weekly quiz that was just one of the homework problems. If the students solved it themselves and understood the solution, then it is really no problem to do it again in class.

I would say that it works, but the students that we teach have a higher degree of motivation than what you might be used to. It is surprisingly rare for me to encounter a straight up apathetic student.

[VonStupp]

What about page citations?  Or is that all listed in AI?  Or is that mostly no longer relevant due to online articles?

K

That is an excellent question. It has been long enough since I have done any formal research that I still have APA, MLA, and Chicago format guides sitting around the house. 

Unfortunately, technology moves much faster than education (and other entities) can keep up with. On the other hand, there are plenty of concerned educators trying to keep up with what is available for youths as these things appear.

VS

[Kalevala]

For term papers/projects and essays, yes. For homework problems... well, there aren't SUPPOSED to be any external sources, right?

Yup! 

I remember having to learn the various rules and typing the references (including page number, etc.).

K

[AnotherSpin]

On routine homework, no. Some instructors will require it if completed by hand, but it is commonplace (and has been for the past 25 years) for instructors to employ online homework services. Physics homework is about solving problems yourself, not doing research. Using Chegg to find a solution and posting a reference to it would be audacious cheating and not fine.

On papers-- yes... but AI can generate references as well. Plus, the common student cheat pre-AI is to patch-plagiarize Wikipedia and then use the Wiki sources as the paper's sources. I am not a humanities instructor, but I do require writing. It used to be formal reports, but I've switched to more informal writing called lab memos this year. Anyway, the common place thing that students will cheat on a formal report is in the introduction. And it is a stupid cheat. It doesn't take much more work to selectively quote your sources and then interpret the quotes yourself.

And if a student were to plagiarize or use AI, what stops them from also lying about their references? A good teacher would personally check those references. But I have a small number of students as compared to normal high school teachers and college and university professors. I still would need to check the sources of 40-60 papers. That is why casual cheating is so effective. Instructors can't afford to scrutinize every single scrap for potential cheating. Just take me with my light load. I assign homework three times a week. Five problems per set. For forty students, you do the math. And I am assigning it on paper, mostly to check that students show their work, not to try to detect cheating.

In modern times, how do teachers do it? First of all AI AI-written anything is so formal, stiff, and uses uncommon language that almost anyone can see when someone used AI to write a paper. You don't even need hallucinations. But what is done in general to catch cheating? Well, online resources are used.

On the front of papers, there is a service called Turnitin that checks for duplication of any other paper in their database. That tests for students copying from themselves, easy to find resources online, and even paying for a paper that someone used at a different school. The difficulty in technical writing is the frequent and long use of known, repeated phrases. For instance, it is not copying for multiple papers to use the phrase "the law of conservation of mechanical energy." One can use filters on the length of the phrase, but it is best if the instructor spends time going through the matches. Turnitin also provides AI checker, which surprisingly is rare at returning false positives.

On the front of online homework submissions, I find that the key thing to look for is the time to complete the assignment. Webassign has become pretty smart about only tracking active use and not just when the tab is open. A typical A or B student will take 1-2 hours to complete one of my homework assignments. If a C student completes in 5-10 minutes, they are most likely cheating in some way. Especially if that level of speed and accuracy is not reflected at all in exams. This unnatural pacing is also one of the ways that the College Board is having proctors observe potential cheating on standardized exams like the PSAT. But most homework cheating is invisible.

Many years ago — and I do mean many; it was in the last century, after all — during my time as a university lecturer in history, I made it a point to allow students to use absolutely any sources they wished during exams. I would often ask them what grade they want — and would duly award it. More often than not, they assessed themselves quite accurately, and I saw no harm in that.

Of course, history has its own particular demands, but to my mind, it was far more important not to stress people unnecessarily. After all, I wasn't testing their ability to memorise names and dates, but rather something altogether different — their capacity for independent thought, their ability to draw their own conclusions, and so forth. An experienced teacher can spot such qualities almost at once.

As for ChatGPT, it strikes me as a rather fascinating tool, and I see no reason why its use shouldn't be permitted freely. The real challenge, it seems to me, lies in giving proper thought to the purpose and substance of the assignments we set.