To teach in the time of ChatGPT is to know pain

LLM use is the most demoralizing problem I’ve faced as a college instructor.

See full article...

 

[wxfisch]

I recall for my undergrad that most of my core classes had very little actual course work, most of our grade was tests. I was in person, so tests were all written or multiple choice which would make an LLM somewhat unhelpful. If we did have an assignment, it was usually document review or synthesis of information as a longer essay which I suppose an LLM is helpful for, but as has pointed out would make things obvious when you are getting As on assignment and Ds on tests.

For my online grad degree though it was asynchronous just like is being talked about in the article. We had a reading assignment each week (usually read chapters a-n then complete either specific exercises from those chapters or answer a writing prompt), and a discussion requirement (post a response to the topic by Weds at 5pm, respond to at least 2 other students posts by the end of the week). The exercises I am sure I could get an LLM to do, but the discussion would be harder. You could likely get an initial topic easily enough, but the responses would be nearly as much work to get an LLM to create as just responding yourself.

I am curious though how group projects fare in the LLM age though. It was always an issue of one person doing more work than others, but how does that change when some students phone it in with LLMs and others actually put in the effort?

 

[wxfisch]

I’ve long been skeptical of long-distance learning.

Primarily I feel that way as: an occasional student, because my motivation flags when I have to spend more time in front of screens. (Though I do recognize that, for some students, long-distance learning is the only possibility.)

But also as a past, and potential future instructor, I’m skeptical because of the cheating plague that has existed long before the rise of AI. LLMs, and other forms of AI are just ‘weaponizing’ that phenomenon, alas…

At this point the only way that I’d agree to teach would be: in-person only; and a student’s grade would rely solely (or at least largely) on their proctored work (e.g., in-class quizzes, exams, etc. ). However I’m not sure if typical schools and administrators would allow that approach…?

This is the same (flawed) thinking that upper managers apply to return-to-work edicts when some people do genuinly work better remote. "It doesn't work for me so no one could benefit from it".

Just because you learn better in person does not make that true for everyone. Some students do better with self-paced courses or asynchronous courses for a variety of reasons. Perhaps they can better focus on the topics they don't understand and breeze past ones that they find intutitve? Maybe it allows them to more easily pull in additional resources to add context or explain a topic in a different way? Or perhaps they are easily distracted in group settings or have social anxity that makes larger in-person classes less effective or more stressful for them? Maybe they just learn better through reading than lectures? There are a million reasons someone may prefer remote learning to in-person that have nothing to do with access to education.

At the end of the day the education is the responsibility of the student. No teacher can force a student to learn, the best they can do is create an environment that fosters learning for their students and be clear and up-front that they students need to put in the effort, especially in post-secondary settings (and even more so in graduate settings). If a student really sees their education as a check box to a future job, no amount of effort from a teacher is really going to solve that.

 

[wxfisch]

Your argument has holes...so we started using skidsteers in factories to increase efficiency, as a result those who wanted had time to go to the gym and workout as a result of the weight yhey gained from using a skidsteer in the factory....even better before the loom everyone had to know how to make shirts, due to the loom we had more time to do other things. Your argument of using a skid steer to lift weights at the gym is backwards, as we used it to make work easier....instead lets look at school lessons. Now we could make a student go to a library and regurgitate a book onto paper....or we can change the goal...instead of writing a paper using the library and pretending the internet doesn't exist, or not using ai....Instead of:

“Write this essay yourself”

It becomes:

“Use AI to generate 5 perspectives, then critique them”

“Find flaws in this AI-generated argument”

“Design a system using AI and explain the tradeoffs”

That’s way harder cognitively than doing grunt work.....just my opinion

I think this is missing the point of education though, especially post-secondary education. The goal is to learn not just the process to get to end result, but also to learn core concepts about that subject (espcially in science and math courses). I won't say that LLMs are unhelpful in day-to-day work in the real world, they certainly can be, and we should teach taht skill as well. But the whole point of the article is that teachers don't have any good tools to combat how easy LLMs have made short-cutting the work required to learn something unrelated to LLMs, and lazy students (or overworked students) are using them to their own detriment either because they don't understand why they need to put in the work, or because they don't care.

The forklift in the gym example is, IMO, perfect here because the entire point of the gym is to do the work, for it to be hard. Making that easier goes exactly against what the end objective is (to be more fit, stronger, look better, etc). The objective was never to pick things up and put them down again, just like the objective in school is not to write essays, it's to learn a topic (and often learn how to learn new things). The exception to that is maybe trade schools where the goal is to learn how to accomplish specific tasks, but even there you still need to understand the why behind a lot of the skills, not just the mechanics of how to do them.

TFA points out that all of your examples can (and are), today, being responded to very convincingly by LLM tools. That is the problem that is being highlighted here.