Chat GPT -Homework now pointless?

There will be detection tools for AI, but they may not keep pace with the AI and disguising tech. Currently, GPT will sometimes give you false references, as a friend of mine found when seeing if it could write an essay. She went to check the citations and they didn't exist. That may become a thing of the past though with the rate of development.

There's been quite a lot of legal action around AIs, it will be interesting to see the outcomes.

Of course you do. Or at least you catch every essay that you've caught. 😉

They are just going to get better and harder to catch

It is just another example of how our education system is not fit for purpose.

Why are we teaching our children stuff that a computer can do better and easier?

The jobs market is going to change and rapidly.

As a side matter-

Insisting students Handwriting is a massively retrograde step. Barely anyone Handwrites anything anymore. It's not needed and its not efficient.

It's a damn shame we spend so long insisting on skills that are rapidly becoming defunct. It would be better to teach every child to touchtype....

Absolutely agree. There's software too that even disguises AI-generated text, so I'd imagine that makes it even harder. One thing I have found that works somewhat is not to compare to documented text (Turnitin), but rather to compare to other people's texts. I've actually used ChatGTP to find unique words or similar phrases between multiple (human?) texts and obtain similarity scores. ChatGTP has a style, and even those corrected ChatGTP texts have styles.
In all honesty, I didn't have this luxury when I did my PhD, so there's an element of jealousy here, but I do pity the laziness that this will propagate in the next generation. We're outsourcing creativity.

That's so obvious OP, there's something about the use of adjectives and the kind of afternoon tv-itis of the meaning it gives things. A lack of purpose if you will.

DD did a written exercise recently at school typed in a shared document so you could see the whole classes work, it was so very clear who'd GPTd their work they may as well have signed it AI.

However if you put that text into an AI checker it comes up 0 probability written by a human.

contentatscale.ai/ai-content-detector/

I teach an essay subject for GCSE and a-level and can always tell a mile off if a student has used AI.
Facts are often wrong, but the quickest giveaway is the language style. We have subject specific terminology and the AI tool will invariably use slightly different terminology. A student would have to work really hard to make the work look like their own - nobody has managed it yet.

We also teach students how to get the best out of AI, rather than discouraging the use of it altogether. It can be a useful summarising tool.

Teachers will become more aware of where DC are simply resorting to ChatGPT. Homework topics such as "find out about x subject" will be linked to "to support a class discussion the next day". Or topics will involve more than just fact finding - they will have to discuss various points of view, for example, or be exepected to voice an opinion and justify it. This sort of homework is already common place.

AI can already produce answers to complete all those tasks. In some ways it’s a good thing, because homework is going to have to become more creative if they’re still going to have it, and should become more
meaningful than the usual ‘write about…’ that lots of homework consists of now.

It's very obvious when something has been written by AI. The stuff it produces also isn't great. It could write you a C grade essay which you could then amend a bit to make it less obviously AI but why bother?

However this will definitely be an issue when the AI gets better- schools will have to rethink the sort of thing they set for homework and do more learning at home and essay writing in the classroom.

I’m currently studying a university module where we’re allowed to use it but it needs to be referenced appropriately. I haven’t actually used it as it tends to get everything wrong (maths-heavy subject)!

You can ask it to write you something "in the style of" for example 'write me a paragraph on the Roman Empire in the style of Robert dinero' it's scarily clever!

They might be able to get away with it but after a certain point, not knowing things because they never did the homework is going to have an effect on them alone. It may mean they don't get the grades for the next step and are stuck in a situation where it was all avoidable had they not relied on AI.

I am a university lecturer teaching undergraduate and post graduate students. This is a really big problem. No, you can't tell whether it's human or AI and there is no programme to detect it. The only reason it's obvious is because the essays don't match the lectures and certainly don't match the students' normal contributions.

However, there is very little we can do. Plagiarism is detected by turnitin, which has been around for years and works really accurately and gives us the evidence we need to penalise students. However, there is very little we can do about chatGPT. It is incredibly frustrating, even upsetting. It makes you distrust all submitted work and I imagine the students are far less interested in the lectures or seminars because they don't think they have to fully engage with it.

Our solution is to return to exams and other modes of assessment like presentations. However, this means that students are not getting the experience they need in academic writing and this will have a knock on effect with PhD students who will be lacking in research and writing skills they need to do research and write a thesis/publish generally.

@youngones1 they've already done that. Except for subjects that need physical pieces (ie art) coursework doesn't exist anymore.

They will have to bring back exams fully and stop allowing coursework to determine your results.

Since starting y7 this year not 1 piece of homework with a mark or even a comment!

And most has felt like busywork.

My eldest seems to do all assessments in class which gets around the AI issue but takes up a lot of time.

My son used chat GP to complete his A Level History coursework. The school reported it , the exam board checked it and he was disqualified from the whole A level.

Honestly, I had an A Level student hand in something that was obviously ChatGPT just this week - I'm giving feedback today and we will have to have this conversation. The first give away was that it was typed when I ask for handwritten unless a word processor is used in exams. Strange, repetitive vocabulary that the student wouldn't use, niche knowledge that we haven't learnt in class applied in an odd way, and never engages with the conversation in the way an exam-style response requires. Bonus points for if the same niche knowledge and weird vocabulary pops up in any other student's work.

If the student swears it is not Chat then I won't pursue it strongly, but only they will then know that they have wasted an opportunity. This is also why more timed written work is done in class, reducing the time for learning the material and revision. They don't do it out of badness, usually an anxiety to get something 100% right 100% of the time.

It will just mean a general dumbing down because people won't be able to write things themselves, much like people can't do simple sums without a calculator

It will become much like the calculator is to maths.

It's getting harder and harder to spot imo (I teach EFL in France) so I don't bother marking stuff done at home anymore (meaning I lose more lesson time as my pupils have more tests in class).

The education system is adapting to the use of AI and it will likely become a useful support tool and not the be all and end all.

When I was at school we had to look up everything in reference books - nowadays it's not considered "cheating" that you just use Google to reference multiple sources very quickly.

Teachers will become more aware of where DC are simply resorting to ChatGPT. Homework topics such as "find out about x subject" will be linked to "to support a class discussion the next day". Or topics will involve more than just fact finding - they will have to discuss various points of view, for example, or be exepected to voice an opinion and justify it. This sort of homework is already common place.

using "just" ChatGPT is akin to just taking your infomation from Wikipedia - which would also be frowned upon. Yes, some children will still do it - but they won't be the ones making most progress academically.

At my work (uni) we are leaning into it and setting assignments where they are asked to use it.