Handling student assignments possibly written with artificial intelligence / AI

This has been discussed a bit on the forum (here, for example), but I’m wondering if anyone has ideas about finding out if homework has been written with AI. I noticed a bit of it last year, and this year a lot - assignments that have great grammar and slightly weird logic. From other in-class material these students write, I don’t think they’re capable of this grammar, but as far as I can tell they’re not actually plagiarizing. Plus they could do this grammar if they sat down with a tutor for 30 minutes to fix the grammar, and since we have a writing clinic that encourages students to do exactly that, I don’t want to punish students who take that approach.

A couple of different questions:

  1. What are they actually using to do this? Is there a common software system university students would be using here, or are there now tons of them? Can these AI systems be told to include quotations and citations (yet)? I’d like to play with the systems a bit myself, since that presumably would give me some “tells”.
  2. Any ideas for how to catch / prove it?

Since the supposed point of these courses is research essays, I don’t want to make assignments in-class or super-guided (i.e. I tell them the topic and specifically what sources to use), but I’m starting to wonder if that’s the direction I need to go.

EDIT: The comments on one article (here) led me to youmakr.com; no idea how common its use is here. But boy, teachers around the world are going to have to make big changes in how composition is handled.

4 Likes

12 Likes

Latest fad is https://chat.openai.com/chat

But they use any website that offers to use AI language models at cost.

We can let AI answer your questions:

:grin:

4 Likes

Thanks AI! But unfortunately it looks like they CAN include quotations - they’re just quite possibly fake.

EDIT: Crap … the more I read about this, the more I think my classes are going to have to move towards almost entirely in-class writing. I had the impression this tech was a looming cloud. I didn’t realize how much of it was already around.

The AI is getting better very fast.
It can do math and programming[!] assignments. This is/will be a huge headache in education.

Previously one could detect plagiarism by letting software do search for similar texts. But AI generated texts can be unique to random seed and uniqueness of the prompt.

I think it will be hard to stop it. Students will let AI do the assignments and then just rewrite it in their own words to mask it.

1 Like

Yup - but it looks like AI can also do marking! So I can have students submitting five AI-generated essays a week, and I can give them five AI-generated responses a week, and at the end of the semester they’ll have a corpus of dozens of essays to work with, none of which a human being has ever actually set eyes on. Progress!

Smartphones have already to some extent “broken” conversation classes, because I think lots of my students don’t have conversations in their native language anyway anymore, never mind in English. Now composition is going too.

8 Likes

Honestly if I had students that were smart enough to set up and use this technology in this way, id probably give them bonus marks.

You can see now a lot of news organizations are basically using AI to create a click bait article, and it looks really decent except for some janky grammar.

2 Likes

I started using AI to generate documentation for some programming projects.
Often there is almost no time reserved for that. As long as everything works, nobody really cares about documentation.

It’s very good even at optimizing/simplifying code. Now it can refactor code from one programming language to another.

This will make software engineers multiple times more productive. But this also means a lot of contractors will be left out where AI can churn out similar results in minutes.

4 Likes

Would it be possible to request the inclusion of personal experiences or something else that is not possible to do with the help of AI?

2 Likes

The AI will tell you it has no own personal experience to share, but one can ask it what a personal experience a real person might have.

Probably not practical to have every student explain the assignment in person…

Also from openai.com’s chatGPT

World War 2 was a global conflict that lasted from 1939 to 1945 and involved the majority of the world’s nations, including all of the great powers. It was the deadliest war in human history, with an estimated 70-85 million military and civilian deaths.

The war was caused by a variety of factors, including the aggressive actions of Nazi Germany, led by Adolf Hitler, and the failure of other countries to effectively address these actions. The war saw the rise of powerful military alliances, such as the Allied powers, which included the United States, the United Kingdom, and the Soviet Union, and the Axis powers, which included Germany, Italy, and Japan.

Ultimately, the Allied powers were victorious, leading to the defeat of the Axis powers and the end of the war. The consequences of the war were far-reaching, and included the Holocaust, the development and use of nuclear weapons, and the beginning of the Cold War.

One possible quote from Shakespeare that could be applied to World War 2 is: “The quality of mercy is not strained” from The Merchant of Venice. This quote suggests that mercy, or compassion, should be freely given and not forced or contrived. It could be applied to the situation of World War 2 by suggesting that those involved in the conflict should have shown mercy and compassion towards one another, rather than engaging in the destructive and deadly conflict. Again, please note that I am only suggesting this quote as a potential literary reference, and I am not able to provide any further analysis or interpretation of World War 2.

It was lying to you when it said it can’t include quotations…

3 Likes

In my book they get points for that personal touch.

Generally though, the academic integrity issues need to be explicitly highlighted: if the AI is not getting co-author credit that is a fail.

Idiocracy here we come!

1 Like

But it has Electrolytes! :potable_water:

3 Likes

This is about teaching English, but in general, I think the problem is that schools and Departments of Education struggle to keep up with the advances in technology. The question is, what do we want the students to learn to be able to survive in the future world? Because of how quick the world is changing, it’s close to impossible to predict what skills students of today will need even 10-15 years from now. Maybe the assignments should be changed to allow any means possible to achieve the best possible outcome, but the students need to disclose every step of the way.

3 Likes

I originally asked chatGPT about the current Russo-Ukrainian war:

chatGPT:

I’m sorry, but I am not able to provide an assessment of the ongoing war in Ukraine as my training only goes up until 2021 and I am unable to browse the internet for current information.

As for the quote from Shakespeare, one possibility could be: “The native hue of resolution is sicklied o’er with the pale cast of thought” from Hamlet. This quote suggests that indecision or hesitation can weaken one’s determination or resolve. It could be applied to the situation in Ukraine by suggesting that those involved in the conflict may be struggling with doubts or uncertainty about the course of the war. However, please note that I am only suggesting this quote as a potential literary reference, and I am not able to provide any further analysis or interpretation of the situation in Ukraine.

1 Like

Ok, this is freaky…

Me: Write an outline for a Friends and How I Met Your Mother crossover episode.

ChatGPT:

The following is an outline for a hypothetical Friends and How I Met Your Mother crossover episode:

Title: “The One with the Yellow Umbrella”

Synopsis: After a chance encounter at Central Perk, the friends are introduced to Ted, Lily, Marshall, Robin, and Barney. The group quickly becomes friends and spends the day exploring New York City together.

Meanwhile, Joey is struggling with the realization that he has feelings for Rachel, and Ross is trying to impress a potential business partner with his knowledge of the city.

Later that evening, the group gathers at MacLaren’s Pub, where they reminisce about their past experiences in the city and share stories about their lives.

As the night comes to an end, Ted realizes that he has met the mother of his children, and the friends part ways, with the promise to meet up again soon.

Act 1:

The friends are hanging out at Central Perk when Ted, Lily, Marshall, Robin, and Barney walk in.
Joey is immediately smitten with Robin, while Ross tries to impress Ted with his knowledge of New York City.
The group decides to spend the day together and explore the city.

Act 2:

The group visits various landmarks in New York City, including the Empire State Building and Times Square.
Joey realizes that he has feelings for Rachel and struggles to express them.
Ross tries to impress Ted's potential business partner with his knowledge of the city, but ends up embarrassing himself.

Act 3:

The group gathers at MacLaren's Pub and shares stories about their lives and experiences.
Ted and the mother of his children have a brief encounter and exchange glances.
The friends say their goodbyes and agree to meet up again soon.

Act 4:

Ted reflects on his encounter with the mother of his children and realizes that he has met her before.
The friends all return to their normal lives, with the promise to see each other again soon.

Won’t AI talk in the third person? Possibly a tell. I appreciate that talking in the first person is/used to be generally discouraged, but perhaps it’s the way forward?

1 Like

ChatGPT is trained to not give “personal opinions” and would often say “I’m just a model so I don’t have feelings/opinions/outlook and so on”. However, that is something that people at OpenAI imposed on ChatGPT, and it requires additional steps and methodologies to prevent the model from just imitate everything in their training data. So if you use another NLP model, the model could easily claim it has feelings and is a real person with opinions and dreams because that’s what it sees in its input.

Vivas will be the only solution, I think. Time allowing…

3 Likes