AI detector.. online course

[purplelucy]

I’m getting quite stressed that I’m not going to pass my course because the tutors are using an AI detector for all our answers and I’ve tested it and even though I’ve 100% used my own writing some of my answers are coming up as AI generated and I’m getting really stressed. I’ve even tried rewording them and it still says it. The website even says it is not 100 % accurate and shouldn’t be relied upon but the course tutors are relying on this as 100 proof that we are using AI to come up with our answers. I don’t know what to do, feel to drop out of the course but I really want this qualification

They should not be using these as anything other than a suggestion that they might need to look at the submission a bit more closely.

I deal with plagiarism at the university I work at. We are not allowed use AI detection software due to the high rate of false positives. There have apparently been legal cases against universities who have relied on the software.

My advice to you would be to keep all your drafts to show how you progressed from start to finish of the assignment. If you are using sources such as journal articles, download and keep a copy or ensure you have the relevant URL recorded.

If you are accused of using AI, challenge it and offer to provide all the evidence that it is your own work. That proof should include you being able to discuss how you went about the assignment and the steps you followed.

I've seen content being marked as AI-generated when it wasn't actually the case. It's a fact these tools are not 100% accurate. I read about AI detection recently; check out the last paragraph here that explains why mistakes of this type happen. And I agree, if needed, one should absolutely challenge it.

Hopefully you've not submitted the answers and are looking for advice...

If your copy has the phrases "stands as testament to" or anything about a "rich tapestry" I'd take that out stat.

You could also try running your answers through Chat GPT and asking it to highlight anything that's going to trigger an AI detector. IRONIC ISN'T IT.

edited to say your post is old! I hope you passed OK

@purplelucy what was the outcome for you in the end? I am just dealing with this at the moment. I wrote answers entirely in my own words but I am being told it is 100% AI written!

And what provider was it if you are happy to share ? Mine was with the Skills Network

Hi, they were called Cidori. It turned out ok in the end. I submitted my work and got my certificate, none of my work was flagged up luckily. Hope it works out for you! It’s frustrating when you write your own work and it says that.

Thank you for the update! I am glad it was ok in the end.

I have emailed and asked for my rejected answers to be reconsidered. I just feel really frustrated and I think it's quite scary that you can have work labelled as 100% AI.when it is totally my own work.

Have a look at my post above. Ask for a meeting to allow you to prove it's your own work and show them drafts, research resources and discuss how you approached it.

They are on very dodgy ground using an AI detection tool as they are notoriously unreliable. You might want to Google court case, university, use of GenAI to get details of which university was taken to court, and lost, for using the tool as sole evidence. Most universities don't use them for that reason.

This article is worth reading - https://www.vktr.com/ai-ethics-law-risk/do-ai-plagiarism-detectors-work-heres-what-the-research-says/

Thank you.
I have some notes I took as I went through the course but these were typed straight into answer boxes online so I didn't do drafts. I think after years of writing for a living these felt easy enough to write straight up so I don't have any rough drafts. But yes I have asked for a reconsideration and I would hope they would look at all my other answers and see that the writing style is the same!

Thank you, this is really interesting and concerning. It's part of the reason I posted, I figure if it has happened to me it must have happened to others.

I didn't use anything other than my notes, taken down as I went through the course, to write the answers.

I suspect it is because after years writing quite formal documents I maybe have a slightly formal writing style, but I don't know.

I can't even imagine cheating because when my trainees try and use AI it always produces total nonsense!

Even the AI plagiarism software makers say they have no legal standing and won't guarantee their products. I would have a word with the tutor up front about your concerns.

It's so utterly frustrating because they just rejected my answers and told me to rewrite them. There was no caveat that "of course our AI checker might be wrong, if so please message....". Just an outright telling off for using AI and some links to mental health helplines/advice about what to do if I feel suicidal .

Surely there should have been a simple option for me to request reconsideration. I have emailed them, but I feel quite cross that the way it was handles implies their checker must be right and I must have cheated.

I'm long in the tooth and can cope with this but it would be absolutely crushing if you were just starting out

AI Detectors are like Lie Detectors they don’t really work, but chuck enough darts in a dartboard, and one or two will hit the right spot.

I work in Digital, and was talking to our Information Security / Cyber security them, and they both said that the detectors don’t work.

If you are genuinely concerned, make a small spelling error, and a grammatical error.

If you are genuinely concerned, make a small spelling error, and a grammatical error.

That's not going to work. We don't use AI detection software but we still manage to identify loads of unauthorised use of AI. That said, I don't doubt we miss loads too when students rewrite it or get the instructions to the AI right.

While writing style is a giveaway, there are loads of other indicators. A lot of the answers tend to be very general and don't address the question in a sufficiently nuanced way

What's particularly frustration though is we are now being discouraged from 100% CA modules and encouraged to include at least 60% in-person exams. We also have to consider how cheatable assignments are, rather than what is the most effective assignment for learning

But what on earth do I do when it was entirely my own work but has been identified as 100% AI? Would you revisit if a student said it definitely was their own work? I don't really know where I go from here, it's put me off ever signing up to study something again.

Believe me, it is just as much of a pisser for the staff

I understand it's tough for the staff too. It's horrible I am sure to read answers and wonder if people are cheating. But I have just had my work rejected with no suggestion they could possibly be wrong about it being written by AI. It's pretty horrific being accused of cheating like that, with no caveat that actually the software can get it wrong, no obvious way for me to challenge the decision.

If you are staff @ColourlessGreenIdeasSleepFuriously can you tell me, would you /the organisation re consider if requested? I mean, I have been offered the option to rewrite and resubmit my answers but I don't see why I should spend another day writing answers I already wrote, just so they can pass a clearly flawed AI detector. And it feels like to agree to rewrite them would be to be complicit in the lie that the detector was correct

I wouldn't have used an AI detection tool in the first place. The procedure where I work is to invite the student to an interview to discuss the submission. If they can show it's their own work, then all is good. If we still believe their is reasonable evidence they cheated, they are penalised but they can appeal that decision.

I would suggest pointing out the inaccuracy of the tools (include links) and asking for an interview so you can prove it's your own work. That would include showing your class notes, describing how you approached the questions and why, and being willing to answer questions on the topic.

Best of luck.

of course. I'd be staggered if your institution didn't especially since the AI people themselves say there's no legal guarantee they are right.

It just seems grim I have to jump through these hoops at all , but I do appreciate your advice.

I managed to get two first class degrees before the internet really existed, there's no way I would be cheating to get through a L2 course!

I was hoping to do further study (I am disabled and wish to redirect my career) but this has completely put me off sadly.

Yes, it is frustrating. Unfortunately though, the level of cheating means that we end up suspicious of all submissions. However, if there is a decent process in place, it is rare for someone to be falsely accused. I've been the plagiarism person for over 15 years and I'd say we've had only a handful of students who we interviewed and then decided there wasn't enough evidence to convict. That's because we examine the evidence rigorously before even contracting the student. The vast, vast majority admit their misconduct when presented with the evidence.

It sounds like the institute you're dealing with doesn't have a proper policy in place.

That makes a lot of sense, I can see how utterly galling it is that there is lots of cheating going on. And as a student it's pretty upsetting that others could get the same qualification by cheating.

From what I can see, my answers were just run through "AI detector" software. I had done the 7 previous units and passed them fine, no suggestion of use of AI. In this unit there were 8 questions, 5 were passed, 2 failed as "50% AI content" and one failed as "100% AI" content. Yet I wrote them all. And surely they would check through first and see that my writing style was consistent throughout? I accept that after years where formal writing was my job I may have a fairly particular style that might be unusual for a student on a L2 course, but it was nonetheless my consistent style throughout. And for me it was the fact that the only option presented to me was to re write my answers that has made me cross. There was no proviso that the software can produce flawed results and this is who you contact if you wish to appeal the decision.