To be naffed off a robot is deciding if my work is ok

[StarDolphins]

I’m doing an online mental health course. My writing and terminology has obviously improved as I’ve progressed through the units.

Yet again, my submitted assessment has failed due to it being flagged as AI. I 100% did not use AI and to me it doesn’t even sound like it. So now I seemingly have to find different ways to say everything to get through zerogpt. I was proud of it and I can’t think of any way to change it and don’t think I should have to.

AIBU to be totally narked and feel like giving up?🙄

One of the major issues with AI currently is that it actually has real problems recognising AI generated writing and as you have discovered issues a lot of false positives. I suspect that many who know how to use punctuation properly would soon fall foul of its algorithms...

It’s ridiculous. The detectors are not remotely accurate. I’ve put articles I wrote 20 years ago in, and they’ve been flagged as 100 percent AI, and AI-generated text that I changed a single sentence in and that’s come up 100 percent human.

Just put some of my own writing through the "humanizer". It has indeed turned it into something chatGPT would not produce, but at the cost of making into it significantly worse prose, as if rewritten by a 10-year-old.

Sorry (lecturer here) but your problem is not with AI detection, but with how your course provider is managing assessments! It does sound like your tutor is being quite passive about this, and if she is claiming her hands are tied due to external verification then you need to escalate it to the programme lead or course provider.

Its an absolute waste of your time to keep messing about with ChatGDP and various checkers and adds no value to your learning.

I increasingly have to deal with AI generated essays and it’s maddening - I bend over backwards to support students who I can see full well are actually putting the effort in.

I had the same thing and I. Was. Furious! 😅 Doing an online course with written elements and some of the answers are run through an AI detector by the tutor during marking. When the feedback was that some of the content was AI generated I just sent a very clear response that I find AI abhorrent, refuse to use it due to the environmental impact, and all responses were my own so they may want to consider a different tool. Unfortunately it's something the tutor 'has' to do but after that point there were no more flags for AI use....

Isn't a hyphen incorrect though? They're different things. A hyphen joins words like compound adjectives (off-road, cross-country) while a dash is used when there is a new clause in a sentence, doing a similar job to a colon or a conjunction. The use of a dash (em dash in typography) rather than a hyphen is correct, it's just also something that AI tends to overuse.

This is absolutely true! I am far from intelligent and I have to really check and change things. For someone with good writing skills, this will be a nightmare.

This will be my second resubmission as the last one came up ‘highly likely AI’ and my redone one still 64%.

I wonder if a more "old fashioned" word processor will help. I use Libreoffice, because it is totally free, safe, and it's sort of more old fashioned I suppose. More manual, which I like. And it free of course, which matters for me.

Are you sure you aren’t AI OP, maybe you aren’t self aware yet 🤷‍♀️

I think we are at the point where text files need an EXIF style data embedded part, same as digital photos.

But of course, one can get EXIF editors. So maybe not.

I think for sure though there needs to be some identifier with AI output that machines can detect. And really, AI companies should be pushing for it themselves, because of AI degradation.

AI degradation is the theory that is AI uses AI to "learn", then it is will start to incorporate all the AI "hallucinations" and errors into itself. The theory is gaining traction.

A bit like inbreeding I think, for want of a better example. Where people with the same DNA issues have kids, and the DNA issues match up and things get worse.

Most universities no longer allow the use of AI detection software as it is so unreliable. It is also biased against certain types of writing. There have been successful court cases taken against universities in other countries where students were wrongly accussed.

I deal with academic misconduct in my university. Firstly, I would say do not change your writing style. Do not put it through GenAI in order to dumb it down. What you need to do is keep evidence that it is your own work. So keep your notes. Any academic sources you have used should be downloaded as you are using them so you can point to it on your laptop. Alternatively, keep the history on your browser so you can show you read it on a certain date. Use Word on OneDrive or similar where you will get version control so you can show the different versions of your assignment as you worked through it.

I would also strongly suggest contacting the student union and asking them what they are doing to ensure the university is not falsely accusing students by relying on the AI detection software as a sole method of assessing use of GenAI. Tbh, I would say it should not be used at all.

Edited to say - I would also ask the tutor what evidence they have to suggest use of GenAi and remind them that the detectors are notoriously inaccurate.

Honestly the obvious answer is a return to hand-written in person invigilated exams.

I was thinking similar, but along the lines of old style word processors. I can see a market developing for this "retro" stuff.

My handwriting is terrible.

If you keep submitting the same document, or versions of the same document, it will be flagged for plagiarism (as well as AI). AI detection software is notoriously unreliable and my university does not use it. I think most places are heading back to in-person exams, in a traditional examination hall.

That’s such a pain. I ran soemthing through an AI checker recently as I had used AI but didn’t want it to sound like I had, it passed almost the whole thing and the only thing it picked up as AI was one of the few things I had written entirely myself?!

A bit worrying these will be used in educational settings.

That shouldn't happen if it's set up to accept multiple submissions.

@DontBuyANewMumCashmere You’re probably fine as it’s likely changing to an n-dash, rather than an em-dash (unless you have a particular setting or are possibly using a US version of Word). The em-dash is a standard American dash and has become a ‘tell’ (using inverted commas here because it’s not actually a tell) simply because many generative AI programs use US English.

For reference, there are three standard types of dash:
Hyphen -
En-dash –
Em-dash —

Examples:
by-product (hyphen)
3–4 (en-dash)
the cat – the black one – was ill (en-dash; British English usage)
the cat — the black one — was ill / the cat—the black one—was ill (em-dash; US English usage)

Why Is My Own Writing Detected as AI? 9 Fixes That Actually Help

The link above has some useful information.

Incidentally, I do a little work for a company that uses AI to generate feedback to send to learners. We include, in the learners' feedback, a warning about using AI. :(

Yes, I don't *want to put a hyphen in, I just type the hyphen as it's easier to find on my keyboard and then it autocorrects to m dash

I was just pointing out that I'm not choosing the m dash (although I do want it!) and I didn't want people to think I am a bot!

I realise this isn't the only thing, just... It's correct use of punctuation isn't it.

Is it possible in these instances to write a note at the end along the lines of I confirm this is 100% human created (myself), no AI involvement. Or does it just get automatically spewed out and rejected by (ironically) AI and not a human.

Ai detector is shit. Because everyone has slightly different writing styles. It leaves no room to be individual because if you don’t follow the “human” script it must be Ai.

Too blunt. AI. Too wishy washy. Al. Over excited. Ai. Too robot like trying to avoid the above. Ai.

Thing is, I use emdashes and semicolons all the time in my writing. And I have done so for years, long before ChatGPT and its AI littermates were even a twinkle in Silicon Valley's eye.

AI is a nightmare for schools/universities and this is just further proof! You have my sympathy - what a waste of your time.

I have no idea whether this is the solution, but your writing in the screenshot you shared doesn’t sound very academic. As a PP has said, you use very short sentences and perhaps that’s reading as an AI summary. You use commas and complex sentences in your comments on here - can you try to do that more in your submitted work?

e.g. Ease of access to services is also important as early intervention and support can prevent problems from escalating.

I'm an academic and this is great advice. You could also show your browser history to show you have been accessing eg academic articles and have not been accessing gen ai sites.

I'm a bit concerned that the academic misconduct appeals process at your provider doesn't seem very robust. You should be offered a chance to put a response to these allegations. Do you have a students union you can involve? If not, I'd follow the academic complaints procedure and potentially involve the Office for Students.

We also do not use AI checkers as they are terrible. But I teach an essay based subject and the question of how we teach and assess extended writing in an age of AI is one we haven't solved yet.

I submitted an academic application this week which contained a few emdashes in and quite possibly a semicolon or two. This thread has me worried!