AI-hunting on the boards

[SoManyTeeth]

I know we're all aware that AI-generated text (and other content) has started appearing all over the place, and understandably people on MN (as elsewhere) are concerned that they're having their time wasted by AI-written posts masquerading as genuine human contributions. But IMO, while AI-hunting is an understandable response, it's becoming a more widespread problem on MN. I feel it might be worth thinking about explicitly including it in the guidelines on troll-hunting in general.

Examples of AI-hunting on this thread: https://www.mumsnet.com/talk/am_i_being_unreasonable/5272283-fleecing-tourists-in-summer-with-high-prices — this is the one that prompted me to post this afternoon, but it's something I see a lot of now.

I understand the urge for posters to demonstrate that they haven't been "taken in", or to warn others against investing in or wasting time on a potentially bad-faith poster. But the same would apply to other types of trolling and bad-faith posting, where posters are advised to simply report rather than engage in troll-hunting. AI posts are presumably an attempt to drive engagement, like much other bad-faith posting. Troll-hunting and AI-hunting can increase engagement with bad-faith posts and threads, disrupt and derail genuine threads, and hurt those falsely accused of posting in bad faith.

There's also a disability aspect that I feel needs to be taken into account here. It's started to be recognised that texts written by autistic people are more likely to be misidentified as AI-generated. Autistic people are being routinely accused and discredited because of natural features of their writing style that they can't help. It's unfortunate that this also plays into old stereotypes and prejudices about autistic people being robotic, computer-like, or otherwise less than fully human, because it means that the false accusations can hurt more when they happen. I'm autistic, and my posts on MN have been dismissed and mocked as AI on MN more than once.

Is this something that could be considered as an amendment to the Talk Guidelines subsection on troll-hunting?

Hey @SoManyTeeth I think it’s me you’re talking about!

You are of-course right that people express themselves in different ways. I’d add that communicating through text strips out nuance and contextualising clues that help with communicating and understanding.

For me it’s not the language of the opening post that’s winding me up. It’s the subject matter or the unusual presentation of particular subjects followed up by no further engagement that’s the tell and that makes me feel like I’m being taken for granted and providing free labour.

I hear you as well about troll hunting being banned but I don’t always agree with MNs unsophisticated application of this. I so often see people posting things about themselves or their kids going straight into the arms of a perv. We are well into the age of online native and the levels of nativity to what’s going on here can sometimes be quite shocking. I’m not surprised that people jump in and say ‘this one isn’t smelling right, have a think about what you’re sharing’

Hi @MurdoMunro, it's absolutely not you personally, or even that thread in particular, that I'm talking about. That thread was just the one I happened to see shortly before writing the post, so it was easy to link as an example to show what I meant by AI-hunting.

It's actually a pretty mild example as these things go — I've had my writing style picked apart in detail and nastily commented on before, as "evidence" that I'm obviously not a human being, and what I've got to say ignored in favour of explaining why my writing style means I'm not real and not worth listening to.

I also have sympathy with the reasons people respond with suspicion to posts they think may be trolls or AI. I know it's annoying to feel you've had your time wasted by a fake poster, or to see less sceptical posters potentially being duped by something when you can see blaring warning signals. And I understand what you're saying about MN's troll-hunting approach having its flaws (it would be better if moderation were more responsive, for a start). But your issue there would apply more broadly to the general troll-hunting policy and its execution, I think?

What I'm trying to say is, given that MN considers troll-hunting a problem, I believe there are good reasons to treat AI-hunting as a type of troll-hunting and manage it in the same way (whatever that way might be). But since the guidelines on troll-hunting don't explicitly mention AI, and from what I see on the boards, people don't think of AI-hunting as something that might be against the rules, it's not currently being treated the same way as other troll-hunting.

Well you certainly sound like a well rounded and articulate human to me so I don’t know what that other poster was going on about.

Rather than ban AI spotting I’d rather than MN spent some of their money on facilitating better. There are patterns here that either they are quite happy with, have not noticed or moderators haven’t been empowered to tackle.

MN earns its money through our contributions, we are both the users and the product here. I would like them to use a bit of that money earned to support the user experience. And to stop other businesses from taking us for a ride.

Aw thanks @MurdoMunro, I've never been so pleased to be identified as human

TBH I think we share similar concerns about MN's moderation, and how their approach (and level of investment) seems to have failed to keep up with the modern internet and their increasing userbase.

If users felt more confident that trolling, AI and other bad-faith content was being adequately prevented/detected/responded to/managed by MN, there'd be fewer people doing what they feel they can to combat it by posting their suspicions on-thread, and thereby less engagement with bad-faith content, fewer false accusations, and less derailing.

Since we don't have that, we have the troll-hunting rules…

Imperfect (and unevenly applied) though those rules are, I think that what they're trying to do applies as much to AI-hunting as to more traditional troll-hunting. It seems inconsistent to me that AI-hunting isn't treated the same way by MN, or at least isn't recognised by posters as being prohibited.

This x 1000000000

I’ve replied to a post asking if it’s AI before. I get that it might just be some people’s posting style or that some people might use AI to help them post.

I’ve also looked at a post, wondered if it’s AI & then discarded the thought because the poster was engaging on the thread.

Surely all someone needs to do if asked is to say they’re human?

I've also asked once thinking about it...I've also seen AI respond to a "how do you feel about x?" post with something along the lines of "as an automated response bot (yes I'm making that up but that was the gist), I don't have any feelings, but here is some info on x..." 😂

It’s obviously an issue that’s going to need some proper thought put to it. Not just here. I am deep into an argument with an energy supplier at the moment and I keep having to find ways to work out if I’m talking to a person or not. The front line people have to stick to such rigid scripts (or haven’t been trained to understand the not-basic questions) that they often come across as robots. Even just in the last year the AI they use has learned to circumvent the tests and I’ve been caught out.

Agree @EmpressaurusKitty at the moment my key test is whether the OP engages with the thread they have created. Likely not AI, might be someone getting us to write their shit content for them, a student looking for us to do their homework or someone hoping to find juicy snippets to copy on to places to take the piss out of women/mumsnet or feed culture wars crap.

Whatever, it’s that create a thread and then sit back that’s had a flare up recently that I’ve noticed.

Hi OP

Thanks for flagging this.We actually already handle these posts in much the same way as troll-hunting, and remove them when they’re reported to us. These posts have a tendency to derail discussions and can be upsetting for those on the receiving end, so we encourage everyone to use the report button.

We do look at posts in context, and so case-by-case may apply on occasion, but accusations of this kind would likely be removed once reported to us.

I guess it's a spectrum, though, isn't it? Whether it's suspected trolls or suspected AI. One or two posters civilly asking a sceptical or clarifying question, or posting a generically-worded caution, is one thing. But that isn't really the sort of thing that would've resulted in the troll-hunting policy.

The thread I linked above is, as I said, a pretty mild one. But it still demonstrates some of the potential problems. Last time I looked, over a quarter of the posts on it were about the possibility that it's AI. If it's AI, each one counts as engagement and bumped it up Active. Maybe the AI-spotters' posts put others off posting, thereby reducing engagement, but there's no way to know. Whether the OP is AI or not, any on-topic conversation that might've happened is harder to have. And if the OP is a real person who spent time writing the post, stuff like this:

I think it's just bad chatGPT.
Yeah me too. It has no point or question.
I agree with this.
These AI threads are awful.
Maybe we should leave the AI and clickbait quacks to talk to each other.
the best solution I heard was that humans should bumbum make weewee their poopoo answers bumbum nonsense fart to bumbum confuse weeeee chatgpt poopoo.

— this many posters chipping in, with a dehumanising and critical tone from some of them — could be quite hurtful.

@LilyMumsnet Thanks for responding.

It looks like most posters wouldn't assume this comes under the troll-hunting guidelines — thanks for clarifying here. Maybe it would be worth making this better-known? I'd guess it's only going to become more of an issue in the future.

It looks like most posters wouldn't assume this comes under the troll-hunting guidelines

Really? It had never occurred to me to treat posts that look like AI as anything other than a form of spamming/trolling and so just report to MNHQ. And so the flip side would be seeing AI hunting posts as equivalent to troll hunting.

Anyway, some amendments to the talk guidelines clarifying re AI generated posts and their reporting would be a good idea.

Yes and if the OP is entirely abstract with no narrative from the poster. Like 'I've been wondering, do we have entirely free will or are there constraints? What does anyone else think?' Lots of those now.

@SoManyTeeth can you share your source for it being recognised that autistic text is being mistaken for AI? Struggling to refine the search terms to find something on Google. Thanks.

If the board is getting spammed with fake AI posts, and seemingly nothing can be done to prevent it, it’s unfair to ban people from discussing whether or not they feel a post is genuine. You are basically suggesting that the level of spamming remains the same, but now no one is allowed to mention it, which is an odd stance.

I think for me it's because AI doesn't always seem to have any sort of malicious agenda when it posts. I've seen a couple of posts I thought were AI and although they were silly, they weren't nasty.

Trolls are different, so I think it's useful to clarify if both need to be treated the same.

Report posts accusing you of being AI and get on with your day. You're not going to be banned as being an AI account, you're not new and obviously human.

I am assuming this is just your personal experience as there are many autistic people on MN and I personally have never seen any accusations of us being AI, rude sometimes but not AI.

I thought AI was used to refine texts, which can be useful for posters with poor linguistic skills or English as a second language. Are you talking about bots maybe?

I like the see all OP posts feature. Anything where the OP posts once and disappears I just assume is bad faith posting.

I do report some AI posts. Yesterday I was just fed up of it and called it out on the thread.

Refining content written by a human should be spelled out as a legitimate use. The sort of posts I've seen which appear to me to be AI generated, some I'd guess are from bots (especially OPs) , some may not be.

I agree @Pickandmixusername . I've come across some posts which are bland and banal, apparently quite harmless but just don't seem like human interaction. One example was a 'poster' who'd put posts with very basic generic gardening advice not really relating to the ongoing discussion on a thread which had loads of really knowledgeable posers on it. I reported, MNHQ 'looked behind the scenes' and agreed it didn't ring true.

The thing about such posts is you have to wonder about why they're being made. It might be some sort of experiment, or it might be to try to establish a username as belonging a legitimate poster with an unobjectionable posting history. I've no idea...hence I report and leave it to MNHQ to decide.

I'm autistic and have been accused a few times of being a bot or journo. Different usernames, of course. Also something along the lines of "Hmm something doesn't seem right about this post/thread" suggesting it's AI or journo writing an article but I never thought to report them because I'm used to my writings (especially when I've started a thread) being seen as too official-sounding or bot-like if it's long and detailed (with lists, bullet points, etc.😆).

I've not had what I'd call a 'nasty accusation' yet so perhaps I'd report if I felt that way or bad about it.