AI bots infiltrating the forum - reply or just click on 'report'?

[DevonDamo]

[Note: this thread probably won't make any sense to you because it's just a few posts split off from another thread in the 'bikes and bits' forum - see link in moderator note below. Scroll down to the 6th post below for my attempt to summarise the issue.]

[Moderator note - split off from viewtopic.php?t=161829 and moved to Using the forum.]

I actually logged in to enter this thread with one of my 'eventually-going-to-get-me-banned' contributions. In this case I was planning to accuse one of the recent respondents of being an AI bot, purely based on their track record of posting tedious, bland, get-it-from-Google replies. But someone's beaten me to it: that reply has been deleted (and hopefully the user has been banned.) Well done to whoever else spotted it.

[slowster]

I actually logged in to enter this thread with one of my 'eventually-going-to-get-me-banned' contributions. In this case I was planning to accuse one of the recent respondents of being an AI bot, purely based on their track record of posting tedious, bland, get-it-from-Google replies. But someone's beaten me to it: that reply has been deleted (and hopefully the user has been banned.) Well done to whoever else spotted it.

Nothing has been deleted on this thread.

[DevonDamo]

My mistake: the bot post on this thread was posted a while back and is indeed still up. I had been intending to respond to their newest vanilla contribution which is the most recent post on a different thread.

I thought it was unusual for anyone on this forum to spot this kind of thing.

[Paulatic]

I thought it was unusual for anyone on this forum to spot this kind of thing.

If it’s OO you’re referring to then it’s certainly been spotted.

[DevonDamo]

I thought it was unusual for anyone on this forum to spot this kind of thing.

If it’s OO you’re referring to then it’s certainly been spotted.

That's the kiddie.

[DevonDamo]

[The following is an edited version of my initial 'throwing my toys out of the pram because my comments were split off' post. Since then, I've had an explanation of what the mods are doing behind the scenes, which is entirely reasonable.]

To make sense of the 5 posts above, you'd need to read them along with the other now-split-off thread - see link in first post. The specific issue on this occasion is a user posting stuff which may well be AI-generated. This isn't a threat in itself, but the internet is being overwhelmed with 'bot' accounts that use AI to blend in, before dropping their 'payload', i.e. links which are either spam or IT security threats. Obviously, with AI becoming so convincing, it's hard to spot this, so here's what caught my eye about this user:

- They posted 'vanilla' content, i.e. on-topic, but really doesn't add anything that you couldn't find from a quick Google search.
- They made a few gross mistakes, e.g. recommending a product which has absolutely nothing to do with the initial query.
- Their brief post history consisted of nothing but the above.

Unlike me, the moderators can't jump to conclusions with insufficient evidence, because they actually get to ban people. They've requested that I click on 'report' when I spot potential threats. I, however, prefer to flag this sort of thing up with the forum users, because our best protection from these threats is 'defence in depth' i.e. having a large number of forum users who know what to look out for. This is especially important when the dodgy users come up with a new modus operandi - the sooner we all hear about it, the less chance they'll have to pretend to be genuine forum members in order to insert their dodgy hyperlinks.

What do we think of the following as a compromise: when the moderators decide that a reported post is worthy of deletion and the user worthy of banning, how about keeping as much of the post and the follow-up discussion online as possible, rather than deleting the whole lot? If that were to happen, forum users would still get the 'heads up' so I'd be happy to report rather than to debate.

[Jdsk]

[The following is an edited version of my initial 'throwing my toys out of the pram because my comments were split off' post. Since then, I've had an explanation of what the mods are doing behind the scenes, which is entirely reasonable.]

To make sense of the 5 posts above, you'd need to read them along with the other now-split-off thread - see link in first post. The specific issue on this occasion is a user posting stuff which may well be AI-generated. This isn't a threat in itself, but the internet is being overwhelmed with 'bot' accounts that use AI to blend in, before dropping their 'payload', i.e. links which are either spam or IT security threats. Obviously, with AI becoming so convincing, it's hard to spot this, so here's what caught my eye about this user:

- They posted 'vanilla' content, i.e. on-topic, but really doesn't add anything that you couldn't find from a quick Google search.
- They made a few gross mistakes, e.g. recommending a product which has absolutely nothing to do with the initial query.
- Their brief post history consisted of nothing but the above.

Unlike me, the moderators can't jump to conclusions with insufficient evidence, because they actually get to ban people. They've requested that I click on 'report' when I spot potential threats. I, however, prefer to flag this sort of thing up with the forum users, because our best protection from these threats is 'defence in depth' i.e. having a large number of forum users who know what to look out for. This is especially important when the dodgy users come up with a new modus operandi - the sooner we all hear about it, the less chance they'll have to pretend to be genuine forum members in order to insert their dodgy hyperlinks.

What do we think of the following as a compromise: when the moderators decide that a reported post is worthy of deletion and the user worthy of banning, how about keeping as much of the post and the follow-up discussion online as possible, rather than deleting the whole lot? If that were to happen, forum users would still get the 'heads up' so I'd be happy to report rather than to debate.

Interesting and important points. I suggest posting them in Using the Forum to encourage as wide discussion as possible.

Jonathan

[DevonDamo]

Interesting and important points. I suggest posting them in Using the Forum to encourage as wide discussion as possible.

Jonathan

Thanks, and we'll put this one down as evidence for the 'it leads to confusion' argument in the debate about the desirability of splitting/merging/moving threads.

[Jdsk]

[Moderator note - the titles of all the posts in this split off thread have now been amended.]

Ah... this is the one that's already been split off?

How about changing the title to cover the wider points and attract attention?

Jonathan

[DevonDamo]

Ah... this is the one that's already been split off?

How about changing the title to cover the wider points and attract attention?

Jonathan

Done. (I'm not really the OP - this 'thread' is just a few posts moved from the original thread in the 'bikes and bits' forum - so I wasn't aware I could change the title.)

[Jdsk]

Thankyou

Jonathan

[Nearholmer]

I think the presence of bots is the best reason there is to allow thread drift, because the little devils harvest sites such as this for expert knowledge, which they then plagiarise and use. Having a discussion about
How best to grow geraniums, right in the middle of a thread about bottom bracket bearings is just the sort of thing that will scupper them when they parrot it all.

[Psamathe]

I wonder how clever many if these BOTs actually are. On my personal website my contact form has a hidden field "website" (name, e-mail & message being visible). BOTs are not clever enough to spot the "website" field is hidden so they fill it in whereas humans can't see it to fill anything so any form with website filled-in is a BOT so it gets trashed.

Surely something similar for phpBB

Ian

[Vorpal]

In addition to the information above, I would like to point out some other tactics that have been discussed on this board before, and are harder to spot. In addition, as spammer tools improve, it can become harder to find & filter out the crap.

Tactics in use by spammers:
-copy a post that is old enough that most of us have forgotten it
-copy material from another forum or cycling website
-post something that is entirely reasonable & personalised
-return later & replace posts or add stuff with advertising, dodgy links, etc.

I'm quite certain that some of this stuff is now being automated, and as it's becoming more sophisticated, it's also harder to spot.

So we very much appreciate users' help in spotting these things.

Unlike me, the moderators can't jump to conclusions with insufficient evidence, because they actually get to ban people. They've requested that I click on 'report' when I spot potential threats. I, however, prefer to flag this sort of thing up with the forum users, because our best protection from these threats is 'defence in depth' i.e. having a large number of forum users who know what to look out for. This is especially important when the dodgy users come up with a new modus operandi - the sooner we all hear about it, the less chance they'll have to pretend to be genuine forum members in order to insert their dodgy hyperlinks.

*Please* report these things, whether or not you post about them. If you want to flag stuff to raise awareness, feel free to do so.

That said, if the flagging of suspicious posting is likely to interfere with or derail the thread, especially a for sale thread begun by a legitimate user, we are likely to remove or split it. 'Using the forum' is a better place to post about it.

[slowster]

As Vorpal has commented, it's getting increasingly difficult to spot some of these bots and the like, and so it's also getting easier to make a mistake. If you submit a report and it turns out to be wrong, no harm is done. Accusing another poster on the public forum, even obliquely, may do more harm than good, especially if your suspicions are mistaken. Moderators have access to more information, so if you report, we can investigate further. Even then we may not be sure, but if we have been put on warning about a suspicious poster by a report, we can then keep an eye on them (as obviously can the person the who submitted the report) and wait to see if subsequent posts confirm the suspicions.

Also, bear in mind that the vast majority of forum members are probably not interested in this issue. I think only a very small number are concerned and interested, as indicated by the number who have posted in this thread and similar discussions in the past. The vast majority come to the forum to talk and read about cycling and other subjects. They will not want the threads they start, participate in, or simply read, to be derailled by discussions of bots, AI or spamming. You might think that more members should be concerned about this and should want to help, but for many the forum is somewhere they can leave the annoyances of real life behind and just discuss subjects they enjoy and of interest to them. They may not even care if a poster on their thread might be a bot, and may not thank you for disrupting the thread to point it out.

Most moderation activity dealing with this is necessarily unseen. If you wish to help, that would be very much appreciated by us, but don't expect much approval, gratitude or even interest from other forum members.

I think that the reality is that this will always come down to a handful of people - both moderators and members - who are naturally quick to notice suspicious behaviour and determined to stop spammers and the like. I would not be hopeful of interesting many others to help. Moreover, I am not convinced that many more people looking out for bots etc. is a good thing: I don't think it would be good/healthy for the forum to have a significant percentage of members on the lookout for suspicious behaviour. I think it's better to have a very small number of people who are naturally and instinctively good at it: they will have better antennae, develop better judgement through having greater experience, and be less likely to waste their own and moderators' time as a result of misjudgements and making false accusations. Think MI5 rather than Stasi. AI and bots are a problem, but they are not such a large problem that it requires active help from a substantial number of members.

For the most part I see no problem with discussion of these cases *after the fact*, but I think that it will usually need to be done in Using the Forum, and that is why in the past I have split threads and moved them here, for example this thread - viewtopic.php?t=160156. However, there will sometimes be information that I would not consider it appropriate to post on a public thread, because it could potentially help spammers and others, but which I might consider it appropriate to share in a PM with someone who submits a report. That is another reason why I think it's easier/better if it's just a small number of members involved.