AI bots trolling forums [Resolved - bot/user banned]

the_amazing_raisin

So today there was a new user trolling one of the forums. I strongly suspect they were an AI bot based on something like ChatGPT or similar.

The user was posting lengthy replies across multiple threads spouting the same BS logic. They had zero spelling or grammatical errors despite posting at almost inhuman speeds

Frankly I feel that this is just the beginning and we're going to see a lot more of this in the near future

I'm wondering what tools and changes boards.ie can take to combat this?

Gloomtastic!

Links please?

HazeDoll

https://www.boards.ie/discussion/2058284734/anti-ev-nonsense

Co Clare Man is the poster in question, I suspect. Now banned.

This is going to keep happening. In itself it's probably harmless enough, though annoying. The problem is that any new poster will be regarded with suspicion, which will have a very negative impact on the general flow of boards.

I don't know what the solution is. Is there a way to filter for posts that are appearing suspiciously quickly? Faster than a normal person can type?

the_amazing_raisin

https://www.boards.ie/discussion/comment/120254466#Comment_120254466

Yep that's the one I was speaking about

I suppose the site owners can rate limit the number of posts new users can make, but the bots could also be rate limited to avoid appearing suspicious

Allinall

They need a stricter process for signing up.

Don’t know what that is, but whatever is in place at the moment is totally inadequate.

The number of banned posters re-registering is unbelievable.

the_amazing_raisin

https://www.boards.ie/discussion/comment/120254433#Comment_120254433

Example below look for posts by Co Clare Man

https://www.boards.ie/discussion/2058284734/anti-ev-nonsense

The user posted across a bunch of other threads but it looks like the mod cleaned their posts up

My main worry is that an attacker could use several bots to swarm a forum with a load of BS posts. The mods are already under enough pressure to clean up after the human trolls without having to contend with an automated one

the_amazing_raisin

https://www.boards.ie/discussion/comment/120254484#Comment_120254484

I agree in general but there's limits to what boards can accomplish without making the user need to provide personal information, which the site owners seem to want to avoid (makes sense since some users can post some very personal stuff at times)

HazeDoll

There's also the possibility that a real person can legitimately sign up and then begin to post by copypasting waffle from AI. I would imagine that's very hard to detect.

If somebody wants to do that for their own amusement, and have a laugh at the foolish boardsies who are taken in by the AI's contribution, it's annoying and trolling but not the end of the world. I know it makes work for mods too, as they can't ignore it. However, I think the real problem will be when a newcomer posts something they've put a lot of thought into and it's expressed in clear language with no spelling mistakes. We'll all point and shout "Bot!" Hardly a welcome that will make somebody stick around.

In Darwinian terms, the posters with half-baked notions and no understanding of punctuation will find the site more welcoming than somebody who has taken the care to express themselves clearly in writing. One group will proliferate and one is headed for extinction.

That’s not a bot

Furze99

https://www.boards.ie/discussion/comment/120254466#Comment_120254466

I saw that but also thought it could be a genuine poster just stirring things up. I would agree with several of their points. If it was an 'ai bot' then pretty sophisticated as it tailored responses in replies - quite possible but authentic. As for spelling errors, well some people are good at spelling....

But where the ai bot account has occurred to me recently on the controversial Russia thread as regards the ongoing war. There's definitely been a series of posters on that thread who tend to word & work arguments the same way, almost Pavlovian. Sometimes they get banned, sometimes they just disappear. New accounts appear who are immediately at the heart of any arguments. These could be real people or just one person or just as easily some kind of ai. I've noted this odd activity in the thread and been warned by mod.

But I really can't see how social media discussion forums can avoid this type of ai activity - maybe it's the beginning of the end.

Markus Antonius

We could do with less human and more bot

Thargor

https://www.boards.ie/discussion/comment/120254650#Comment_120254650

It most definitely was or at least it was someone copying and pasting replies for it, it was typing reams of paragraphs too fast and there was an uncanny valley feel about it straight from the first thread I was reading until I checked my bookmarks and saw he was posting in all the EV thread at the same time practically and replying to every single user that quoted him.

Have you watched ChatGPT in action or tried it yourself? Thats how it talks, borderline human unlike all previous chatbots, you could literally tell it to pretend to be a Boards.ie poster who posts negatively about electric vehicles and it would do it. Tom Scotts latest Youtube is about how its going to change everything and you wont know who or what you're speaking to soon, theres no way Boards.ie is going to be able to defend against it. Twitter is just going to be legions of bots arguing with each other, not that it isnt currently.

Furze99

One way to monitor it is I guess to see what threads or forums an account is active on. If it's continuously one thread or one forum or related threads, then suspicions would arise. If a poster is thread banned but continues to post on a variety of other threads, more likely to be genuine as it would be more typical human behaviour. But of course, I'm sure someone with the smarts could mimic this anyway and set up an ai bot to target a cross section of threads and forums.

the_amazing_raisin

https://www.boards.ie/discussion/comment/120254782#Comment_120254782

Yeah that's the thing with these new ChatGPT bots, they're very sophisticated but they do have this inhuman feel about them. The uncanny valley is a good analogy

I don't know for sure that the poster was a bot or a human and I've no way to prove it, but my suspicion is it was a bot

The thing that got it for me was the way the poster was giving lengthy replies rapidly across threads for several hours. Any human trolling the site would have gotten tired and bored after an hour or so and taken a break, not keep it going through the whole day

Oddly enough, I think the defence is also going to be an AI of some description. It's quite possible that an AI model could recognise text usage patterns of frequent posters and flag ones that seem like they could be bots for further review

the_amazing_raisin

https://www.boards.ie/discussion/comment/120254802#Comment_120254802

The poster in question seemed to be asking a whole load of dumb comments across several forums, a lot of which have now been deleted

I agree in general though, the trolls tend to target a specific forum rather than going after the whole site, so maybe that's a way to spot it

HazeDoll

Maybe there's a way for an AI to keep an eye on how posts are typed in the "Leave a comment" box.

Most of us make a mistake or two when typing, or delete a sentence or pause and think about what we'll say next. We edit and make changes before we post. Maybe this sort of thing could be identified as a pattern of human behaviour, which would expose the bots among us. It might be a bit like the 'no CAPTCHA reCAPTCHA' software that looks at your behaviour, watching for human characteristics. There might have to be some sort of allowance for real people who occasionally copypaste text like a piece from a news article. (There are also people who fire off a post without ever reading back over a single sentence but I think their haphazard mode of expression would identify them as human quickly enough.)

Oscar_Madison

I spotted him (Clare man) early on - I pitied the people engaging with him as they were trying to engage as if he was a new poster but it was very obvious what was happening.

I assumed this was a guy cutting and pasting answers he got to other peoples posts from ChatGPT.

So can I ask, can the bot independently go off and start posting gibberish where it wants on boards or does it still require human intervention? I just thought it was someone cutting and pasting content the bot was supplying them.

Either way it Might be an idea to educate or inform posters on boards to expect more of this tripe and to report as early as possible and not to engage- have to say it was easy enough to spot and a poor attempt but they’ll only get harder to detect I’m sure.

Calahonda52

https://www.boards.ie/discussion/comment/120254650#Comment_120254650

On watt basis do you make that assertion, I was in a number of the threads that were running and it was unreal

Calahonda52

The next thing will be a bot based DoS

the_amazing_raisin

https://www.boards.ie/discussion/comment/120269993#Comment_120269993

Already exists, and there's ways to combat it that some cybersecurity companies can employ

the_amazing_raisin

https://www.boards.ie/discussion/comment/120269964#Comment_120269964

It should be possible in theory to automate a chat bot into boards. I'm not sure exactly how the backend software of boards works but it could be possible to send an API request directly that looks like it comes from a browser

Failing that, you could probably use something like selenium to drop text directly into the reply field in a browser and post it. You should even be able to refresh the page and see if there's any replies and feed that back to the chat bot automatically

So yeah, very possible to automate it. All the technology exists already

Oscar_Madison

https://www.boards.ie/discussion/comment/120270182#Comment_120270182

Cheers- I know nothing of such things hence the question- , - obviously whatever interface he used, he was able to spam the site quite quickly with ease -