That a police department is unable to distinguish thieves from the rest of citizens do not prove that there are no thieves but just that the police department is incompetent.
1. Do spam bots exist
2. Do "social bots" exist
For (1) the answer is clearly yes and nobody is claiming otherwise. However, academics don't see to care about spambots. That's a problem industrial researchers care about but not the sort of people who publish 'scientific' papers.
For (2) the correct answer is "there is no reliable evidence to support this". The term social bot is of course vague, but these papers are usually defining it as bots that attempt to subtly manipulate political discourse for ideological or geopolitical agendas. And when you go looking for evidence that such bots exist you end up finding what Gallwitz & Kreil found - that not just some of the research is bad, but all of it is. Where are the papers proving this problem actually exists in a clearly convincing way? There aren't any.
There's a very deep problem here which crops up in lots of other contexts e.g. why is there so much research into psionic abilities, why is COVID modelling so unreliable, why do scientific journals keep publishing auto-generated gibberish as "peer reviewed science" [1], why do so many papers in so many fields fail to replicate incl. not just social sciences but fields of critical importance like oncology? The root causes are the same, namely, academic institutions don't actually seem to care if research uses the scientific method or not. So you get things like Botometer as featured in this article, where it's just some ML model that's producing noise. It classified half of US Congress as bots and nobody cared. Thousands of papers have been published on the back of this model and zero have been retracted. It's because in that world what matters is publishing things that are interesting, not publishing things that are true.
[1] https://dailysceptic.org/2021/10/03/436-randomly-generated-p...
Society is currently designed to force the have-nots to LARP gladiatorial combat, not enrich each other’s lives.
> For (2) the correct answer is "there is no reliable evidence to support this". The term social bot is of course vague, but these papers are usually defining it as bots that attempt to subtly manipulate political discourse for ideological or geopolitical agendas. And when you go looking for evidence that such bots exist you end up finding what Gallwitz & Kreil found - that not just some of the research is bad, but all of it is. Where are the papers proving this problem actually exists in a clearly convincing way? There aren't any.
I'm not sure clear on your distinction between "spam bots" and "social bots," but clearly there is reliable evidence to support "bots that attempt to subtly manipulate political discourse for ideological or geopolitical agendas," exist, like plans to procure accounts for them.
https://www.nytimes.com/interactive/2021/12/20/technology/ch...
> Suppliers should provide accounts on overseas social platforms to purchasers at any time. The platforms include Twitter, Facebook, etc., and the supplier should provide about 300 accounts per month on each platform. Note: Content for this project is highly time sensitive. Suppliers should be able to supply accounts at any time.
I'd imagine those accounts would be deployed and managed in an automated fashion. Is someone automating accounts to send out propaganda operating a "spam bot" or a "social bot"?
a. The NYT has been a prime pusher of conspiracy theories about Russian foreign influence on US elections, none of which turned out to be reliable. Claims of Chinese foreign influence are very similar.
b. The media have published a great number of stories on supposed Russian/foreign conspiracies to use "social bots" and on investigation they are invariably relying on third party research of the sort that this paper shows aren't reliable.
c. The New York Times has become extremely biased in the last decade, especially post 2016 and especially on anything related to why people vote the way they do.
I would say though, that 10 accounts per day doesn't seem like enough to have much impact on anything at the scale of social networks. The claim the paper is making is twofold:
1. The research asserting the existence of large scale social bots is all flawed and doesn't really show that.
2. Even if there are people trying to do such things, there's no evidence to suggest it would or could work.
Given the saturation level media coverage of this topic and the widespread ideological belief in certain political quarters that you can rewrite people's politics with hashtags, I'm sure there must be some people, somewhere, trying it out. But if there are, the academic papers claiming to have spotted them aren't really doing so.
It's reliable, unless you're willing to go so far as to say the NYT would lie about Chinese government procurement documents: "The documents, which were part of a request for bids from contractors, offer a rare glimpse into how China’s vast bureaucracy works to spread propaganda and to sculpt opinion on global social media. They were taken offline after The Times contacted the Chinese government about them."
> a. The NYT has been a prime pusher of conspiracy theories about Russian foreign influence on US elections, none of which turned out to be reliable. Claims of Chinese foreign influence are very similar.
What outlets, specifically, do you think are more reliable on that and other topics?
> b. The media have published a great number of stories on supposed Russian/foreign conspiracies to use "social bots" and on investigation they are invariably relying on third party research of the sort that this paper shows aren't reliable.
On what basis can you personally judge the relative merits of research papers in this area?
Re: other outlets. I don't think any outlets are reliable on those topics because the whole idea this is really happening at scale seems to be made up.
Re: on what basis. Having read them. See the paper this thread is about. Also, having worked on bot fighting professionally before at Google.
>> What outlets, specifically, do you think are more reliable on that and other topics?
> Re: other outlets. I don't think any outlets are reliable on those topics because the whole idea this is really happening at scale seems to be made up.
So the other outlet is really just your own personal bias?
Honestly, I get the impression that you have a preconceived notion that social media manipulation isn't important (i.e. either it doesn't exist, but if it does it doesn't happen often, and if it does happen often, it doesn't work, etc.), and are moving goalposts and/or engaging in confirmation bias to reject anything that contradicts that notion. For instance, I don't think a serious outlet would claim you can "rewrite people's politics with hashtags" and the precise scale or effectiveness isn't super important (e.g. if someone's shooting at you, in important ways it doesn't matter exactly how many shots they take or if they actually hit you or not).
I'm also not clear on the distinction between "social bots", "spam bots", and a human manually operating many social media accounts like a puppeteer. I'm quite prepared to believe "social bots" don't actually exist (e.g. some kind of semi-autonomous "AI"-driven thing), but I'm quite sure the latter two do exist for political manipulation (not D to R reprogramming, more amplifying existing divisions and sowing confusion).
"I don't think a serious outlet would claim you can "rewrite people's politics with hashtags"
That is exactly the claim made by some social bot research papers. Until you spend some time reading the literature I don't think it's a great idea to form strong opinions about this field.
Try watching the talk version of this paper: https://archive.org/details/hopeconf2020/20200726_2000_Peopl...
In particular the section where he starts talking about causality inversion addresses this.
"the precise scale or effectiveness isn't super important"
Taxpayers are paying for this bad research because granting bodies and academic administrators incorrectly believe that it's a large and important problem. They believe that because they can't reliably detect pseudo-scientific research, which seems pretty important to me. Our money should be spent on researching things that are both real and important.
"Social bots" is a very poorly defined term (this is one of the issues with the research field as a whole) but generally means automated accounts that pretend to be real people, but which aren't, and which are engaging on social media to influence politics. A lot of claims about social bots imply Russia has developed general AI far in advance of anything in the west and is keeping it a secret just so they can troll Twitter, which is a bizarre and absurd claim on its face - probably why it's rarely stated outright, just implied.
"Spam bot" is easier to define because there's no implication of intelligence. Just getting a link to your webshop out there is sufficient.
Tools that help firms and people manage a single Twitter/FB account aren't a part of either definition.
>these papers are usually defining it as bots that attempt to subtly manipulate political discourse for ideological or geopolitical agendas.
Sorry but the issue smells recursive