[Thread] 1/ Here are some reflections on the nomenclature related to bots, trolls, & 'influence campaigns.' It's something that comes up a lot & I wanted to reflect on. Firstly, lots of people use the terms troll and bot interchangeably. While this might be a convenient #polbots
2/ shorthand for discussing the vast array of 'weirdos' online they are distinct. Bots are usually automated or semi-automated accounts, programmed to engage in specific behaviors usually at frequent intervals. On Twitter they may look like real accounts, and are designed to
3/ deceive. Trolls are usually real people behind an account. While the term 'bot' describes the likelihood of automation, troll is more a behavioral characteristic. Trolls engage with people, intimidate, harass, criticise, deflect, use logical fallacies, and spread
4/ disinformation. Bots can do this too, although I am not convinced the level of sophistication exists yet. I usually tend to find bots boost certain accounts, or mass tweets certain points of view. Of course maybe bots are so sophisticated now I can't tell them from trolls (I
5/ doubt we are there yet though). For the most part, bots can be set up with a minimum level of expertise. On the other hand, vast amounts of bots probably require a little more investment and expertise to operate. Regardless of how you cut it, bots can be the work of
6/ individual entrepreneurs selling their services to the highest bidder, or more institutional entities like PR companies or other advertising agencies. This can be confusing when looking for attribution. If people who own bot networks have multiple clients over time, then it
7/ likely that those bots will bear traces of multiple, at times incoherent or unrelated promotion campaigns. Political bots or political spam accounts may only be political for the duration of a specific contract, unless they are specifically run inhouse by some political
8/ campaign team. Similarly, political bots will tweet content that is designed to humanise them, so between their political campaigning they might tweet any amount of bilge or content designed to make them look more credible and less spammy. Trolls are a different beast (
9/ literally and figuratively). If we see trolling as a behaviour then anyone can engage in trolling. Anytime you harass, intimidate, spread propaganda etc you are most likely engaging in 'trolling'. If this behaviour is habitual then you are probably a troll. I would not say
10/ a troll needs to have any inherent political definition. This is why qualifiers are important. Are we talking about 'political trolls', 'paid trolls' etc? For the most part, when we refer to malicious activity and influence campaigns we are talking of multiple accounts
11/ engaging in a pattern of similar behavior involving attempts to intimidate, harass, spread disinformation, and use logical fallacies to spread a certain idea. I would say an 'influence campaign' is a concatenation of such actions, each with a specific, coherent message.
12/ Now the question remains is that can such 'influence campaigns' be described as behaviors by real (authentic) or non-real (inauthentic) accounts. Well I'd argue both, but first what is the difference. An authentic user is presumably a living human being whose beliefs are
13/ are, for the most part, not perceived to be the views of an employer who is paying or coercing that individual into engaging in online behaviour designed to spread a particular political message. An inauthentic user is, then, presumably, a non-human or a person who is
14/ paid, co-opted or coerced into spreading a specific message. But now we have a difficult question. At what point does the agent of coercion/payment/co-optation play a part in the definition. Is someone a troll only if they are directly exposed to one of the above conditions?
15/ Perhaps, but how do we determine that - well without a smoking gun we cannot. What if an authentic user who has been exposed to extreme propaganda and fake news absorbs those ideas and reproduces them at their own free will, acting aggressively to spread them? What if
16/ that person just watches a lot of Fox News and Alex Jones and is emulating that style of behaviour? This is a big problem of course, especially when we are trying to define an 'influence campaign'. We can thus have a broad and a narrow definition of an influence campaign
17/ A Broad definition may be one in which a number of seemingly connected accounts (connected by biography, aesthetic, communication) tweet or engage in high volumes on a specific issue, repeating un-nuanced and similar talking points, aggressively. A narrow definition
18/ could be a campaign where those users engage in the above behaviors but are explicitly connected to an entity known to procure specific services, whether a PR company, the IRA, the CIA - however you want to perceive it. The narrow definition is extremely difficult to
19/ evidence, the broad definition a large net that will suck in all sorts of people. However, we have to start somewhere, and articulate those points. That's why I have generally been careful in my multiple data threads to avoid being specific about attribution
20/ However, I do believe the broad definition is useful. In my experience, it does not take a great number of 'inauthentic users' working hard to pollute debates among real people. When successful, this will mean authentic users then replicate talking points prescribed by
21/ inauthentic users. Here the functioning of the bot or troll masters, and whoever pays/coerces/coopts them becomes truly dangers. Just as the old media models rightly examined the impact of small groups of people determining editorial content, there is the risk that
21/ such behaviours replicate themselves on social media. If purveyors of fake news can successfully reproduce their ideas among hosts, like a virus, then that's when the line between influence campaigns and 'achieved influence' becomes nebulous.
22/ Anyway, on that dystopian note, happy tuesday and happy reading!
