No, #Bots are not to blame for #BREXIT

In fact, it seems to be a good example for the unimportance of #SocialBots

                                                                                                                 

The paper by Philip N. Howard (@pnhoward) and Bence Kollanyi (@bencekollanyi) has brought attention to the topic of social bots, again. This is very important because bots are already active in all social networks and there is a real risk of manipulating political decisions with this tool. 

Was the BREXIT-referendum manipulated by bots?

Howard and Kollanyi answer this question with "yes, but just a little bit":

"We find that political bots have a small but strategic role in the referendum conversations: (1) the family of hashtags associated with the argument for leaving the EU dominates, (2) different perspectives on the issue utilize different levels of automation, and (3) less than 1 percent of sampled accounts generate almost a third of all the messages. " (p.1).

OK, lets say the empirical findings are valid (I have some concerns, see below). There are bots! There are more tweets for leave and the total number of leave-tweets by bots is higher than the number of bot-remain-tweets.

What still is missing is any argument (empirical or theoretical grounded) that links the quantity of bot-tweets to an effect on voting behavior. There are bots and this reveals the possibility that they have an influence, but the possibility always means that it is not necessary.

In the conclusion, the authors provide an argument that links the data to a theory of voting behavior:

"The pervasive use of bots over social media heightens the risk of massive cascades of misinformation at a time when voters will be thinking about their options and canvasing their social networks for the sentiments of friends and family" (p. 5).

This idea is plausible: Undecided voters look in the social networks what their friends and family is thinking and then follow the opinion of the majority.

But data used in the paper does not tell us anything about the friend-follower connections of the bots. Normally, bots do not have many human followers. I think, it is very unlikely that someone is voting for Brexit, just because a fake-user to whom he is not connected is posting something. The whole argument was only convincing if the network-structure would be very dense. The number of messages tells us nothing if we do not know who is listening and how the listeners think about the information process: Are they receiving information from friends or do they think of the messengers as spam-bots?

Bandwagon and underdog effects

If we connect the whole topic with established patterns of political psychology, we might come to a different interpretation. In general, we can expect two different kinds of effects of the perception of upcoming elections. Voters might tend to support the majority, because it is always better to be on the winning side: bandwagon. Or voters might support the outsider position, because it is fair to be for the underdog. (Of course, Herbert Simon has written something exciting about this.) Both contradicting effects would be expected to be stronger in case one side is very likely to win. In the Brexit-case, I think it is reasonable to argue that the remain side had an edge. At least the bookies were in favor of remain and the traditional media, as well. Let's assume, on Twitter the Brexit-side was stronger. We know, that social media is used more frequently by younger people. So, if the bandwagon effect is working (like the authors seem to assume), we would have a pro-Brexit effect on young people. But we know, that this is not what has happened. Young voters have strongly supported the remain-side. Taken the bot-action in consideration, I think it is more likely that the bots caused an underdog effect on young voters, than that they had a pro-Brexit influence.

Some remarks to the data

I strongly doubt that the Twitter-streaming API is unbiased when you are looking for 20 trending hashtags. But this is something, Twitter is to blame for and not the authors.

Looking at table 2 in the paper, I have the impression that the data is not really supporting the thesis of the authors: The percentage of "heavy automated tweets" is nearly the same for #Brexit and #strongerin. So, if I am looking for #Brexit, the chance to see a bot tweet is not higher than for #strongerin.

The bot rate is higher in tweets that mix pro and contra hashtags. I think this is not surprising, because many bots are just spam-bots who do not care about brexit but just post links to e.g. Russian video platforms. But it is very unlikely that these tweets will convince anyone to vote for or against brexit.