Nov 9th :  Whatever the result, but not for the reasons you might think! Or Angry People Click!

Head over to to read our most recent blog.

“The current presidential campaign is amongst the most polarised in US election history, and it is no coincidence that it is also the first election to be so completely dominated by Social Media. Against a background in which both campaigns are using Software robots, or Polbots, of varying degrees of sophistication and one of the campaigns, is apparently being led from a twitter account, this is a defining moment in political history.”


Against a background in which both campaigns are using Software robots, or Polbots, of varying degrees of sophistication and one of the campaigns, is apparently being led from a twitter account, this is a defining moment in political history. In recent weeks commentators on both sides of the political spectrum and one of the candidates have begun to warn us of some of the consequences of this election, not least the prospect of widespread dismay and anger when the results of the election are declared. As a researcher and engineer working in the field of Political Bots, I feel it is essential to make it clear that the strong emotional response that is undoubtedly about to break out is driven to a significant degree by a factor which few political commentators are, as yet, even aware. I will describe a self-reinforcing feedback loop which will, I suggest, lead to a situation without precedent in a major democratic country.

I say without precedent, but in fact, political scientists have already begun to describe this phenomenon, and indeed we have an excellent comparative case in the form of the 2015 Israeli national elections, and I will discuss those events in more detail below. So what is happening and why should it lead to new levels of dismay and worse, amongst the electorate of a stable country with what will be a legitimate and democratically elected president.

I must, first of all, make it clear that this argument is not derived from the nature of the candidates, their policies or positions but instead arises from the nature of this new situation for political discourse. Today in the US the broadcast media no longer holds power to establish or promote either an agreed agenda and criteria for the content or even manner of delivery for the discussions surrounding the events of a campaign. In the present situation, we must recognise that heterodoxy is the order of the day, no one institution, organisation, group or class can ‘manage’ an individual’s experience of the campaign. Because in 2016 the means and methods of creating, disseminating and commentating upon events lie in a multitude of channels of communication. Channels which are almost as diverse as they are numerous. I am not arguing that PolBots or deliberately partisan software agents of any kind will help frame this election, although that may very well be the case to some extent given the levels of their use. What I am arguing is that the loss of dominance, the decline of the hegemony of the Broadcast media is part of the issue. More important is the significance of software algorithms designed and used for commercial and not primarily political reasons when combined with natural human tendencies.

The frightening heart of this problem is that the Internet is not so open as some might assume. It is not a free for all in which all voices can be heard and all stories told. The reason for this is not some form of censorship or a gigantic conspiracy funded by one side or the other but rather an unseen side-effect of the way in which the Internet is funded, operated and ‘optimised’. Perhaps the most worrying fact here is that the source of much of the trouble we will face comes from the extent to which people tend to listen to the people, voice, message or event that speaks to their preconceptions. As psychologists would put it, we all seek out the confirming instance; that piece of evidence which confirms the way we think. We human beings do not seek out the stories, interpretations or evidence which contradict our positions. Political scientists use the term Homophily and are already describing how the tendency to associate with and to like people similar to ourselves is framing discourse online. Now if that were not bad enough and it is terrible for the democratic process under any circumstances, but it has always been present. The Internet, unfortunately, is designed — explicitly — to amplify just this behaviour.

How then can something as complex and abstract as the Internet take up a simple human characteristic and multiply its effect in such a way as to destabilise the democratic process? Well, the answer is perhaps all too shockingly simple. The application of a few, very straightforward rules to a series of actions that are being carried hundreds of thousands of times an hour by tens of millions of people. The people are connected not by politics but by algorithms, algorithms which feed their core predilection for the same or similar people, ideas, views, events, products and votes. I will look at two different types of Internet experience, each carried out hundreds of millions of times a day and showed in simple terms what is happening.

The internet is designed regarding the political assumptions that underpin its operation; it is not value-free just because these foundational concepts are not explicitly labelled political. One of the founding principles of many Internet products and services, in fact, one of the significant claimed benefits of the Internet is the drive to create and locate services and products for the individual. In fact, the primary source of revenue for the operation of the internet itself is advertising, and the key goal for advertising is the drive to personalise the content of web pages. Everything from searches to the content of the page and the advertisements it carries are designed to match the location, needs, desires and values of the individual user.

So an essential feature of the internet that is leading to the creation of false expectations around this election is the extent to which the individual can follow their desires to create or curate if you will, a series of moments which will inform and amplify their chosen ‘story’. Their chosen environment is always adding evidence and argument to bolster their existing ‘take’ on the world in general and the election in particular. So for example, we will browse social media feeds, favourite blogs, messages, news websites, discussion forums and address the call-outs from all of our subscriptions and notifications. The entire range of which is usually, designed to appeal to us for the simple reason that we chose it so efficiently thanks to modern algorithms. Now every one of the agents with which we have chosen to interact will have used one or more data mining techniques to craft the deadlines, notifications, tweets, etc. to capture our attention. In short, they will, as far as millions of dollars and human ingenuity allow, have helped us create our echo chamber. It is this effect that lies at the heart of the problem. So as each day passes more and more of the electorate will go, for example, report via polling that they believe the election to be ‘rigged’. The evidence that is piling up for this is that wherever some of us look, someone, or indeed everyone is saying just that. No amount of traditional media ‘corrections’ can overturn that form of knowledge, because they (traditional media and our personalised experience of the ‘news’) are not equivalent either concerning their visibility or salience to the individual. It is not so much that we live in a post-fact world, but that we are also profoundly immersed in forms of knowing that are not now nor ever will be fact-based at all.

The second aspect of Internet-based products and services designed to support the production and consumption of ‘information’ is once again application of simple algorithms to locate individual subscribers and users in the context of a geographical, social and economic space. Targeting individuals using ever more refined techniques to create the ‘hyperlocal’, ‘personalisation on steroids’ feed. So, for example, Facebook timelines, Twitter feeds, discussions on news and blog sites and even the results of Google searches all will be adopted by marketing companies and service providers to supply our choices and serve to amplify the echo further. Of course, all of these platforms also feed to a certain extent off one another, so the combinatorial effect is also to intensify in a variety of ways the echo chamber effect. Google search has increasingly included social data as the recent deal between Twitter and Google only serves to illustrate.

Researchers have demonstrated that Facebook presents users with only a (small) fraction of the posts generated by their friends. Furthermore, This selected feed is further limited as Facebook prioritises homophilous content. Because, as we have said, people on the internet exhibit Homophily. The tendency of individuals to associate and bond with similar others or those with whom we are more likely to agree. Facebook deliberately amplifies just this tendency in constructing our timelines. So much so that even if our friend was likely to disagree with us or mention an article, product or event that we would not like then they are far less likely to appear in our timeline. Facebook’s EdgeRank algorithm that decides which piece of content should be displayed on user timelines takes into account engagement (likes and comments) on posts as part of this process.

If Facebook is a platform seemingly designed to operate as an echo chamber, then the Twitter platform will surely echo to the beat. Here we can be confident that unless you have chosen to follow a user or source, then you will receive nothing and the usual commercial constraints shape the material added to user feeds. Twitter is home to the PolBot, and during this present election researchers have repeatedly described the action of large, well organised and resourced twitter projects to drive Twitter activity in favour of a particular political position.

One further example will serve to illustrate how the commercial instincts that drive the Internet shape our experience of political discourse. Disqus Inc. is one of the leading companies in the world for providing internet tracking services and closely allied to this advertising services for content providers. As they have said, ‘We have the largest and deepest audience profiles on the web…’ Whether or not a site allows advertising Disqus is following what is said and by whom. That data is sold. Hence the words and ideas we employ or comment upon are part of the information set at the disposal of other service providers such as marketing companies to use in shaping our experiences both on and offline. Disqus also provides the Comment space service to marketers, this also offered directly within the comments; Sponsored Comments delivered right to where you live. Especially if you are there to express your anger, advertisers are well aware — angry people click!

The experience of the political left in Israel in 2015 may serve as the marker here, although to a degree the Republican party or at least some of its strategists seemed shocked at the defeat of Mitt Romney in 2012. In Israel, people spoke of waking up the morning after the election to discover a burst bubble — much as the left in the UK felt after the Brexit vote. The centre-left in Israel had mobilised actively and mainly on social media; they were feeling very upbeat. Because as one newspaper reported; ‘everyone they knew was either voting Zionist Union or Meretz’. (The opposition parties trying to prevent a further the term of office for Prime Minister Benjamin Netanyahu). Most important was the fact that ‘Nobody was speaking to the traditional Likud voters or the religious voters outside the camp.’ The atmosphere of shock and deep despair that overtook the defeated voters from the left of the electorate was all the more devastating because their own experience of the election campaign had been positive and reinforcing. In other words, those who lived inside the bubble finally had to share the same reality as those who had lived in another bubble during the election. A very similar reaction to that described in left-wing newspapers in the days after the 2016 Brexit vote in the UK. Here we had the almost laughable situation of millions of citizens searching for information on what the EU is and what leaving might entail but after the event. Raising the question as to what had been the basis for the decision in the first place?

The phenomenon of an echo chamber or bubble in which the electorate spend their time during an election campaign is probably an extension of the impact of traditional partisan media in the past. However, it is clear that the Internet is expressly designed on many different levels to amplify this effect and it would be reasonable to assume that on November the 9th there will be many millions of people who wake up to something other than what they will have expected — let alone wanted. Much of the intensity of the reaction that will inevitably follow can lay at the feet of the internet — where much of the subsequent anger, doubt and despair will no doubt find a home.

It is tempting to hypothesise that many of the changes we are observing in the present campaign are not so much a product of the candidates themselves but rather a reflection of the roadmap laid out for political discourse in the 21st century by the addition of non-traditional forms of communication. The production and dissemination of a highly personalised and massively diversified space for conversation is leading to an increasingly partisan tone of the debate. A deeply polarised and polarising phenomenon which itself leads to the death of facts. The truth is made for momentary consumption and lies travel the world before the truth has its trousers on. It would also seem that truth need not have bothered; many will ignore it and most never ‘believe’ it even if they noticed, which is becoming increasingly unlikely. In the third week of October, those using Google to find out about the position taken by Mr Trump on the economy are so far exceeded by those concerned with his statements concerning the wall that they are in danger of being missed by Google. Notwithstanding the views of professional economists, administrators, politicians, foreign heads of state or most conventional news sources about the insecure nature of his economic proposals. What people are interested in, simply put, what they are interested in — the wall. Consider the top 5 questions about Donald Trump trending on October the 18th as shown above. What matters to those deciding how to vote and interested in Mr Trump is how are other significant people voting and the age of the Trump family. Facts such as these are at the heart of peoples’ narratives and understandings about the election. Hardly shocking in the age of the Internet but still revealing.

The candidates themselves would seem to have devoted both rhetoric and advertising spend to the task of addressing those with whom they share a ‘bubble’, speaking the truth to their own. Donald Trump has made it clear that he is loyal to those who are loyal to him, the ‘message’ is the man. Conventional attempts to marshal evidence and argument, to engage in debate and win over those who do not share his ‘values’ are as relevant here as Republican Party policy documents.

The internet both affords and amplifies these tendencies — partisanship and fact-free campaigning, reinforcing a view that practical politicians will have to follow Mr Trump down this rabbit hole if they are to prosper. Perhaps like the experience of Israel, we will witness a call to legislate for the effect of social media upon political campaigns in some way, but it seems unlikely that such an initiative could succeed. Instead, we should prepare ourselves for more of this unwitting and deeply unpleasant change in the way we both conduct campaigns and live through the aftermath.