On Big Data and Post-Truth

In his book “A Mathematician Reads The Newspaper”, John Allen Paolos writes on how an understanding of mathematics can change your understanding of daily news. One of the examples he offers is to imagine the following (and plausible) scenario on gun control in America. Opinion polls show that up to 80% of the public support some form of gun control. Yet all politicians are reluctant to touch this subject. Why?

Paolos posits that that we naively believe that in a democracy the majority opinion carries the day. However, he argues that it is not the question of the comparative numbers in the majority or minority, but how these numbers break down among themselves. In this scenario on gun control, among the 20% who oppose gun control, three-quarters of them (NRA members, etc.) are so fanatical on the issue that they will make a voting decision purely on this one issue. Three-quarters of 20% is 15% of the electorate. Of the 80% who favour gun control, most of them do so among a vast spectrum of ethical or libertarian issues, and only 5% of them (who may have been provoked as victims, or closeness to victims, of gun-related crime) will make a voting decision solely on this issue. 5% of 80% is 4% of the electorate. In terms of single-cause constituencies, you have 15% of the electorate on the minority side and 4% on the majority side. The 11% difference will swing most elections, the politicians know it, and therefore the reluctance to touch the issue.

This means that democratic politics does not like diffuse majorities, and prefers the commitment of fanatical minorities. Campaigns depend on generic promises that aim at attracting a certain mass base, and then courting the swing votes provided by fanatical minorities. There was a time, until a few years ago, when any political campaign had to make a choice on a small number of these single-cause constituencies it wished to court as the swing votes. And given the narrow range of choices, the ideology of the campaigner had to be clear, logically consistent, and perceptible by the general public.

Has this changed in this era of big data, especially the sophistication of the analytics that exploits big data? It is significant that a single British firm, Cambridge Analytica, was involved in two of the most globally significant political campaigns of 2016: the Brexit Referendum and the US Presidential Election, and worked for the winning side in both campaigns: the Donald Trump campaign, and leave.eu (one of the major groups in favour of Brexit). For more on this see an article in The Spectator, a translation of an earlier German article found on the online journal AntidoteZine, and the firm’s own website. Significant common features of the Brexit and Trump campaigns were that both produced results that were not predicted by traditional polling techniques, and were criticised for not caring whether their public postulations had any foundation in objective facts or truth.

Cambridge Analytica examines large data sets that were not available earlier, such as Google searches and Facebook likes and posts, to construct a personality model described by the acronym OCEAN (openness, conscientiousness, extraversion, agreeableness, and neuroticism) that predicts with surprising accuracy the choices that a person may make. With enough data, the analytical model can predict choices with greater accuracy than someone very close to the person whose personality is being mapped; or if the data set is large enough, even more accurately than the person himself/herself may be able to comprehend.

There is some controversy over the significance of the role that Cambridge Analytica played in the Brexit and Trump campaigns, and the extent to which the OCEAN model was applied.  But it is clear that the existence of these new analytical models allows mapping of the geography of single-cause constituencies with a hitherto unparalleled accuracy, offering new opportunities for electoral campaigns. The problem is that the richness and fine grain of the data analytics reveal a set of constituencies that can be quite diverse, and may have insufficient overlap for the articulation of a consistent ideology. If an electoral campaign wishes to mobilise as many of these fanatical groups as possible, it cannot afford to articulate a logically consistent position, and must base itself on a common ground that is most easily found in a language that centres on the spectre of fear and loss, raising emotions to a pitch where there is little concern for seeking validation in facts. Perhaps, that is why the Oxford Dictionary has had to declare “post-truth” as the word of the year for 2016.

Advertisements