Algorithms and the persuasion machine

In the current BBC series Secrets of Silicon Valley Jamie Bartlett (technology writer and Director of the Centre for Social Media Analysis at Demos) explores the ‘dark reality behind Silicon Valley’s glittering promise to build a better world.’ Episode 2, The Persuasion Machine, shines a spotlight on several of the issues we are investigating in UnBias.

The episode focuses on social media and the ways in which user posts, interactions and likes etc. on platforms such as Facebook can be collected and used as data by advertisers and political marketers. These algorithm-driven processes are in fact so powerful they appear to have the capacity to influence the results of national elections.

In the programme Bartlett talks to a psychologist at Stanford University who explains how highly sophisticated algorithms and tracking procedures can collect up masses of data about an individual via his/her online digital footprint. This can then be combined with psychometric testing to build up a startling accurate prediction of that user’s personality – including political persuasion, professional background and religiosity. Bartlett allows his own Facebook posts to be tested against the psychometric model and is very surprised that the model is able to pinpoint his religious upbringing, profession and academic history – even though he had not posted explicitly about any of these things.

These data collection and analysis processes are hugely valuable to advertisers and marketers as they enable them to direct highly personalised messages to online users. Bartlett gains unique access to find out how these techniques played a crucial role in the election of Donald Trump. Alexander Nix at Cambridge Analytica explains how the company used telephone survey and online platform data to provide a dataset to the Trump campaign. This dataset provided information on over 220 million US voters and enabled the Trump campaign to target potential supporters and donors with personally tailored messages.

As Bartlett points out, many social media users are likely to be unaware of the extent to which their digital footprints can be used to identify personal information about them. Furthermore they also may be very uncomfortable about the ways in which this personal information is used to target them with highly personalised messages. But this discomfort does not appear to be shared by the proponents of these innovations. Alexander Nix at Cambridge Analytica describes personalisation in (political) advertising as ‘exciting’ and ‘inevitable’. A director of Trump’s Digital Content campaign team reports that representatives from Facebook, Google and You Tube all made frequent visits to their campaign headquarters. She describes them as ‘hands on partners’ who played a key role in helping the campaign to identify ways in which online platforms can be used as a mechanism for political persuasion. Bartlett reflects that social media is now being used as a persuasion machine contributing to the new, unpredictable political landscape we see around us. This is an unintended, unexpected consequence of our hyper-connected digital age: do we accept it, and the turbulence it brings, as ‘inevitable’ or do we seek to find ways to regulate and control it?

3 thoughts on “Algorithms and the persuasion machine”

  1. Thanks for raising this discussion Helena. As a citizen, I would definitely like to know how a politician (especially someone who is an individual Presidential candidate rather than a PM who represents a party) relates to me personally, and in this I see a huge potential for personalised political advertising (e.g., I may care more about monetary policies rather than defence or foreign policy, so show me things I care about more.). But at the same time, this personalisation allows the candidates to be janus faced and present themselves differently to different citizens, which is hugely troubling.

    1. Thanks for your comment Nishanth. I agree with you in finding this kind of personalisation troubling despite its potential positives. Directed political messages are not new but this is on a much larger scale. Personalised messages can build up a rapport between voter and politician but the techniques behind them here would seem to undermine that authenticity and allow for a worrying level of manipulation.

Leave a Reply