It feels urgent, with elections approaching in several democracies, to focus on lessons from 2016, when the Brexit referendum and the United States presidential election were both marred in ways still coming to light. Among several valuable reports, a recent study for the European Parliament’s Panel for the Future of Science and Technology usefully encapsulates the challenges without unremitting alarmism and pessimism. It is clear-eyed about both the benefits and dangers to democracies of technologies that are in the hand-held devices of most voters.
Against a backdrop of increasing polarisation, “new digital technologies have taken centre stage in political processes — both as a source of information and a campaigning platform”, says the study, Polarisation and the use of technology in political campaigns and communication. “Such new and relatively unregulated platforms create new opportunities for nefarious actors to deliberately push false content and distort information flows for political gain.” But artificial intelligence will also offer new opportunities for better accountability and transparency.
A cited example is Google’s Perspective, trained “to automatically detect toxicity in comments in real time”. The machine learns from millions of comments made by humans. Toxicity is defined as “a rude, disrespectful, or unreasonable comment that is likely to make you leave a discussion”, and comments are ranked on a scale from “very healthy” to “very toxic”. Initiatives such as this ought not substitute for human judgement, but they may be helpful indicators, able to flag what humans need to focus on in the fractious, febrile, time-pressured campaign atmosphere.
A striking feature of many analyses of what went wrong in 2016 is the harnessing since then of the power of computing to map the vast amounts of data that have been gradually made available about how electorates communicated through social media in the lead-up to voting. Some yield instructive surprises, such as the fact that Russian trolls used Instagram, which Facebook owns, more than Facebook, to which the Cambridge Analytica scandal later drew most attention.
The European parliament panel’s study cautions against hasty legislative responses, and notes that laws in a jurisdiction can, if misjudged, be ready models for anti-democratic action elsewhere.
The Australian parliament demonstrated the hasty mistake last week when, in one day, in response to the Christchurch shootings and just before they dispersed to fight an election due next month, parliamentarians created far-reaching new criminal offences about the sharing of abhorrent, violent material on digital platforms. A product of laudable intent but lousy process, the law is predicted to have unintended overreach.
Two of the lessons are old, but have to be relearned in a bewilderingly fast-changing environment. First, verify, and with care that increases in proportion to the electoral impact of the information being presented.
And demand transparency, especially from the tech giants, on whose platforms so much of contemporary election campaigning takes place, both as paid advertising and in the informal but consequential many-to-many sharing of links and views.
The panel’s study also lays stress on transparency: “Fundamental algorithms remain ‘blackboxed’, the collection and use of personal user data is opaque, and content moderation frequently lacks public oversight ... Shedding further light on online information ecosystems, their functionalities and core mechanisms is not just a concern for the general public. It is also a concern for political actors and producers of journalistic content who are the traditional gatekeepers and agenda-setters of political information in Europe. As they rely more heavily on social media platforms in order to inform the public, connect with citizens and mobilise voters, they also become subject to algorithmic promotion and filtering processes that lack transparency and oversight.”
— Guardian News & Media Ltd
Paul Chadwick is the Guardian’s readers’ editor.