Against opacity, outrage and deception in digital political campaigning

In the 2015 UK General Election, £3m was spent on advertising on platforms, companies, consultants and strategists. In the 2017 General Election, over £6m was spent. This includes increased use of data analytics (automated insights into datasets using data-mining techniques), and data management approaches to “profile” audiences (via mathematical techniques to discover patterns in “big data”). It includes increased use of iterative, large-scale, rapid testing of ads online (“A/B” tests) to identify and deploy the most persuasive ad; and to gather data on, and target, the most important voters with tailored messages. 

While using such techniques to mobilise activists and engage voters is good for democracy, these processes are opaque. Potential harms arising include fragmented national conversations (A/B testing allows secret negative messaging to niche audiences without alienating the broader electorate); and undue political influence over voters by exploiting their vulnerabilities (such as inability to recognise deception).

In 2019, the UK Parliament’s House of Lords Select Committee on Democracy and Digital Technologies asked whether greater transparency in digital political campaigning would improve the UK’s electoral process. To answer this, we analysed the Leave groups’ campaigns from the 2016 Brexit Referendum. We concluded that we need an ethical code of conduct for transparent, explainable, civil and informative digital political campaigns (see Table 1). The 2019 General Election confirms this need.

Table 1. Ethical Code of Conduct for Digital Political Campaigns

TransparencyMake clear if political messages online come from a party, how much campaigners spend on digital campaigning, and on what.
ExplainabilityIn campaigns that extensively use AI to profile voters, give all voters an explanation of the profiling. 
CivilityCampaign material should be civil (e.g. not nasty, aggressive, disrespectful, or pitched to provoke anger and outrage) and must not incite others to commit crimes (e.g. making false statements of fact about candidates’ personal character or conduct). If campaigners deliberately breach civililty codes to become righteously uncivil (for moral reasons), then rationally justify why.
InformativenessCampaigns should give voters enough information to freely make informed judgments. The information provided should be true, complete, undistorted and relevant.

Unethical campaigning: the 2019 General Election

Contravening transparency. Since 2018, Google and Facebook provide publicly searchable libraries of election ads and spending on their platforms: each Facebook ad also says who paid for it. However, abuse continues. Cambridge-based Green Party activists complained that campaigning group 3rd Party Ltd managed by Vote Leave’s former chief technology officer (Thomas Borwick) was pretending to be the Green Party by buying Facebook and Instagram ads encouraging people in swing constituencies to “Vote Green”. 3rd Party Ltd also ran social media ads to “Save Brexit”. Borwick boasted that he could use proxy groups on Facebook to “split the vote” of Conservative opponents.

Contravening explainability. Facebook’s political ad library includes information on broad geographic targeting (England, Scotland, Wales, Northern Ireland), broad demographic targeting (age, gender), reach and amount spent on ads. Since February 2019, users can click on a Facebook button ‘Why am I seeing this ad?’ which tells them the brand that paid for the ad, some biographical details targeted, if and when the brand or one of their agency/developer partners uploaded the user’s contact information, and when access was shared between partners. However, there is no information on finer-grained targeting such as use of psychographics and probabilistic data inferences from metadata; or to what campaign end they were targeted (e.g. voter mobilisation, suppression).

Contravening civility. Early in the campaign, the Liberal Democrats tested 13 identical Facebook ads attacking Corbyn, with differently provocative headlines on Corbyn’s personal lack of trustworthiness, weakness and failure to stamp out hate, discrimination and anti-Semitism in his party. Accusations of anti-Semitism grew increasingly shrill, with Chief Rabbi Mirvis urging people to “vote with their conscience” as “A new poison – sanctioned from the very top – has taken root in the Labour Party” (selectively evidencing Corbyn’s past comments and associations).

Contravening informativeness. Numerous examples of deception include the Conservative Party press office assuming a false identity online, creating a parody Labour manifesto site, circulating doctored, misleading videos of the opposition, and contravening Google’s ad rules. Labour ran misleading Facebook ads on how much Labour’s policies would save the “average family”.

Transparency is Not Enough

Unethical campaigning proliferates despite large tech companies’ efforts to reduce opacity via greater transparency and explainability. Since campaigning started in October, Twitter committed to stop accepting most political ads (from 22 November onwards). On 20 November, Google said that targeting of election ads would be limited to general categories (age, gender, post code location); and advertisers would no longer be able to target political messages based on users’ interests inferred from browsing or search histories. While tech companies should continue to reduce opacity, the onus lies on political campaigners to meet the ethical demands for civility and informativeness.