Democratising software? Situating political campaigning technology in the UK’s EU referendum
Laurie Waller and David Moats
In 2016, in the wake of the UK’s EU referendum, a series of controversies emerged about the use of personal data and computational tools by political campaigns to target their messages to potential voters.1 These controversies were sparked by a scandal involving the political consultancy firm Cambridge Analytica gaining access to large volumes of social media data and using psychological profiling to advise political campaigns in both the UK and the US. In the UK, public inquiries have since been conducted by the Electoral Commission (2018a, 2018b), the Information Commissioner (2018) and the Digital, Culture, Media and Sport’s Select Committee (2019) about the harvesting of personal data from digital platforms, campaigns’ spending irregularities on social media advertising and the roles of tech companies and political consultants in online disinformation campaigns (see also Howard and Kollanyi 2016). Some commentators have argued that such technologies, and the companies that run them, potentially compromise the very basis for free and fair elections.2
In a blog post not long after the UK’s referendum result, the director of the Vote Leave campaign, Dominic Cummings, announced the ‘open-source’ release of a piece of software called VICS that, he argued, played a critical role in the campaign’s success.3 He asserted that the campaign’s management software enabled Vote Leave to interactively link sophisticated data analysis processes with ‘on the ground’ canvassing, continuously improving and testing the predictive models used by the campaign via the incorporation of ‘live feedback’. Cummings argued that it was precisely this technological form of interactivity between the campaign and citizens that gave the campaign its authentic edge and enabled it to mobilise ‘people who usually ignore politics’.
What are we to make of such claims about the ‘democratising’ or ‘anti-democratic’ influence of technology in relation to a political event like Brexit? The editors of this volume argue that it is no longer sufficient for STS researchers to mobilise theories of democracy in analysing technological processes, without also attending to the ways in which related ideas and notions are deployed and contested in empirical settings. The events above demonstrate that, on various sides of the Brexit debate, the democratic legitimacy of the referendum is, to varying degrees, inflected through controversies about technologies, and their roles in politics. In these controversies, actors are constantly redrawing or questioning boundaries between ‘the technical’ and ‘the political’. Thus, some technologies are invested with political capacities, like increasing participation (see the chapters by Papazu and Pallett & Chilvers, this volume), while others fade into the background, as mundane, everyday aspects of political practice. Attending carefully to these empirical mobilisations of democratic ideals may be particularly important, we argue, in situations like the UK’s EU referendum, where technologies are linked to practices some deem to be anti-democratic, such as misinforming the electorate or attacking experts and the judiciary, and political rhetoric conflating popular sovereignty with isolationist and, at times, explicitly xenophobic nationalism. Indeed, in the case of the Vote Leave campaign, claims about participation and freedom seemed often to coexist with a reactionary, almost Schmittian, understanding of democracy as merely national sovereignty. Situations like Brexit, we will go on to suggest, may pose a challenge (or at least give pause) to STS approaches which might take for granted that they are on the side of democracy.
In the first part of this chapter, we consider how political campaign platforms, like the one released by Vote Leave, have been analysed by political sociology, media and communications researchers and STS researchers alike. We argue that existing analyses tend to rely on taken-for-granted notions of various democratic ideals, and we draw on STS discussions of technological controversies and ‘anti-politics’ to help avoid this trap. We illustrate how democratic ideals are deployed empirically using promotional materials for political campaign platforms. We show how these companies mobilise various notions of democracy in order to sell these software technologies, but do so in highly ambivalent, and arguably, contradictory ways.
The second part of this chapter examines how similarly ambivalent and contradictory democratic ideals emerged in Vote Leave’s software release and Cummings’ explanation of the campaign’s technological strategy, mentioned above. In the blog post accompanying the software release, Cummings makes a series of elaborate claims about technological change and politics but, as we will show, these claims also take on very particular meanings in relation to Brexit as a ‘democratic situation’ and, specifically, a parliamentary inquiry into misinformation in political campaigning.4 We suggest that the controversy over the use of technology in Vote Leave’s campaign illustrates how moves to democratise technology are not necessarily incompatible with anti-political practices.
In concluding, we discuss the significance of this case for recent STS debates about populism and democracy. Situations like Brexit, we suggest, allow us to see both how political domains are unsettled and reordered by software and data infrastructures, and the consequences of this for democratic political practice. We argue that attending to particular ‘gerrymanderings’ (Woolgar and Puwluch 1985) of the technical and the political in situations like Brexit can help specify the stakes of contestations over expertise and infrastructures for democratic politics.
Campaign software as ‘democratising’ technology?
The events introduced above have brought into view just how much technologies orient and organise activities that are typically considered routine features of political processes. For example, in The Victory Lab, Sasha Issenberg (2012) catalogues a host of political scientists, data-driven market researchers, advertising consultants and data analytic tools which are now seen as integral to winning elections. While political campaign platforms, like the one released by Vote Leave’s director, are often associated with politically mundane or technocratic practices of management and administration, in this section we examine how they also become invested with ideals about democratic politics.5
So, what are political campaign platforms? Issenberg (2012) notes that there is a long history of political campaigns using software packages, dating back at least to 1983 with the development of ‘Campaign Manager’ by John Aristotle Phillips and his brother Dean. These packages have been used to manage volunteers, synch schedules and collate information about supporters or potential voters. It is this latter feature – the ability to house extensive data on people, enabling the practice of ‘microtargeting’ – which has been at the centre of recent debates. Microtargeting is a process by which potential voters and supporters can be grouped into increasingly narrow subsets (based on psychological profiles, online purchases, cultural interests, behaviours and so on) so that political messages can be tailored to these groups. This requires that data about voters (email addresses, home addresses, demographic information) can be linked up to other types of data (personal data culled from our phones and other devices). Microtargeting dates back to the early 2000s, when the ubiquity of email and email lists offered different possibilities for engaging with supporters, although one might trace back the rationale to theories of mass communication, long critiqued by sociologists (Katz and Lazarsfeld 1955). Today, advocates of microtargeting tend to link these practices to well-worn tropes about ‘big data’ (Kitchin 2014; Burgess and Puschman 2014) in order to distinguish these campaign technology start-ups from so-called ‘traditional’ political consultants and political scientists (Ansted 2017).6,7
What does perhaps separate the current crop of software packages from their predecessors is the way in which they are often presented as highly networked ‘platforms’. While not all of these players refer to their products as platforms, most have adopted key rhetorical and technical features of these Web 2.0 entities (O’Reilley 2005). For example, most of the software specifications we looked at promoted their use of Application Programming Interfaces (APIs), which facilitate the downloading and uploading of data and streaming with other applications, as well as the integration of different software systems (Helmond 2015).8 It has frequently been highlighted that digital platforms paint themselves as a kind of public utility, like water pipes, which is provided as a service but does not overdetermine how it can be used (Helmond 2015; van Dijck 2013). It has long been argued that this neutral rhetoric conceals both the business models associated with many digital platforms and important asymmetries of access and visibility baked into their technological design (Bucher 2012; Gillespie 2010). As we highlight below, while much software is promoted as applicable to myriad activities, it is also typically coupled with consulting business models and licensing agreements; the technology is offered principally as a service rather than a discrete product.9
Much of the scholarship on campaign technologies – which comes mostly from political communications and political sociology – has debated whether forms of interactivity made possible by digital technologies necessarily facilitate the formation of publics or new modes of participation in politics (Bennett and Segerberg 2013). In the context of US political campaigns, Howard (2006) has argued that the development of what he terms ‘hypermedia’ campaigns led not to broader but rather to more narrow forms of managed interactivity between campaigns and voters. A similar version of Howard’s argument can be detected in work by Barocas (2012), which analyses micro-targeting. Where promoters of micro-targeting suggest that such techniques can enhance a campaign’s relationship with its audiences (for example, Bartlett, Birdwell and Reynolds 2014), Barocas argues that by allowing campaigns to deliver different messages to different people, they undermine the sense of a common conversation in politics. In this way, micro-targeting is said to contribute to the development of advanced (neo-)liberal forms of democracy in which political participation is reduced to a matter of individual preferences, and in which some individuals (such as swing voters) matter more for campaigns than others.
Kriess (2012, 2016) has chronicled the organisational work involved in integrating data infrastructures within US political parties. Kreiss argues that novel forms of interactivity between campaigns and voters afforded by social media and other Web 2.0 technologies are contingent on the local organisational processes of particular campaigns. In Prototype Politics, Kreiss (2016) shows how the two major US parties have developed strikingly different information infrastructures, inheriting many of these specificities from past campaigns, for example, those of Howard Dean and Barack Obama on the Democrat side. Kreiss highlights that the development of interactive networked politics in US elections can be seen as an achievement of the organising practices of campaigns as much as of software development. In other words, these developments are both social and technical.10
Both Barocas and Kreiss do an excellent job of politicising these technologies – that is, examining the possible effects of campaign management software on political processes, as well as showing the asymmetries of power that shape their development. However, while they argue that technological change has consequences for democratic politics, they do not show that it fundamentally threatens or reconfigures what counts as democracy, which remains largely assumed. Indeed, we can arguably detect certain normative stances towards democracy in such studies. For instance, Barocas proposes that voters should be treated equally, and assumes a common conversation as the aim of democratic politics. For Kreiss this involves positioning interactivity as a positive characteristic of campaigns. Such stances may very well suit the specific political cultures the authors are referencing, but in affirming a particular version of democratic politics they potentially imply that contestations over the capacities of these technologies are not properly political. Empirical sensitivity to the dynamics of technological contestation, we suggest, is therefore crucial to analysing how technologically inflected ideals about interactivity and bottom-up organising can potentially serve what some might see as anti-democratic ends – something we will deal with in the next section.
Democratising science and technology can be understood in various ways: for instance, as holding experts accountable, enhancing the power of publics over technological infrastructures and participation in design and innovation processes – in other words, bringing notions of democracy into science and technology. STS studies have long focused on showing how what counts as the capacities of technologies is always contestable (Bijker and Law 1992; MacKenzie and Wajcman 1999). They highlight how what counts as technical (as opposed to social or political) is contingent and its boundaries often strategically performed by actors in order to remove certain practices and forms of knowledge from scrutiny and contestation. Rather than accepting common sense understandings of what counts as the political, such studies have sought to extend politics to technical expertise and practices. In the analysis that follows, however, we are also interested in how technologies (and related expertise) come to be included in, or are excluded from, the domain of politics, and the extent to which the political practices engaging them are claimed to be democratic or not. In a situation like Brexit, we suggest, such moves are required since technological and political contestations are often difficult to clearly separate from one another; in the case of Vote Leave, technological development is presented by its director Cummings as the primary strategy of the political campaign.
Foucauldian and post-Foucauldian studies have long attended to instruments and techniques which are presented by proponents as facilitating democratic politics. Osborne and Rose’s (1999) influential account of the opinion poll, for instance, showed how the proliferation of methods for studying public opinion provided the basis for exchanges and interactions between social science and government in the latter half of the twentieth century. Informed by insights from critical theory, such studies have tended to be cautious about understanding methods like polling as unequivocal agents of democratisation. Osborne and Rose, in particular, highlight why methodological descriptions of polls as procedures for analysing public opinion are, alone, of limited use for analysing the social phenomena polling produces (opinionated societies) and the political practices developed to organise and mobilise them. Their account makes clear why attending empirically to the specific technologies involved in enacting public opinion (for instance, questionnaires, ballots and, in our case, software and machine learning) is critical to a social analysis of contestations like those emerging around the Vote Leave’s campaign in the Brexit referendum.
STS studies focused on politics and policy issues have widely attended to technological controversies as occasions when the boundaries between technology and society are unsettled and the political domain is reconfigured (Jasanoff 2020; Nelkin 1979). Much research in this tradition has studied novel participation procedures – often promoted by political scientists as a means by which technical disputes can be democratically resolved. STS studies have shown how, in practice, procedures like juries or consensus conferences can ‘frame’ the issues at stake and foreclose competing problem definitions and public concerns (see overview in Chilvers and Kearns 2016).11 Yet critiques of how procedures frame issues and order participation have, arguably, tended to give too much credit to the state agencies and regulatory authorities attempting to govern technological society, and too little to the substantive matters that give rise to public concerns in the first place (Marres 2007). This is a helpful reminder that we can only understand certain consequences of procedures for engaging publics (and other technologies) in relation to particular political events and issues. In some versions of STS research on politics, such questions have been elaborated in discussions about the potential value of technological controversies to enrich democratic politics. Callon, Lascoumes and Barthe (2009) argue that the relationship of technological controversy to democratic politics is poorly understood by political theories that enact demarcations between democracy as a system of governing technological societies, on the one hand, and as a set of ideals about social and technological change, on the other. In other words, questions about whether technological controversies give rise to novel political objects and forms of collectivity are therefore not well served by concepts and analytical tools developed primarily for the study of democratic states (de Vries 2007).
Many STS scholars have therefore turned to approaches that aim to more thoroughly empiricise relations between technology and political ontology (Marres 2013; Woolgar and Lezaun 2013). Marres (2013), for instance, discusses the example of eco-show homes, which are invested with notions of citizenship and environmental participation ‘by design’. She argues that it is not enough to note that these objects ‘have politics’ but that we need to empirically attend to how some technologies become politicised and explicitly invested with normative political capacities. Similarly, this approach proposes to attend to how other technologies are rendered as sub-political (Marres and Lezaun 2011) or as mundane aspects of political practice. Woolgar and Neyland, for example, discuss the work that is done when certain technologies of governance – recycling bin bags, security cameras and so on – are rendered as ‘mundane’ (Woolgar and Neyland 2013). Such sensitivities have analogues in fields such as infrastructure studies. Bowker and Star’s (1999) classic Sorting Things Out, for example, invites researchers to consider how infrastructures become invisible for some actors while remaining unavoidable for others. Such approaches thus highlight why contestations over democratic politics in technological societies may in practice centre around objects and settings that appear mundane to the institutional forms and procedures associated with the politics of government and state power.
Thus, rather than merely revealing the ‘hidden’ politics of technology, such approaches also ask how certain technologies become politicised by actors in a given situation, while others become taken for granted as background infrastructure. Andrew Barry’s (2005) distinction between politics as a commonly understood domain of social life ordered by governmental actors, and ‘the political’ – which he normatively defines in agonistic terms as a space of disagreement and contestation (see also Mouffe 2005) – is particularly pertinent to the case considered in this paper. Barry’s distinction makes clear that activities in the domain of politics can ironically have what he refers to as ‘anti-political’ effects by circumscribing possibilities for engaging in disputes and debate. This, we argue, provides a helpful guide for our analyses, because even if we need to bracket our own specific political commitments to democracy, Barry reminds us that we are broadly speaking in favour of open contestation, allowing issues to unfold and take their course – even though it is never self-evident how this is to be done. It is also particularly valuable, as we highlight below, for understanding how technological change promoted as having democratising effects can be in principle compatible with anti-political practices.
The marketing materials for some popular campaign software platforms can provide a helpful illustration of how ideals of democracy or participation can be used in ambivalent ways and potentially in the service of anti-political ends.
Having an organized community gives you an advantage. The concept of community organizing has been around for thousands of years. Moses was a community organizer. The tools of organizing should not be controlled by anyone. The organizing technology that we’re building at NationBuilder is helping people come together to support the leaders and the causes they believe in. It’s not a weapon to wield; it’s infrastructure that’s enabling democracy.12
In the above statement from NationBuilder, ‘community organising’ evokes ideas of bottom-up participation: the phrasing ‘come together to support’ proposes that agency lies with supporters (as opposed to leaders or technocratic campaign managers). In this formulation, NationBuilder is not controlling things; it is ‘helping’, positioned as merely an infrastructure. NationBuilder is tasked with increasing both the role of supporters in campaigns and also democracy in society, yet at the same time it is depoliticised, distancing the software from connotations of propaganda or control.
Similar uses of ‘engagement’ are found throughout the publicity materials, again assigning agency to supporters or ‘the community’, but in ways which also invoke engagement as a marketing term: customers are, for example, encouraged to actively promote products themselves through their networks.
Share pages tend to make excellent follow-on calls-to-action from other action pages. For instance, after a constituent completes a signup form, you may daisy-chain to a share page which encourages the constituent to encourage their online peers to sign up as well.13
Blue State Digital, which (as its name highlights) emerged from the US Democratic party, talks about ‘encouraging’ constituents. Yet it seems, as suggested by the quote above, that the ways in which supporters can engage are very much set by the platform parameters and the campaign managers (for example, signing up a friend or distributing leaflets). So, while the rhetoric often evokes more active supporters, perhaps even to the extent of making it appear that the campaign is accountable to them, the architecture of the platforms also appears equally compatible with campaigning styles long critiqued as ‘machine politics’: organising public opinion and manufacturing the consent of the governed (Lippmann 1922).
While participation and interactivity are often invoked, specific political stances and potential disagreements seem to be actively supressed. The Groundwork, a (now defunct) software company linked to Google’s Eric Schmidt, represents a particularly conspicuous example. Although associated with Hillary Clinton’s presidential campaign, the company made no mention of political parties or politics in general in its publicity, instead presenting its software as a tool for social movements, community organisers and non-profit campaigns alike.14 This flexibility is possible because of the way these platforms are typically presented as a neutral (apolitical) infrastructure (as in the quote from the NationBuilder website above), to be used in any particular political context. This assumes that technology can remain politically neutral regardless of the place, process or regime in which it is instrumentalised. As we will go on to illustrate, the limiting of politics to a mere context for technology is empirically problematic in a situation like Brexit where both constitutional matters and what counts as legitimate political practice are at stake in the controversy.
To make this even more apparent, many of these companies offer versions of their platforms which are geared toward (non-political) marketing as well. Where political advertising is today often seen as the application of marketing principles to the practices of political campaigning, several software companies also trade on specific notions of politics, prompting advertisers to run their campaign ‘like a candidate’ or promote products ‘at the speed of politics’.15 This invokes a version of politics as mobilising followers (in contrast to audiences or consumers). So, these particular presentations of the software as a democratising infrastructure could also, arguably, be seen as anti-political in Barry’s sense by flattening out radically different ways of doing political engagement and disagreement.
Marketing materials for these platforms also frequently claim to level the playing field of the political landscape in general; that is, they assert the technologies to be inherently democratising. In doing so, they often evoke notions of equal access to software common in the Free Libre and Open Source Software (FLOSS) community. For instance, by offering their software at a supposedly low price, many platforms claim not to discriminate between wealthy political campaigns (such as those backed by major US political parties) and other types of NGOs and issue-based initiatives. This is to call forth a notion of democracy as fundamentally ‘fair’, in which particular parties or candidates are not given an advantage over others. Yet alongside such egalitarian gestures, other technological divisions may be created: several firms might price their service based on the total number of subscriber email addresses being stored, but at the same time these more established organisations are more likely to receive bespoke support and ultimately shape the development of further software features.
The above brief examples highlight some of the ways in which democracy is performed as a marketing device for technology firms aiming to sell software and computational expertise to political actors. These companies trade on fuzzy ideals of democracy, in which interactivity, grassroots organising and civic participation are moral constructs unmoored from any particular instantiation or struggle. Notice how in some cases the platforms themselves are invested with the capacity to increase democracy (or install democracy) while at other times they are made to fade into the background as a neutral infrastructure for politics. It’s not hard to see how such deployments of democracy could be made compatible with anti-political practices, delimiting politics as simply a market segment and context for technology deployment.
In the next section, we discuss in detail the campaign software used by Vote Leave (called VICS) during the UK’s EU referendum, critically attending to how similar-sounding claims about the democratising influence of technology take on a distinctively anti-political significance in relation to debates about the legitimacy of the referendum process and Brexit.
Vote Leave’s (anti-)political software release
The UK’s EU referendum arguably unsettled many taken for granted ideas about the role political campaigns play in processes like referendums. With its high voter turnout, the referendum is widely considered to have been an occasion in which new sorts of actors and groups participated, outside traditional partisan lines (Davies 2016). Although the referendum may have increased participation, debate still rages about its effects on, and consequences for, democratic politics in the UK. We thus propose to approach Brexit as a ‘democratic situation’ in the sense proposed by the editors of this volume (Birkbak and Papazu 2022).
The primary material we use to analyse this democratic situation is the blog post by Dominic Cummings16 (Director of the Vote Leave campaign) which accompanied the open-source release of Vote Leave’s campaigning software, as well as correspondence with campaign personnel and details of the software itself, which we also attempted to set up ourselves (see p. 241, below). The software release coincided with the aforementioned investigations into Vote Leave’s campaigning practices by regulators, specifically the Digital, Culture, Media and Sport Committee (2019) inquiry on disinformation during the referendum. This latter inquiry (which concluded several years later) explicitly highlighted the lack of transparency around Vote Leave’s campaigning practices and judged the campaign, in addition to its violating laws relating to campaign spending, to have been complicit in the data protection violations of the consultants it employed. The inquiry’s final report notes that Cummings refused to give oral evidence to the inquiry and ignored a formal order requiring him to appear before the committee. Against this backdrop, Cummings’ software release could be seen as an attempt to stage a counter-demonstration of transparency, purporting to lift the hood on the inner workings of Vote Leave’s campaign and contradict criticisms, later formalised by the parliamentary inquiry, that the campaign deliberately misinformed the electorate.
What we are interested in here is the way notions of the democratising influence of technology, and arguments about the social construction of expertise, can (potentially at least) be made compatible with what others claim are anti-political practices. Cummings’ account draws on many of the same tropes about the democratising influence of technology as the software firms discussed above but, as we show here, these have very particular consequences in a situation like Brexit.
In the blogpost, which was released shortly after the referendum result, Cummings announced the open-source release of a piece of software, called the Voter Intention Collection System (VICS), that he claimed enabled an interactive approach to campaigning and was able to mobilise the ‘silent majority’ of eurosceptic voters. Crucially, Cummings argued, the software allowed the Vote Leave campaign to develop data analytics and test predictive models that would inform, in real-time, the campaign’s social media messaging approach.
In a text laced with popular anti-establishment rhetoric, Cummings presented Vote Leave’s use of data science expertise as an attack on an out of touch political class and the ‘traditional’ expertise of its political machinery:
This included a) integrating data from social media, online advertising, websites, apps, canvassing, direct mail, polls, online fundraising, activist feedback, and some new things we tried such as a new way to do polling … and b) having experts in physics and machine learning do proper data science in the way only they can – i.e. far beyond the normal skills applied in political campaigns (Cummings 2016).
In a broadside against metropolitan pollsters, economic forecasters and marketing specialists, Cummings proposed that campaigns no longer need to rely on traditional techniques of political strategy and should instead ‘hire physicists’ to mine data. Extolling the novelty of VICS, he stated that:
Amazingly there was essentially no web-based canvassing software system for the UK that allowed live use and live monitoring. There have been many attempts by political parties and others to build such systems. All failed, expensively and often disastrously (Cummings 2016).
Cumming’s claims about the ascendance of data science and the decline of what is positioned as ‘traditional’ polling or campaign tactics mirrors much of the positioning we find in the campaign software publicity, discussed in the previous section. Yet, when we approached the data scientists working with Cummings for an interview, they declined, saying their firm did not ‘talk about the polling it does’. This may be taken to suggest that the analysts themselves do not see such a clear separation between polling and data analytics. Similarly, a blog by an ex-Labour staffer who set up the software suggests many commonalities with the technologies used by conventional political party bureaucracies. He notes: ‘For all talk about how radical and new this software is, this is worse than other parties’ alternatives’.17
Cummings’ claims to inject science into politics, demarcating a new data science from the old polling, have broader connotations in relation to the referendum process. While such invocations of science-led politics are in many senses simply technocratic, they are clearly intended by Cummings to be continuous with Vote Leave’s broader attack on the EU. In one of the most publicised moments of the campaign, the leader of Vote Leave, MP Michael Gove, announced on prime-time television that ‘the British people have had enough of experts’. Conflations between policy experts and political elites have subsequently played a central role in populist stagings of Brexit as a victory for the people against an out-of-touch political class. In this sense, the drive to replace political scientists with data scientists is not only an issue related to knowledge about voter behaviour but also invokes images of the political elite that were the target of the Leave campaign’s anti-EU rhetoric. While we are accustomed in STS to showing the politics of different ways of constructing knowledge claims, we can see here how some actors might leverage similar arguments to reduce such contestation to simplistic oppositions – a hallmark of populist thought.
Not unlike the marketing materials surveyed above, Cummings adopts many tropes relating to grassroots mobilisation, emphasising the role of Vote Leave in mobilising ‘people who usually ignore politics’ against ‘the Government machine supported by almost every organisation with power and money’. Such populist political rhetoric also appears to inform the ways that other, more technically literate, actors involved in the campaign understood the VICS software. One of its developers, for instance, described VICS in the following way:
People on the ground would … gather data to enrich the model in an iterative process, so the model improved organically over time as more data came in from the ground team. VICS facilitated the whole thing – highlighting [sic] key geographical areas to volunteers and telling them which streets to go to (based on the evolving model) (Personal communication 16 December 2016).
In this account, the software is presented as enhancing interactivity between campaign managers and volunteers via improved feedback. However, reading through the software code, we found a pdf guide designed for activists on the ground canvassing for the campaign. The data gathering process appears distinctly inflexible and closed to a range of bottom-up forms of political interaction. Much as with the other software, individual views were sought only on predefined issues prioritised by the directors of the campaign;18 what counts as interactions appeared delimited by the management.
It is also worthwhile interrogating similar claims about how VICS promoted interactivity between Vote Leave and its supporters. Cummings deploys many tropes of grassroots political activism, but it is also clear that part of Vote Leave’s strategy was to discover potential supporters. Indeed, in the blog post Cummings outlines one strategy the campaign developed for discovering potential Leave voters which involved a seemingly unrelated competition in which players were asked to predict sporting results. This allowed Vote Leave to gather data about a specific type of participant (sports fans) who they believed traditionally ignored politics and used this data to train their models. Despite the implication that data analytics allows campaigns to expand what counts as politically active people, financially incentivising groups to participate passively, or unknowingly, in campaigns implies a distinctly technocratic and paternalistic view of political mobilisation. Vote Leave’s analytics models might be responsive but gathering data about people surreptitiously hardly seems interactive. While we can only speculate on the inner workings of the campaign, our point is that these claims about the politics of science and expertise take on distinctively populist meanings in relation to Vote Leave’s claims to be wrestling back democracy from a technocratic elite.
Much like the marketing materials of other platforms, Cummings’ account of VICS paints the software as ‘levelling the playing field’ for diverse political movements. In announcing the release of Vote Leave’s software, Cummings referred to some basic tenets of the open-source movement, stating that its release was ‘…strictly on the basis that nobody can claim any intellectual property rights over it’. However, as with the other platforms, open-source software can be made available in different sorts of ways; the openness of software is in practice shaped by its ongoing maintenance as much as its legal status (Kelty 2008).
We attempted to set up VICS ourselves.19 However, in its open-source form, the software is missing a key module and in its released form doesn’t perform its basic functionality. According to one of the developers we spoke to: ‘VICS depends on a [sic] upstream application which we called the “Voter API”’. This API sits in front of a database which contains information about the voters, and some aggregated measures (scores) to enrich the data and is essential for allowing the platform to communicate with the database. Without it the software is effectively an empty shell. By announcing its open-source release as a contribution to democratising the field of campaigning technology, Cummings seems to draw on a similar understanding of democratising as ‘fair’ used by firms like NationBuilder. However, there is no obvious community of coders contributing to or forking the code, and the ReadMe begins with the disclaimer ‘This project is no longer maintained’. The incomplete VICS code and lack of engagement with its release suggests that Cummings’ claims to be open-sourcing Vote Leave’s technology fall considerably short of open-source conventions.
What is interesting about Cummings’ blog post accompanying the software release is that it could be said to perform moves similar to those that many STS scholars have been seen to make in the past. Cummings is claiming that Vote Leave’s approach was democratising; releasing VICS software, he claims, demonstrates not only that Vote Leave had the technology to understand the electorate better than its rivals, but that the software enabled Vote Leave to upend the traditional hierarchical way of involving experts in campaigns and make the campaigning process interactive and responsive. By releasing its software open-source, Cummings suggested, Vote Leave’s campaign developed a technology that could expand participation, to open the door to novel interactions between a new breed of experts and the voters marginalised from existing political processes. While we do not think that STS researchers would ever fall into such blatant technological-determinism, STS notions associated with democratising science and technology, such as ‘public engagement’ or ‘counter-expertise’, are clearly not immune to populist variants of political thought.
As our analysis suggests, there are many reasons to be sceptical regarding Cummings’ claims about the software, and against the backdrop of various inquiries into Vote Leave, we have suggested that the blog could also be read as a superficial attempt to demonstrate transparency and to justify campaigning practices that critics claimed deliberately misinformed the electorate. Indeed, based on the materials we have, Vote Leave’s approach appears to have been more concerned with fitting voters into pre-defined framings of issues than with articulating their specific concerns. However, even though we are sceptical about the substantive claims Cummings makes, his text should still be sobering for an STS audience because it (rhetorically) claims to be enacting a democratising move on campaigning technology – in both senses of making technology available to more people and making it political. While we might disagree with Cummings’ politics or position on the EU, we still might concede that his account draws attention to the importance of technology in the political arena and the ways in which it is often bracketed as a mundane aspect of political processes.
Discussion and conclusions
Brexit is a situation that has brought into centre stage the kinds of relations between technology and politics which STS scholars have long raised, but it has also provoked questions about what makes specific political forms, like campaigns, democratic. Much of the debate about the referendum has focused on misinformation on social media, and in STS this has been discussed in the context of concepts like ‘post-truth’ and a populist rejection of expertise (Jasanoff and Simmet 2017; Sismondo 2017). Some (e.g., Fuller 2016, cited in Sismondo 2017) have raised questions about constructivist approaches to knowledge in STS, given that all sorts of actors in politics now seem to be using constructivist arguments for partisan political gain (see also Latour 2004). Such debates highlight that STS analysis of politics cannot be entirely agnostic when it comes to matters of epistemology in deciding what counts as a knowledge claim and what is simply myth, rumour, slander or lying. However, the focus on prominent scandals around firms like Cambridge Analytica has tended to bring only certain technologies into the foreground of these debates, while rendering other technologies as merely the background infrastructure of political practice.
Our analysis of Vote Leave’s software release has shown that online misinformation is not only (and not even primarily) a matter of epistemology (how we know facts from falsehoods) but an issue that emerges alongside attempts to technologically innovate traditional formats of political practice. Vote Leave may well have used social media adverts to peddle myths about the EU, although in many respects this could be seen as simply a longstanding political campaigning technique deployed in a novel medium. What is more interesting for STS, we suggest, is the way in which an actor like Cummings deploys a gesture of open sourcing the campaign’s software as a means of justifying such political practices as democratic. Claims made by Cummings about how technological innovation made the campaign interactive and responsive to voter concerns appear hard to substantiate from the software release. Yet the open-sourcing move by Cummings makes clear that technology lies at the heart of Vote Leave’s politics: the campaign’s relationship to voters, the referendum process, government and the EU are all, in Cummings’ account, refracted through ideas about relations between science, technology and democracy, albeit that these may sometimes be considered contradictory. This case, we suggest, demonstrates why the political aims of democratising science and technology cannot be taken for granted, and why attending to the situations in which science and technology become invested with political capacities is crucial.
We have argued that, in analysing the role played by campaigning technologies in the referendum process, it is important to look at how actors draw boundaries between the technical and the political. It is not just important to question the tendencies of actors to fetishise technology and reduce democracy to a mere ideal, but also to attend closely to the political practices that such notions are deployed to justify. For example, Vote Leave’s open-source software release claims to demonstrate transparency, yet this gesture towards technological openness (which we don’t find particularly convincing) coincides with Cummings’ refusal to give evidence before a parliamentary inquiry into misinformation during the referendum. One way to distinguish what Cummings is doing from our STS-informed position is through the concept of ‘ontological gerrymandering’ (Woolgar and Pawluch 1985), to bring a political term, appropriated by STS, back into the political arena. In STS, this term was used to describe how, in literature on social problems, an arbitrary division was created between ‘social’ factors and other, ‘natural’ or ‘technical’ factors. Similarly, in the case we have presented here, certain technologies are rendered as political, but not others, and this division affects claims about what counts as legitimate (or illegitimate) political practice. What this means for the present analysis is that STS scholars should not find themselves confronted with a choice between advocating for the democratising of technology or not; they should instead take a position on particular gerrymanderings of technology and politics. While in other situations, we might be able to get behind a critique of experts, or advocate for the open-source release of proprietary software, we have shown that Cummings’ use of both of these moves – the staging of technology as a site of politics – is here compatible with a form of anti-political action.
We reserve a few final words for the substantive matter of Brexit. While the European Union has been widely understood as a primarily technocratic endeavour that suffers from democratic deficits, it has arguably also occasioned novel forms of politics (see Ehrenstein, this volume). As Barry’s (2001) study of the EU highlighted, an analysis of the politics of European integration requires attending to technological controversies, from bathing water to air quality standards, and the contestations that emerge around practices like demonstration or testing that take place beyond the traditional sites of government and administration. The EU may be built on political talk about technological innovation and networks, but in practice the political contestations that shape it, Barry argues, are far more situated around sites where measurements are made or infrastructure is being developed. Such a view of EU politics was largely bracketed by the main protagonists in the Brexit referendum. As far as we can see, few attempts were made by any campaign to connect the sociotechnical issues underpinning EU politics to referendum debates about democracy. There was a missed opportunity, in other words, to connect public concerns about EU politics to issues (like air pollution) that shape the technological infrastructures and environments of everyday social life and have material consequences for democracy in the UK. Instead, somewhat ironically given the campaign’s populist rhetoric of ‘taking back control’, the case of Vote Leave illustrates precisely why attempts to technologically innovate traditional forms of politics, like campaigns, are likely to simply reinstate expert-centric forms of political practice that may have little to do with addressing the democratic deficits between governing authorities and the publics they claim to represent.
1 See, for example: https://www.theguardian.com/technology/2017/may/07/the-great-british-brexit-robbery-hijacked-democracy (accessed 30 Jun 2018).
2 For example, the investigative journalist Carole Cadwalladr. See: https://www.ted.com/talks/carole_cadwalladr_facebook_s_role_in_brexit_and_the_threat_to_democracy#t-903238 (accessed 08 May 2020).
3 The software was released on the open-source platform GitHub and can be accessed here: https://github.com/celestial-winter/vics/ (accessed 02 May 2018)
4 We use the term ‘misinformation’ in a broader sense than is defined in the parliamentary inquiry (Digital, Culture, Media and Sport’s Select Committee, 2019), which distinguishes between ‘disinformation’ and ‘misinformation’ as, respectively, the intentional or unintentional propagation of ‘false information’. For our purposes the notion of ‘false information’ is unhelpful because it suggests that misinformation can be easily separated from contestations over expertise and the knowledge infrastructures of political campaigns.
5 It is notable that to date we have come across less than a handful of political commentaries written about Vote Leave’s software. A blog by the BBC’s Laura Kuenssberg, for instance, largely repeats the claims made on Cummings’ blog, albeit with a few qualifying ‘maybes’. See: https://www.bbc.co.uk/news/uk-politics-37841605 (accessed 30 Jun 2018).
6 The 2016 US election saw a series of particularly strong attacks on traditional polling methods and the use of aggregate pollsters like Nate Silver to predict election results (Loukissas and Pollock 2017). There is also a range of longstanding problems recognised by the polling industry, including low response rates, tensions between representative phone and unrepresentative internet polling, and public distrust of pollsters.
7 Such distinctions between ‘new’ data science expertise and ‘traditional’ political science techniques like polling, presented as a fait accompli, are harder to substantiate in terms of their applications to campaign management. Many of the tools we looked at centre on the not-so-ground-breaking technique of A/B testing, a longstanding practice within market research, used to test and trial messages and advertisements but also assign ‘responsiveness scores’ to contacts in the database.
8 Making them a ‘type 1’ API (Helmond 2015).
9 In its basic premise, ‘software as a service’ is not unlike the dominant business models of the software industry, which are widely organised around licensing agreements rather than individual ownership.
10 In the UK context, Anstead (2017) has similarly interviewed politicians and campaign managers on the effects of data driven campaigns in the 2015 general election.
11 Such studies, importantly, make clear why participation should not be conflated with democratisation.
12 Available at: https://nationbuilder.com/myths (accessed 30 April 2020)
13 Available at: https://tools.bluestatedigital.com/kb/article/what-is-a-share-page (accessed 20 April 2020)
14 Available at: https://www.thegroundwork.com/ (accessed September 2017, site no longer accessible)
15 Available at: https://nationbuilder.com/allsaints (accessed 30 April 2020)
16 Since we started working on this topic in 2016, Cummings has gone from being a little-known campaign director to the UK Prime Minister’s Chief Adviser, a position that he left during the Covid pandemic. He is now a polemical media figure who is both credited as a mastermind strategist behind Brexit and caricatured as a Silicon Valley-inflected Rasputin in equal measure.
17 Available at: https://twitter.com/jshmrtncrrngtn/status/793845659946266624 (accessed 10 May 2021).
18 This point is also raised by the ex-Labour staffer introduced above, who wrote up his blog in an article for the New Statesman (Carrington 2016)
19 We set up the software on a virtual machine, using the Ubuntu 16.04 operating system with some help from our colleague Marcus Burkhardt.
We are grateful for the comments and feedback provided by participants at the authors’ workshop convened by the editors in November 2017, as well as those of the editors themselves and the anonymous reviewers.
Anstead, N., ‘Data-driven Campaigning in the 2015 United Kingdom General Election’, The International Journal of Press/Politics, 22 (2017), 294–313.
Barocas, S., ‘The Price of Precision: Voter Microtargeting and its Potential Harms to the Democratic Process’, Proceedings of the First Edition Workshop on Politics, Elections and Data, (2012), pp. 31–36.
Barry, A., Political Machines: Governing a Technological Society, (London: The Athlone Press, 2001).
Bartlett, J., J. Birdwell, and L. Reynolds, Like, Vote, Share, (Demos, 2014), <https://www.demos.co.uk/files/Like_Share_Vote_-_web.pdf?1415749150> [accessed 3 July 2018].
Bennett, W. L., and A. Segerberg, The Logic of Connective Action: Digital Media and the Personalization of Contentious Politics (New York: Cambridge University Press, 2013).
Birkbak, A., and I. Papazu, ‘Introducing Democratic Situations’ in A. Birkbak, and I. Papazu, eds, Democratic Situations, (Manchester: Mattering Press, 2022) [this volume]
Bijker, W. E., and J. Law, Shaping Technology/Building Society: Studies in Sociotechnical Change, (Cambridge, MA: The MIT Press, 1992).
Bowker, G., and S. L. Star, Sorting Things Out, (Cambridge, MA: The MIT Press, 1999).
Bucher, T., ‘Want to Be on the Top? Algorithmic Power and the Threat of Invisibility on Facebook’. New Media and Society, 14 (2012), 1164–1180.
Callon, M., P. Lascoumes, and Y. Barthe, Acting in An Uncertain World: An Essay on Technical Democracy (Cambridge, MA: The MIT Press, 2009).
Chilvers, J., and M. Kearnes, Remaking Participation: Science, Environment and Emergent Publics (London: Routledge, 2016).
Carrington, J., ‘The Vote Leave Campaign Wasn’t as Clever as It Thinks It Was’, The New Statesman, 10 July 2016, <https://www.newstatesman.com/politics/staggers/2016/11/vote-leave-campaign-wasn-t-clever-it-thinks-it-was> [accessed 3 July 2018].
Cummings, D., ‘On the Referendum #20: The Campaign, Physics and Data Science – Vote Leave’s ‘Voter Intention Collection System’ (VICS) Now Available For All. In: Dominic Cummings’s Blog, 20 October 2016, <https://dominiccummings.com/2016/10/29/on-the-referendum-20-the-campaign-physics-and-data-science-vote-leaves-voter-intention-collection-system-vics-now-available-for-all/> [accessed 30 June 2018].
Davies, W., ‘Thoughts on the Sociology of Brexit’, in Verso Books, eds, The Brexit Crisis: A Verso Report. (London: Verso, 2016), <https://www.versobooks.com/books/2352-the-brexit-crisis> [accessed 30 June 2018].
De Vries, G., ‘What is Political in Sub-Politics? How Aristotle Might Help STS’, Social Studies of Science 37, (2007), 781–809.
Digital, Culture, Media and Sport Committee., Disinformation and ‘Fake News’: Final Report. (House of Commons, 2019), <https://publications.parliament.uk/pa/cm201719/cmselect/cmcumeds/1791/1791.pdf> [accessed 1 November 2019].
Ehrenstein, V., ‘Technocratic Activism: Environmental Organisations, Carbon Markets and European Bureaucracy’ in A. Birkbak, and I. Papazu, eds, Democratic Situations, (Manchester: Mattering Press, 2022) [this volume]
Gillespie, T., ‘The Politics of “Platforms”’, New Media & Society, 12 (2010), 347–364.
Helmond, A., The Web as Platform: Data Flows in Social Media (PhD Thesis, University of Amsterdam, 2015).
Howard, P. N., New Media Campaigns and the Managed Citizen (Cambridge: Cambridge University Press, 2006).
Howard, P. N., and B. Kollanyi, ‘Bots,# Strongerin, and# Brexit: Computational Propaganda During the UK-EU Referendum’, (2016), <https://ssrn.com/abstract=2798311> [accessed 30 June 2018]
Information Commissioners Office, ‘Investigation into the Use of Data Analytics in Political Campaigns: A Report to Parliament’, (2018) <https://ico.org.uk/media/action-weve-taken/2260271/investigation-into-the-use-of-data-analytics-in-political-campaigns-final-20181105.pdf> [accessed 1 November 2019]
Issenberg, S., The Victory Lab: The Secret Science of Winning Campaigns (New York: Broadway Books, 2012).
Jasanoff, S., ‘Science and Democracy’, in U. Felt, R. Fouché, C. A. Miller, and L. Smith-Doerr, eds, The Handbook of Science and Technology Studies, (Cambridge, MA: The MIT Press, 2020), 259–288.
Jasanoff, S., and H. R. Simmet, ‘No Funeral Bells: Public Reason in a “Post-Truth” Age’, Social Studies of Science, 47 (2017), 751–770.
Katz, E., and P. Lazarsfeld, Personal Influence: The Part Played by People in the Flow of Mass Communications, (Columbia University: Bureau of Applied Social, 1955; repr. Taylor & Francis Group, 2005).
Kelty, C. M., Two Bits: The Cultural Significance of Free Software (Durham, NC: Duke University Press, 2008).
Kitchin, R., The Data Revolution: Big Data, Open Data, Data Infrastructures and Their Consequences. (London: Sage, 2014).
Kreiss, D., Taking Our Country Back: The Crafting of Networked Politics from Howard Dean to Barack Obama (Oxford: Oxford University Press, 2012).
——, Prototype Politics: Technology-intensive Campaigning and the Data of Democracy (Oxford: Oxford University Press, 2016).
Latour, B., ‘Why Has Critique Run out of Steam: From Matters of Fact to Matters of Concern’, Critical Inquiry, 30 (2004), 225–248.
Lippmann, W., Public Opinion (New York: Harcourt, Brace and Company, 1922; repr. Jefferson Publication, 2015).
Loukissas, Y., and A. Pollock, ‘After Big Data Failed: The Enduring Allure of Numbers in the Wake of the 2016 US Election’, Engaging Science, Technology, and Society, 3 (2017), 16–20.
MacKenzie, D., and J. Wajcman, The Social Shaping of Technology: How the Refrigerator Got Its Hum (Milton Keynes: Open University Press, 1985).
Marres, N., ‘The Issues Deserve More Credit: Pragmatist Contributions to the Study of Public Involvement in Controversy’, Social Studies of Science, 37 (2007), 759–780.
——, ‘Why Political Ontology Must be Experimentalized: On Eco-Show Homes as Devices of Participation’, Social Studies of Science, 43 (2013), 417–443.
Marres, N., and J. Lezaun, ‘Materials and Devices of the Public: An Introduction’, Economy and Society, 40 (2011), 489–509.
Mouffe, C., On the Political (London: Routledge, 2005).
Nelkin, D. E., Controversy: Politics of Technical Decisions (London: Sage Publications, 1979).
O’Reilley, T., ‘What is Web 2.0: Design Patterns and Business Models for the Next Generation of Software’, (2005) <http://oreilly.com/ web2/archive/what-is-web-20.html> (Accessed 11 May 2020).
Osborne, T., and N. Rose, ‘Do the Social Sciences Create Phenomena? The Example of Public Opinion Research’, The British Journal of Sociology, 50 (1999), 367–396.
Puschman, C., and J. Burgess, ‘Metaphors of Big Data’, International Journal of Communication, 8 (2014), 1690–1709.
Sismondo, S., ‘Post-Truth?’, Social Studies of Science, 47 (2017), 3–6.
The Electoral Commission., Investigation: Vote Leave Ltd, Mr Darren Grimes, BeLeave and Veterans for Britain, (2018a) <https://www.electoralcommission.org.uk/who-we-are-and-what-we-do/our-enforcement-work/investigations/investigation-vote-leave-ltd-mr-darren-grimes-beleave-and-veterans-britain> [accessed 1 November 2019]
——, Investigation: Leave.EU Group Limited, (2018b) <https://www.electoralcommission.org.uk/who-we-are-and-what-we-do/our-enforcement-work/investigations/investigation-leaveeu-group-limited> [accessed 1 November 2019]
van Dijck, J., The Culture of Connectivity: A Critical History of Social Media, (Oxford: Oxford University Press, 2013).
Woolgar, S., and J. Lezaun, ‘The Wrong Bin Bag: A Turn to Ontology in Science and Technology Studies?’, Social Studies of Science, 43 (2013), 321–340.
Woolgar, S., and D. Neyland, Mundane Governance: Ontology and Accountability, (Oxford: Open University Press, 2013).
Woolgar, S., and D. Pawluch, ‘Ontological Gerrymandering: The Anatomy of Social Problems Explanations’, Social Problems, 32 (1985), 214–227.