Data, Democracy and Discourse: A study about the security threats of the use of political microtargeting in political campaigns in the United Kingdom

(1)

1

Institute of Security and Global Affairs

Leiden University – Faculty of Governance and Global Affairs

Data, Democracy and Discourse

A study about the security threats of the use of political

micro-targeting in political campaigns in the United Kingdom

Written by: Joris Roelofs Student number: s1381970

Supervised by: Dr. Bibi van den Berg and Ms. Ilina Georgieva Program: Crisis and Security Management (MSc)

Word count: 23.913 Date: 9-6-2019

(2)

2

Acknowledgements

I would like to sincerely thank my supervisors Dr. Bibi van den Berg and Ms. Ilina Georgieva for providing me with feedback on several occasions and showing great patience in guiding me in the right direction.

Furthermore, I would like to thank my family and friends for supporting me during the thesis process, in particular my mother Letty de Graaf and my father Michiel Roelofs.

Last but not least, I am grateful for the music of Jimi Hendrix, who's following quote from the song ‘The wind cries Mary’ got me through some of the harder moments in the writing

process:

(3)

3

Chapter 1 - Introduction

1.1. Political micro-targeting

The freedom to choose leaders in a democratic election is the most important freedom there is, according to French president Emmanuel Macron (2019, p. 22). Political actors engage in political campaigns to compete for the votes of the electorate. Political actors are eager to use more efficient and effective strategies to gather more votes and outsmart their opponents in elections (Dobber et al., 2017, p. 1). One of these new strategies is political micro-targeting that originated in the corporate sector due to technological innovations in data analytics.

Political micro-targeting is the “ability of campaigns to contact the right voters with the right issues” (Endres, 2016, p. 771). The ‘micro’ part indicates that political actors aim their campaign communication to the level of individuals or small segments of the electorate. The communication they receive is ‘targeted’ and tailored to their political attitudes based on data analysis. Jim Messina, the chief campaign manager of the 2012 Obama campaign, explains political micro-targeting in an interview for an American audience. The campaign had the “singular goal to run a personalized campaign, where you got a different campaign than you did, all based on our ability to move you and persuade you to vote for Barack Obama” (Milken Institute, 2013). The presidential campaign of Barack Obama in 2012 was arguably the first campaign that adopted political micro-targeting on a large scale (Bimber, 2014, pp. 140-141).

Not only political actors in the U.S., but also political actors in other democratic states such as Germany, the United Kingdom (UK) and the Netherlands adopted political micro-targeting in their campaigns as well (Bennett, 2015, p. 371). The actual effect of political micro-targeting on voting behaviour is unknown, as it is hard, if not impossible to measure (Bennett, 2015, p. 375). However, political actors believe that this strategic way of

communication is more efficient and effective than other communication methods (Rubinstein, 2014, p. 884). Since there are certain parts of the electorate that are unlikely to be susceptible to a campaign message (e.g. members of a different political party), communicating to this group of voters is a waste of resources (Nickerson & Rogers, 2014, p. 53). Political micro-targeting provides for better-informed communication efforts to voters that could be persuadable. Political campaign organizations usually have limited resources, which they must allocate wisely (Kreiss & Jasinski, 2016, p. 559). Therefore, it is likely that political

(6)

6

actors will continue to exploit the strategy of micro-targeting (Nickerson & Rogers, 2014, p. 53).

1.2. Security threats of political micro-targeting

Although political micro-targeting might be an efficient strategy for political actors to communicate in their campaigns, there are threats associated with political micro-targeting that could harm democratic values. Popular journalistic accounts have described campaign organizations that use political micro-targeting as all-knowing Big-Brother-like organizations, because of the large amounts of data that the campaign organizations gather and analyse (Brennan, 2012; Grassegger & Krogerus, 2017). Several academics identify multiple threats as well that relate to two fundamental rights of people in democracies. These are the right to free and fair elections and the right to a private sphere.

First, the use of political micro-targeting might threaten the free- and fairness of elections. Free and fair elections are elections where voters can inform themselves freely about political issues and candidates and where their freedom of choice is free from manipulation (European Commission, 2018, p. 1). There are academics that argue that political actors manipulate the freedom of choice by using political micro-targeting (Kruschinski & Haller, 2017, p. 17). Manipulation could happen because a voter could increasingly receive political advertisements adjusted to the specific political attitudes of that individual and are different from the messages other voters receive. Political micro-targeting raises the issue of transparency as well, since it is an opaque technique (Rubinstein, 2014, p. 897). Voters do not know they are seeing selective and tailored information from political actors, which might affect how they estimate that information. Furthermore, political actors are supposed to be transparent in a democracy, which also applies to elections (Rubinstein, 2014, p. 897). The perceived effectiveness of political micro-targeting provides an incentive for political actors to increasingly advertise on social media where advertisements are not public. This limits the transparency of political campaigns (Rubinstein, 2014, p. 897).

Furthermore, the narrowcasting of information to small groups of voters by political campaigns contributes to filter bubbles, according to Pariser (2011, p. 85). This could harm the freedom of elections, since people will increasingly think and receive political information in small, like-minded groups (In ‘t Veld, 2017, p. 4). Therefore, people might not know about all the political options available, but only those that their filter bubble presents. Finally, the goal of political micro-targeting is to communicate messages only to those people that might

(7)

7

be persuadable (Bodó et al., 2017, p. 4). This could mean that political actors leave some groups of voters out of the targeted campaign communication, because political actor perceive them as unpersuadable. Political redlining is the term to describe this issue (Endres & Kelly, 2018, p. 14).

Secondly, the use of political micro-targeting might threaten the right to a private sphere. The use of political micro-targeting requires “the gathering and combining personal data about people on a massive scale” (Zuiderveen-Borgesius et al, 2018, p. 87). This creates the risk that campaign organizations gather personal data without people giving explicit consent for that purpose (Bennett, 2015, p. 262). People must give consent for the processing of personal data. If people know that political actors watch them, they might handle

differently. Moreover, the issue of transparency listed two paragraphs before, also limits privacy rights (Rubinstein, 2014, p. 893). Voters are unable to exercise their rights over their data, when they do not know that political parties gather personal data about them.

Furthermore, the increased gathering of personal data by political actors could result in a system of voter surveillance, where the right to privacy is structurally undermined (Barocas, 2012, p. 34).

These related issues illustrate that political micro-targeting might threaten the right to privacy and the right to free and fair elections, which are basic democratic values. As such, political micro-targeting might form a security issue for societies based on democratic values. The term ‘security’ means, “a low probability of damage to acquired values” (Baldwin, 1997, p. 13). Traditionally, the referent object of security would entail only the survival of the state. However, other referent objects in other domains are also possible since the work of the Copenhagen School of security (Peoples & Vaughan-Williams, 2015, p. 97). Security is about different issues in different contexts.

Therefore, this thesis considers anything that has a high probability to damage these democratic values as a security issue. The security logic in relation to the mentioned threats, regards political micro-targeting as a security threat if fundamental democratic values are under pressure. A security logic entails government actors to enact policies more quickly and with fewer checks than other legislation (Baldwin, 1997, p. 20). One should always be critical and attentive when government or academic actors frame issues as security issues.

(8)

8

1.3. Academic puzzle and research question

By contrast, there are also academics who argue that the mentioned threats against democratic values are overstatements. In their view, the emphasis is too much on the ‘all-knowing’ political actors that use information about the electorate to manipulate their votes (Kreiss, 2017, p. 2). Kreiss posits that political actors have a hard time in directing their communication efforts during political campaigns. Hersh (2015) argues in a similar vein. He contends that political actors are more likely to act by chance in their campaign

communication, than that they precisely know which voter is susceptible to what message through sophisticated data analyses (Hersh, 2015, p. 197). What is more, Lynch et al. state that some accounts overstated the role of data analytics companies in elections, because these companies themselves overestimate their influence (Lynch et al., 2018).

These academics argue that the identified security threats about the use of political micro-targeting in political campaigns are overstatements to some extent. This presents a puzzle, which is essentially about the subjective dimension of security (Baldwin, 1997, p. 14). Both accounts seem convincing on first sight, so it is unclear what the probability of damage to democratic values is by the use of political micro-targeting in campaigns. Moreover, the associated threats of political micro-targeting might relate to campaigns in a certain political context (Bennett, 2016, p. 274). Most of the academic literature on political micro-targeting is about campaigns in the United States, which is where political micro-targeting originated. The critique from Hersh, Kreiss and Baldwin-Philippi is mainly about the U.S. as well. The U.S., however, have an exceptional political and cultural context (Zuiderveen-Borgesius, 2017, p. 89). For example, the U.S. have very limited privacy laws (Kruschinski & Haller, 2017, p. 2). Such a political context matters for the use of political micro-targeting and thus also for the threats associated with political micro-targeting (Dobber et al., 2017, p. 2).

There are some accounts about political micro-targeting in other states, such as a study into the 2015 General Election in the UK(Anstead, 2017, p. 295). Moreover, the UK is a promising case, since there was widespread public attention of the use of political micro-targeting in political campaigns in the aftermath of the UK EU referendum (Shipman, 2016, p. 441). A whistle-blower revealed documents from the data-analytics company Cambridge Analytica, which allegedly played a role in the EU Referendum providing micro-targeting messages (Cadwalladr, 2019). Subsequently, there have been several reports from different government organizations that address the use of political micro-targeting in the campaigns in the EU Referendum mostly, and to some extent in the elections afterwards. Therefore, this

(9)

9

case seems like a crucial case to study the use of political micro-targeting in political

campaigns as a potential security threat. The main aim of this study is to identify what threats against democratic values are associated with the use of political micro-targeting by

government actors for the UK EU Referendum and thereby contributing to the academic puzzle. The main question in this thesis is:

To what extent can the use of political micro-targeting in political campaigns in the UK be framed as a security threat?

This question examines the underlying security discourse in government reports. Studying a security discourse is relevant, because the framing of a security issue would make resources available to address problems that would not have been available otherwise (Zedner, 2003, p. 162). There are several academic studies and reports that mention one or several of such threats on democratic values as explained earlier, but none of them address a potential underlying security logic in the discourse about political micro-targeting. Furthermore, most of the academic literature about political micro-targeting focuses on the U.S. (Bennett, 2016, p. 274). Therefore, this study contributes to the academic puzzle by focusing on the UK and critically examining how government actors interpreted the use of political micro-targeting.

1.4. Structure of the thesis

The structure of this thesis is as follows. The next section explains the body of knowledge. It does so by providing an overview of how political campaigns developed from pre-modern, modern to postmodern campaigns and how this led to the advent of political micro-targeting. Additionally, the second chapter provides a conceptualization of political micro-targeting and provides a review of the academic literature that identifies the main threats that academics associate with political micro-targeting. The third chapter is about the basic research design, which for this thesis is a single case study. The research method is a discourse analysis. Moreover, the third chapter explains the justification of the case selection and posits the object of analysis. The fourth chapter explains the context of the EU

Referendum and subsequently analyses the discourse of government reports according to three levels: a text level discussing discursive statements, an intertextual and a linguistic level. The fifth chapter presents the conclusion and a discussion with suggestions for further

(10)

10

Chapter 2 - Body of knowledge

This thesis focuses on the use of political micro-targeting in political campaigns and the threats associated with this strategy. This chapter establishes why political actors adopt political micro-targeting as a strategy in the first place. Section 2.1. aims to establish that the use of political micro-targeting does not originate out of a vacuum but is a result of political and technological developments and is part of a larger trend of scientificisation and

professionalization of political campaigns. Subsequently, section 2.2. establishes a

conceptualization of political micro-targeting as consisting of two parts. Thereafter, section 2.3. explains the main threats that the academic literature mentions most often.

2.1. Origins of political campaigns

Elections are a competition for votes. Political actors compete in elections, where they try to gain as much seats as possible at the cost of their opponents (Bennett, 2016, p. 267). Political actors have made many different communication efforts to persuade voters to vote for their candidates and policy plans. A political campaign is an organized communication effort by political actors leading up to an election or referendum (Farrell-Schmitt-Beck, 2002, p. 30). The goal of this communication effort is influencing voters. Political actors in this context are political parties or candidates and the organization which they pertain to. If a political campaign hires external experts and includes them in their organization, these are part of the campaign organization as well (Trent et al., 2016, p. 263).

The political actors must follow the logic of maximization for their strategies during campaigns, because their opponents are likely to do the same (Norris, 2000, p.

148). Therefore, political actors are always improving their communication methods in campaigns, which makes campaigns subject to continuous change (Farrell & Schmitt-Beck, 2002, p. 12). The political campaigns have become more professional and technologically advanced throughout the decades (Tenscher et al. 2012, p. 145). Political actors adjust their campaigns to new circumstances in every election and modern technologies are one of the salient factors that drives this constant change (Benoit, 2007, p. 31). The professionalization and scientificisation of campaigns are the largest general changes, which occurred in most post-industrial democracies (Norris, 2000, p. 136). Norris (2000) categorizes political

campaigns into three groups: pre-modern, modern, and postmodern campaigns. The following section uses this categorization to explain the major developments about political campaigns.

(11)

11

2.1.1. Pre-modern campaigns

Political actors communicated based on personal contacts between the political actors and the voters at the local level during the pre-modern campaign (Norris, 2000, p. 137). Political parties had many members and most people voted consistently for one party during their whole life. There was a strict partisan alignment. The media were also partisan, with each political party having its own medium to communicate with their own part of the

electorate. Radio and newspaper were the main sources of political information for the public, apart from direct campaign communication. The direct communication of political parties was decentralized, because there were small local campaign organizations that rang doorbells, organized party meetings, and spread pamphlets and posters (Norris, 2000, p. 137). Norris (2000) places the time of the pre-modern campaigns roughly until the beginning of the 1960s.

2.1.2. Modern campaigns

The modern campaigns took place from the early 1960s until the late 1980s (Norris, 2000, p. 138). The mass media originated in this period (Trent et al., 20016, p. 87). More media reported on politics and these media reached people representing all ethnic, social, and religious backgrounds. Particularly, the television became a widespread medium through which people followed the news and politics (Norris, 2000, p. 138). There was a decline in more active forms of local political communication, such as local meetings or door-to-door conversations.

According to Norris, there became an emphasis on the party leadership because of the importance of the television (Norris, 2000, p. 139). This led to the centralization of campaign organizations. There was a central level of hierarchy in the campaign organization that set out the main communication strategy. Furthermore, political actors hired professional marketers and advertisers who professionalized the communication of campaigns by using marketing strategies from the corporate sector (Gibson & Römmele, 2001, p. 33). Gibson & Römmele term this development the professionalization of campaigns.

Moreover, political actors increasingly used scientific methods for opinion polls to inform communication efforts (Trent et al., 2016, p. 7). The goal of these opinion polls was to

understand the public opinion about certain issues. The outcome of these polls informed changes to campaign communication. The term to describe this development is the

(12)

12

scientificisation of political campaigns. The budgets of campaigns increased as well. Whether the increased efforts of campaigns influenced voting behaviour is subject to a wide academic discussion. It is not the scope of this research to discuss this; however, academics assume that political campaigns matter and that information through media influences undecided voters to some extent (Farrell & Schmitt-Beck, 2002, p. 13 or; Trent et al., 2016, p. 256).

One of the main causes for the changed strategies of political actors during campaigns was a decrease in partisan alignment of media and the electorate (Norris, 2000, p. 138). The

electorate loosened its strict ties to a specific political party and voters increasingly switched between parties in different elections (Norris, 2000, p. 144). This meant that an increasing number of voters could be persuaded in political campaigns, which is a reason political campaigns become more professional and adopted ‘catch-all’ strategies through the mass media. Media became diverse and independent, also in states where there were strong ties, like in the UK (Norris, 2000, p. 144). This decline in party membership and the diversification of the media was visible in most post-industrial democracies (Norris, 2000, p. 144).

Communicating via the mass media has the advantage of reaching a large audience, but this audience will be diverse, and people will differ in how they sympathize with the message (Blaemire, 2003, p. 225). For example, an advertisement placed in a newspaper is likely to reach many people that are unlikely to sympathize with it. Political actors would generally target their messages via the mass media to different groups based on demographic and geographic segmentation of the electorate (Ridout et al., 2012, p. 4). The term ‘demographic’ relates to quantitative characteristics of a group, such as ethnicity, gender, age or education level and the term ‘geographic’ refers to the place where people live (Trent et al, 2016, p. 286). The problem is that the difference of political attitudes within demographic or

geographic groups can be just as huge as outside these categorized groups (Ridout et al., 2012, p. 3). Political actors’ level this problem in the postmodern campaigns by returning to more decentralized communication strategies.

2.1.2. Postmodern campaigns

Postmodern campaigns are the campaigns from the 1990s until the present and Norris characterizes them by a strong national coordination like the modern campaigns and by more decentralized communications through the increased use of the internet (Norris, 2000, p. 147). Nielsen (2012) argues that personal political communication matters (Nielsen, 2012, p. 19). Mass media might be ineffective because of oversaturation, audience fragmentation and limited effects (Nielsen, 2012, p. 18). There was an increase in ‘field efforts’ aimed at

(13)

13

mobilizing voters ‘on the ground’ (Bimber, 2014, p. 135). This means that campaign communication reaches voters personally and direct, by door-to-door canvassing or phone calling.

There was a further decrease in partisan dealignment, which figure 1 shows. The ‘pool’ of undecided and switching voters increased in most post-industrial democracies (Trent et al., 2016, p. 3). This development created more incentives for political actors to persuade voters. Although some scholars criticize this partisan dealignment, Dalton posits that the evidence for it is visible in many democracies and argues: “across most established democracies, fewer voters enter elections with standing party

preferences that endure to Election Day, and more voters are deciding later in the campaign based on the events of the election” (Dalton, 2016, p. 13).

A fragmented media environment typifies the postmodern campaign as well. The further fragmentation and diversification of channels and broadcasts on the television medium exemplifies this fragmentation (Norris, 2000, p. 147). Most importantly, the rise of the

internet and all digital innovations that use the internet, characterize the postmodern campaign (Norris, 2000, p. 138). New forms of decentralized interactions are possible between voters, political actors, and their campaigns on the internet (Trent et al., 2016, p. 276). Almost all political actors responded to the widespread adoption of the internet (Norris, 2000, p. 148). Trent et al. (2016) argue that the internet has “deeply altered the communication of political campaigns” (Trent et al., 2016, p. 6). Furthermore, campaign information can reach voters without an intermediate medium.

The internet developed into the Web 2.0. There is a wide discussion on the meaning of this term, however there is consensus that Web 2.0 indicates that the internet got more user-generated content and provided a platform for social media (Hosch, 2018). Social media have become an important feature of political campaigns (Kruikemeier et al., 2017, p. 367). These are programs on the internet designed for social interactions and information exchange. The

Figure 1: This graphic shows the increasing partisan dealignment in four democratic states (Dalton, 2016, p. 11).

(14)

14

largest and most important ones for political actors are Facebook and Twitter (Spierings & Jacobs, 2014, p. 138).

Facebook is a communications platform based on social interactions (Trent et al, 2016, p. 295). People have their own ‘pages’ where they can post photos, stories or videos which in turn is read by other people who follow this person by ‘liking’ her. Every individual has a continuous news feed, which differs for every individual. It is based on the online behaviour of an individual, so what content a person interacted with, what pages he/she visited and what he/she ‘liked’ In 2019, Facebook has worldwide more than two billion monthly active users (Statista, 2019). About 66% from this amount are daily users. Twitter is an online social platform based on ‘tweets’, which are short messages containing a maximum of 140

characters (Spierings & Jacobs, 2014, p. 138). For political actors with many followers, it acts as some sort of an own press agency. There are about 323 million monthly active users and 126 million daily users (Wagner, 2019).

Facebook and Twitter both provide the opportunity for companies and political actors to target messages to individuals or specific small segments of the electorate. The main

advantage of using social media for political campaigns is that it creates a direct relation between the political actor and the individual voter, without an intermediate medium (Spierings & Jacobs, 2014, p. 138). Political actors have full control over the content of the message, while is not the case when communicating via the mass media. In addition, there is space for interaction between political actors and voters where voters can express themselves (Bimber, 2014, p. 133). Although the social media have become important communication media for political campaigns, this has not completely decreased the use of mass media. These media complement each other rather than substitute (Enli, 2017, p. 51).

The hiring of more marketing strategists from the private sector by political actors solely for framing communication messages further professionalized the postmodern campaign (Norris, 2000, p. 140). The goal of these efforts is to gain more control in the diverse and complex news environment (Norris, 2000, p. 140). The political actors increasingly adopted marketing techniques for their campaigns where emphasis is on the ‘customer’, which in this the case are voters.

This emphasis on the customer means that campaign staff increasingly try to figure out the needs, wants and drives of voters through opinion polls, surveys or focus groups (Norris, 2000, p. 147). The scientificisation of campaigns intensified via this way. Campaign organizations hire more opinion pollsters and scientists who organize focus groups to get a better grip on the opinions of the electorate. This scientificisation of campaigns combined

(15)

15

with the Web 2.0, led to the increased adoption of the analysis of large and detailed datasets to create predictive models about voters (Nickerson & Rogers, 2014, p. 51). These datasets contain data gathered about human behaviour from mostly digital sources, such as smartphone applications, Facebook, or Google.

It is generally argued in the academic literature that the campaign of Barack Obama in the U.S. presidential election of 2012 was the first to employ large-scale data analytics and predictive modelling, where the quantity of the data and the quality of the analysis was unprecedented (Bimber, 2014, p. 131). The predictive models that resulted from the data analysis, used scores about ‘likelihood to vote’ in general, ‘likelihood to support Obama’, ‘likelihood that non-partisan voters could be persuaded toward Obama on the basis of specific issues’, ‘likelihood of donating’ and ‘likelihood of volunteering’ to categorize voters (Bimber, 2014, p. 143). These theoretical categorizations directed personalised campaign

communication to individuals. The main difference with previous efforts in campaigns by modelling voters is that political actors gather data about the individual, rather than about demographic subgroups (Anstead, 2017, p. 296).

The campaign hired about 50 data analytics specialists (Bimber, 2014, p. 141). Their campaign director claimed that the campaign had spent 100 million dollars on technology (Bimber, 2014, p. 141). The Democratic Party, just as the Republican Party have built up the so-called ‘voter files’, which are databases containing data from the electorate and which are typical for the U.S. These data stem most importantly from state elections registration offices, where Americans register to vote. The basic information in those databases are name, address, gender, date of birth and party affiliation of the individual and are public (Rubinstein, 2014, p. 868). Voter files are the backbone of the micro-targeting operations in the U.S. The Obama campaign added data they bought from commercial databases to these voter files (Bimber, 2014, p. 141). Those databases contain e.g. data about magazine subscriptions, musical preferences, or online shopping behaviour. Those data about people are worth nothing on itself, but combined they reveal important patterns about individual behaviour which can be related to the likelihood to vote and support (Bimber, 2014, p. 142).

The result of the use of data analytics and predictive modelling by the Obama campaign is summarized in the following way: “By the end of the campaign, it became clear that the Obama campaign had predicted with enviable accuracy the behaviour of voters in swing states and reached a sufficient number of them with the right message to prevail in a contest in which Obama received many fewer votes nationwide than he had in 2008” (Bimber, 2014, p. 143). The term ‘political micro-targeting’ captures the combination of data analysis,

(16)

16

predictive modelling and targeted political communication as described in the previous paragraphs about the Obama campaign.

2.2. Conceptualization of political micro-targeting

Political micro-targeting is defined as the process of campaign experts analysing many detailed data, often using increasingly sophisticated techniques, to uncover previously

unrevealed but key relationships, that can be used to identify more precisely potential and persuadable voters and therefore to micro-target more persuasive messages that specifically appeal to them (Murray & Scime, 2010, p. 144). This definition consists of two parts. The first part is about data gathering, data analyses and the ability of campaign organizations to predict the political preferences of individuals (Murray & Scime, 2010, p. 144). This happens through the analysis of large datasets, with generally four sorts of data. It depends on the legislative context and the resources available to campaign organizations to what extent these data are available. The next paragraph presents the possible sorts of data based on the U.S. situation, but campaign organizations in other democratic states also try to incorporate these four categories for political micro-targeting (Dobber et al., 2017, p. 3).

The first are political data, which are data about the political attitudes of individuals, such as voting history, (previous) party membership, party donations, etc. (Rusch et al., 2013, p. 1615). In the U.S., such information is publicly available, but this is not the case in most other democratic states. The second form of data are consumer data, which are data about “names, addresses, address histories, driving records, criminal records, and consumer purchases like magazine subscriptions, mortgage information, credit-card purchases, gun ownership, etc.” (Schipper & Woo, 2012, p. 4). Data brokers sell these consumer data to political parties (Murray & Scime, 2010, p. 144).

The third form of data are online data from social media or web browsing activities. The companies behind websites store, analyse and sell many of the mouse clicks, website visits or interactions online (Kosinski et al, 2016, p. 493). For example, the data from behaviour on Facebook can provide important insights. It is claimed that a 100 ‘likes’ on Facebook can reveal a person’s personality better than that person’s parents could do this (Youyou et al., 2015, p. 1037). Some online data are publicly available, such as open social media profiles. Fourth, campaigns generate data themselves from surveys, focus groups or the people ‘on the ground’ (Rusch et al., 2013, p. 1616). This is a continuation of the

(17)

17

personally persuade voters, fill in ‘walk sheets’ about the political attitudes of the people they talk to (Bimber, 2014, p. 141). Campaign staff add this information to the data analyses. Other possibilities are sending out questionnaires about political perceptions of voters in strategic regions to ask them or the organizations of focus groups to test the effectiveness of certain frames or advertisements by different groups of voters.

The result of the analyses of these different sorts of data are models that predict the likelihood of individual voters to go vote and the likelihood that they support political actors (Ridout et al., 2012, p. 3). This is generally the method, but political campaign organizations differ in how they make predictive models, what data and what algorithms they use and what kind of predictions they precisely make about individuals. It is unknown what the

mathematical algorithms precisely are, because campaigns will not disclose such information (Farrell & Schmitt-Beck, 2002, p. 12).

Political actors use these scores about likely voting behaviourin the predictive models to send tailored messages to those parts of the electorate that are likely to be receptive to a message (Rubinstein, 2014, p. 883). That is the second part of political micro-targeting, which is about the new forms of advertising on social media. These media make it possible to direct political advertisements to a very small group of people, which allows the advertisements to be personal and specific (Madsen & Pilditch, 2018, p. 2). However, micro-targeted campaign communication includes certain offline communication as well, for example by phone calling or door-to-door canvassing (Nielsen, 2012, p. 12). The political actors personalize the

communication in these offline methods as well, because the predictive models inform the issues discussed in conversations.

2.3. Identified threats in the academic literature

There are several threats associated with the use of political micro-targeting in campaigns that could harm the right to free and fair elections and the right to a private sphere. Political micro-targeting could be regarded as a security threat if it threatens to damage these

democratic values. This section explains the most important threats that a review of the academic literature identifies. This thesis codes the discourse analysis according to these categories, which chapter 3 explains in detail. First, there are the issues that threaten the free- and fairness of elections.

There is the threat of manipulation of people’s voting choice (Kruschinski & Haller, 2017, p.17). Gorton (2016) posits that campaigns “know with near pinpoint precision how to

(18)

18

resonate with them” (Gorton, 2016, p. 79). A voter could increasingly receive political

advertisements adjusted to the specific political attitudes of that individual. This could lead to asymmetric information flows where every individual could receive a different message with a different frame about the policy views of political parties (Kruschinski & Haller, 2017, p. 17). This is a form of manipulation, because voters will only receive selective information on certain topics and miss the ‘whole picture’. It is important for fair elections that voters can inform themselves freely about political issues and candidates and not just receive those views that are likely to interest them (European Commission, 2018, p. 1). Political micro-targeting might be a threat to that principle.

Secondly, there is the threat of campaigns becoming non-transparent processes

(Rubinstein, 2014, p. 897). Voters do not know that they are seeing selective information by targeted advertisements, since political micro-targeting is an opaque technique. Political actors increasingly advertise on social media during political campaigns. These

advertisements are not public, but are only visible for those that receive them. Therefore, the campaigns of political actors become non-transparent, while the fairness of elections requires openness and transparency (Rubinstein, 2014, p. 897).

Subsequently, the narrowcasting of information to small groups of voters in political campaigns contributes to filter bubbles, according to Pariser (2011, p. 85). A filter bubble is a personalized information flow, typically on digital media. People will only see that kind of information that is within their bubble. The algorithms of the digital media programs create the filter bubbles. This could damage free and fair elections, because it becomes uncommon that people receive information from outside their bubble since the bubble filters this away (Endres & Kelly, 2018, p. 26). Therefore, people might not know about all the political options available, but only those that their filter bubble presents. People will increasingly think and receive information in small, like-minded groups, according to Pariser, which creates a ‘sorted’ public sphere that is hostile to dialogue (Pariser, 2011, p. 164).

Finally, there is the threat of political redlining. The goal of political micro-targeting is to communicate only to those people that might be persuadable (Bodó et al., 2017, p. 4). This could mean that political actors leave some groups of voters out of their campaign

communication based on algorithmic decisions (Endres & Kelly, 2018, p. 14). That limits the fairness of elections since some groups will receive no political information about the

campaigns. As Kruschinski & Haller argue: “in a worst case scenario this could lead to a defect in democratic discourse when specific minority issues are not addressed by politicians for strategic reasons (Kruschinski & Haller, 2017, p. 17).

(19)

19

Secondly, there is the threat that political micro-targeting poses to the right of a private sphere. For example, there is the risk that campaign organizations use certain data for political purposes while people have not given explicit consent for that purpose (Bennett, 2016, p. 262). That is a violation of the right to a private sphere. There is also the risk of data breaches. Hackers could increasingly try to breach the databases of political parties, when these parties have large and valuable databases (Bennett, 2016, p. 273). There have been cyberattacks on the databases of political campaigns, where hackers tried to unauthorized access to the data about the political attitudes of voters (Bennett, 2016, p. 273). Via this way, political micro-targeting might lead to insecurity and the evasion of the private sphere.

Rubinstein posits as well that voters do not know that political actors gather many data about voters and what kind of insights they derive from these data (Rubinstein, 2014, p. 893). This lack of transparency makes it impossible for people to exercise their rights over their personal data. Moreover, the gathering of many data about individuals by political actors might create a system of voter surveillance (Bennett, 2015, p. 371). This is a threat to the right of a private sphere, where people are free and secure of being ‘watched’ (In ‘t Veld, 2017, p. 6).

These issues have in common that the use of political micro-targeting might damage the principle of free and fair elections and the right to a private sphere. As such, the use of political micro-targeting has a high probability to damage acquired values and can be regarded as a security threat. These two values in particular are fundamentally important for many post-industrial democratic states and should be protected against anything that would threaten to damage these values. The right to free and fair elections is enshrined in human rights law (OHCHR, 2011, p. 4). The Office of the High Commissioner for Human Rights posits that free and fair elections hold that, most importantly, the outcome of the election “must represent the real and informed choice of voters” (OHCHR, 2011, p. 8). The European Commission argues that democratic values are “imperative for a society where pluralism and tolerance prevail, and where European citizens can vote with the security that they are not being misled” (European Commission, 2018, p. 1). European Governments should protect the fundamental right of people to vote independently and freely in elections (European Union, 2012).

Thereafter, European citizens have the right to a private sphere. This is enshrined in in the Charter of Fundamental Rights of the European Union (European Union, 2012). This means that they have a right a private life at home and their communications where they are free from being watched, as article 7 of the Charter of Fundamental Rights of the EU states

(20)

20

(European Union, 2012). Furthermore, the right to privacy is about personal data as well. People have the right to know which organizations processes data about them and with what purpose, as article 8 states of the Charter. Moreover, organizations must ask consent from individuals to process their personal data.

The identified threats could damage these two democratic values and could therefore be regarded as a security threat. However, there are scholars that question these identified threats. For example, Kreiss argues that some academics exaggerate the threats about manipulation or privacy (Kreiss, 2017, p. 11). He argues that there is no prove of an actual effect of political micro-targeting on voting behaviour. Kreiss points to ‘decades of

communication research’ that prove “that it is really hard to change people’s minds through campaigns” (Kreiss, 2017, p. 2). Moreover, Kreiss contends that the academics, whom associate threats against democratic values with political micro-targeting, have an idealized concept of democracy (Kreiss, 2017, p. 2). In such a democracy, there is an ideal place of public deliberation, where rational voters make their choices based on an open and free debate in the public sphere. The threats identified in the literature would threaten such an ideal public sphere. However, in practice most democracies do not work like this (Kreiss, 2017, p. 3).

Baldwin-Philippi (2017) has the same reservation as Kreiss: “the concern that we are being driven away from democracy by data is overstated. While campaigns certainly try to control citizens’ actions and information, and often succeed, this system does not have the strict control that many of these concerns rely on” (Baldwin-Philippi, 2017, p. 631). In particular, she undermines the threat of segmentation of voters and the creation of filter bubbles. Instead, the information flows of diverse group of voters often overlap and duplicate. She contends that voters are not receiving information only in small, like-minded groups (Baldwin-Philippi, 2017, p. 631). Nickerson & Rogers also argue that academics overdraw many of the threats about the data-gathering practices of campaigns (Nickerson & Rogers, 2014, p. 57). They argue about the threat to privacy that micro-targeting might form. They posit that most of the data that political actors use in campaigns are publicly available, therefore not limiting citizens’ privacy (Nickerson & Rogers, 2014, p. 57).

Hersh (2015) studied the way in which the Democratic Party in the United States uses political micro-targeting and got access to their databases. His conclusion is that political campaigns are far from all knowing and are not limiting the private sphere of voters as much as certain accounts argue (Hersh, 2015, p. 197). Rather, political campaigns often do not have a coherent strategy to specifically target specific individuals and improvise on many

(21)

21

about the manipulation of voters and the threat about the creation of a system of voter

surveillance (Hersh, 2015, p. 214). Rather, campaigns have a hard time figuring out what the preferences of voters are, even after the increased usage of data analyses by the Obama campaign. Moreover, Lynch et al. (2018) found that many of the reporting about political micro-targeting tends to rely on the statements of data analytics companies themselves. However, data analytics companies are commercial companies, so they have an incentive to overstate the influence of micro-targeting practices, which might result in selling more products or services (Lynch et al., 2018).

These accounts of Hersh, Kreiss, Baldwin-Philippi, Nickerson & Rogers and Lynch et al. are downgrading the threats associated with political micro-targeting in political

campaigns. These contrasting ideas are puzzling, since the arguments of both ‘sides’ seem convincing. The main research question aims at contributing to this puzzle, since studying an underlying security discourse would reveal whether the associated threats of political micro-targeting are related to convincing arguments or whether they are overstatements. The next section explains the research design for answering the main question.

(22)

22

Chapter 3 - Research design

3.1. Single case study and case selection

The research question already indicates that the basic research design of this thesis is a single case study. This design aims at a “detailed and intensive analysis of a single case” (Bryman, 2012, p. 66). The advantage of this design is that it enables an in-depth analysis about one case, where the study establishes a thorough picture of the case and its context. Having multiple cases limits the deep description of a case. This thesis studies the potential security threats of the use of political micro-targeting in the United Kingdom.

The focus of the analysis is the discourse following the 2016 United Kingdom European Union membership referendum (EU Referendum) where several political parties campaigned. Covering the discourse over multiple democratic states and multiple elections is beyond the reach of this study, so the focus is on the UK for several reasons. It has become clear that political actors in the UK used political micro-targeting in their campaigns as well (Anstead, 2017, p. 501). In particular, campaigns for the EU referendum in the UK in 2016 were associated with political micro-targeting (Cadwalladr, 2019). There was widespread attention for the micro-targeting company Cambridge Analytica that played a role in facilitating political micro-targeting in both elections after a former employee of that company revealed their practices and methods (Cadwalladr, 2019). This launched several government investigations and reports. For these reasons, the EU Referendum is a discursive event and is crucial to examine the security threats of political micro-targeting (Bryman, 2012, p. 538).

Another reason for the focus on the UK, is that the sceptical accounts from Kreiss, Hersh, Nickerson & Rogers and Baldwin-Philippi focused largely on the U.S. context. Much of the academic literature about political micro-targeting is about the U.S. as well. Therefore, the threats identified in the literature review or the critical accounts of Kreiss, Hersh and

Baldwin-Philippi might be different in another electoral context. Anstead compared the data practices of political parties in the UK in 2015 with political parties in the U.S. and argues the main difference is that parties in the UK have far less possibilities to gather and analyse data (Anstead, 2017, p. 305). The main cause for this difference is the lack of the voter files. The UK falls under a jurisdiction with different privacy laws, a different electoral system and regulatory framework than the U.S. (Dobber et al., 2017, p. 16). This might be a factor in

(23)

23

framing political micro-targeting as a threat (Anstead, 2017, p. 501). A study into the UK helps to limit the bias about the U.S. in the academic literature about political micro-targeting.

3.2. Critical discourse analysis

Discourse analysis is part of social constructivism, which is a methodological conception within social research that emphasizes the subjective dimension of research and denies the possibility of a definitive portrayal of social reality (Bryman, 2012, p. 529). Rather, the interpretation of social reality is subjective and constructed by actors like politicians, organizations, or researchers. The philosopher Michel Foucault laid the theoretical foundation for discourse analysis (Bryman, 2012, p. 528). For Foucault, academics should not treat discourse merely as a group of signs, rather, discourses are “practices that systematically form the objects of which they speak” (Foucault, 1972, p. 49). A discourse frames the way we understand an object by depicting it in a certain way and assigning a set of linguistic categories relating to that object (Bryman, 2012, p. 528). This way of attributing linguistic categories to frame an object provides a justification for the power of practitioners to understand a problem and to act on that problem (Bryman, 2012, p. 528).

Political actors try to accomplish things with language in speeches or texts (Bryman, 2012, p. 529). By doing so, their discourse “entails a selection from many viable renditions and that in the process a particular depiction of reality is built up” (Bryman, 2012, p. 529). That interpretation of selecting a viable rendition is subjective and rules out other

interpretations in the process. Actors explain or disseminate their subjective interpretations of social reality using language in the form of speeches or texts. A discourse is about these language acts and make certain actions possible, desirable, or even inevitable (Bryman, 2012, p. 536).

There are different kind of discourse analysis, but there is one which is concerned with the analysis of political texts. That is the Critical Discourse Analysis (CDA), which this thesis uses for the analysis. CDA aims at revealing the meaning of a particular phenomenon by examining how the discourse relates to other discourses, how texts employ the discourse, how the discourse chooses an interpretation of social life and there makes certain actions possible and how the actors use the discourse to legitimate their actions (Bryman, 2012, p. 537). This thesis adopts the CDA to study whether the use of political micro-targeting is associated with threatening democratic values after the UK EU Referendum by government agencies. By doing so, the goal is to construct how British government agencies interpreted the use of

(24)

24

political micro-targeting in political campaigns and whether they regard it as a threat to free and fair elections and/or as a threat to the right to a private sphere.

3.3. Object of analysis

The discourse plane are government reports following the EU Referendum; thus, the object of analysis is the language in these reports. The conceptualization explained political micro-targeting as consisting of two parts, so the material should address one of these two parts or both. There are six main government reports about the use of political micro-targeting following the EU Referendum. These address either part one or part two of political micro-targeting, or both. Two reports are from the Information Commissioner's Office (ICO) and two reports are from the Digital, Culture, Media, and Sports Committee (DCMSC) of the House of Commons.

Subsequently, there is a report from the Electoral Commission called “Digital

Campaigning: increasing transparency for voters” (Electoral Commission, 2018). Finally, the UK Government proposed legislation to regulate the internet in a White Paper called ‘Online Harms’ (Wright & Javids, 2019). For the ease of reading, this thesis refers to the White Paper sometimes as ‘report’ and classifies the UK Government under ‘government organizations’ when referring to all organizations. Furthermore, the term ‘government’ means the wide range of public bodies which includes the ICO, DCMSC, Electoral Commission and UK

Government. That term is different from the ‘UK Government’ which is the central

government. This thesis acknowledges that there are more agencies in the UK that deal with micro-targeting practices, but the mentioned organizations seem like the most important ones in dealing with political micro-targeting. Moreover, their reports are the most extensive ones.

3.4. Analytical steps

The academic Florian Schneider offers a methodological approach to conduct a CDA (Schneider, 2013). He made this approach by synthesizing the work of three leading

academics in discourse analysis, which are Siegfried Jäger (2004), Paul Chilton (2004) & Norman Fairclough (1994). This thesis uses the ten steps in Schneider’s approach (2013) to analyse the discourse, and supplements this with information from Van Dijk (2001).

Moreover, these steps are in accordance with the explanation of CDA from Bryman (2012), that this thesis mentioned in section 3.2. Since the first and second step are very similar, just like the eighth and ninth step, this thesis combines them in two separate steps. The emphasis

(25)

25

of the analysis is on step five, six and seven, because Fairclough and Bryman argue that these are the three most important steps for CDA (Fairclough, 1995, p. 98; Bryman, 2012, p. 538). In addition, step two, three and four are working steps. This thesis does execute them, but does not incorporate them in the written analysis of chapter four.

Step one is about establishing the context. This section must explain how the chosen sources fit into the ‘big picture’ (Van Dijk, 2001, p. 356). Van Dijk proposes to examine factors that deal with the context in which an actor produces a speech or text. In addition, he adds that the context should explain the social or institutional role of the actors. Bryman adds to this step the explanation of a discursive event (Bryman, 2012, p. 538). That is an event which generated discourse. As indicated, the discursive event in this thesis is the UK EU Referendum.

Step two is about preparing the material for the analysis. This step holds that the material should be placed in a format that makes it easy to analyse, which is what this thesis did (Schneider, 2013).

Step three is the coding of the material. Coding is the assigning of attributes to specific units of the analysis (Schneider, 2013). Schneider (2013) argues that the analysis should start with coding categories that the researcher theoretically derives from a review of the literature. The review of the academic literature about micro-targeting resulted in several identified threats, which section 2.3. explained. This thesis takes these threats as discourse strands and codes the issues in the government reports according to these threats. For example, the section about ‘data breaches’ discusses all the arguments or discursive statements relating to concerns over data breaches. All the threats in the government reports about political micro-targeting related to some extent to these categories, so this thesis made no extra categories.

Step four is the examination of the structure of the text (Schneider, 2013). This part deals with the classification of the chapters and how the headers guide to text. It is important to examine the role of the introduction and conclusion. This thesis places emphasis on the executive summaries of the reports, since they are often most read.

Step five is about the collection and examination of discursive statements. This part aims at ‘zooming in’ on the individual fragments and discursive statements of the texts (Schneider, 2013). This section discusses the actual content and meaning of the text (Bryman, 2012, p. 538).

Step six is about intertextuality. Schneider posits that this section must look at the references of the texts and what meaning they attribute to these external sources (Schneider, 2013). A text or speech is usually based on other events or statements. Intertextuality aims at

(26)

26

uncovering the function of these external sources for the arguments in the material under investigation (Van Dijk, 2001, p. 417). This thesis examines these external sources as well, since some of the arguments in the government reports relied on external sources. For example, some government agencies commissioned reports alongside their own report. The section about intertextuality discusses those reports.

Step seven is about identifying rhetorical mechanisms. There must be an examination of the form of discursive interaction used to communicate meaning and beliefs (Bryman, 2012, p. 538). Van Dijk argues that a discourse analysis must also examine the form and style of the discursive statements (Van Dijk, 2001, p. 357). Schneider, identifies several linguistic tools to examine texts (Schneider, 2013). This thesis examines two linguistic mechanisms that were most prevalent throughout the government discourse. These are rhetorical mechanisms and modalities. Rhetorical mechanisms function to frame or simplify specific statements (Schneider, 2013). The texts mostly employ hyperboles. Furthermore, modalities are about words like ‘should’ or ‘could’, but are also about words that imply a sense of urgency (Schneider, 2013). According to Baldwin, the framing of a security threat often implies a ‘sense of urgency’ (Baldwin, 1997, p. 2). Therefore, this thesis analyses linguistic features of the texts that posit a sense of urgency.

The last step contains interpreting and presenting the findings of the reading, analysing, and coding process (Schneider, 2013). This thesis places the interpretations within each analytical step.

3.5. Limitations

This research design has several limitations. One limitation is that the data retrieved are all secondary. This study is dependent on how others have interpreted the use of political micro-targeting. It only reviews written documents with no consideration for field experience. Moreover, this thesis does not measure any effect of political micro-targeting. Another limitation is that the findings of a single case study are often not externally valid (Bryman, 2012, p. 69). It is only one case, which makes it hard to generalize to other cases. Furthermore, the thesis tries to cover all the potential threats that the government agencies associate with political micro-targeting, but there is no fixed list of threats. The chance exists that the analysis leaves out some threats, making the study potentially unreliable (Leung, 2015, p. 325). Furthermore, there are more government agencies in the UK that deal with political

(27)

27

(28)

28

Chapter 4 – Critical discourse analysis of UK government reports

This chapter executes the critical discourse analysis. The first part establishes the context by explaining the discursive event and the role of the government actors that made the reports. Section 4.2. provides the discursive statements according to the coding categories and provides. The section thereafter examines the external sources about intertextuality, while section 4.4. examines the linguistic mechanisms in the discursive statements.

4.1. Establishing the context

4.1.1. The discursive event: the UK EU Referendum

Ever since the UK joined the European Union, there have been Eurosceptic sentiments within the British public (Swales, 2016, p. 4). These have been steadily growing, culminating in a national referendum about the EU (Swales, 2016, p. 4). The percentage of the British people willing to leave the EU was about 26% in 2015 (Curtice & Ormston, 2015, p. 4). The referendum took place on June 23 2016 and posed the following question to the British

electorate: “Should the United Kingdom remain a member of the European Union or leave the European Union?” (Liberini et al., 2017, p. 6). The choice was between either ‘remain’ or ‘leave’ the EU. Many commentators termed ‘leave’ option as a ‘Brexit’. The outcome was that 51.9 % of the people voted ‘leave’, while 48.1% of the electorate voted ‘remain’. The outcome resulted in the new Prime Minister Theresa May officially opting out of the EU on 29th March of 2017 (European Council, 2017).

There were two main campaigns and two unofficial campaigns. The ‘Vote Leave’ Campaign (Leave) was a cross-party campaign. Officially, the two main parties in British parliament, Labour and the Conservative Party supported the Remain campaign (Hobolt, 2016, p. 1261). However, since the Referendum question divided both parties, there were members of parliament from both the Conservative and Labour party and members of the cabinet that openly joined the Leave (Curtice, 2017, p. 25). The focus of the campaign’s communication was about the negative effects of EU regulated policy domains, such as the economy and immigration (Hobolt, 2016, p. 1262).

‘Britain Stronger in Europe’ was the name of the official campaign that supported the ‘remain’ vote. This was a cross-party campaign as well (Mullen, 2016, p. 89). 96% of Labour Members of Parliament (MP) were joining this campaign, as was their leader Jeremy Corbyn (Curtice, 2017, p. 25). Many members of the Conservative party joined this campaign, as

(29)

29

were members of the cabinet such as Prime Minister David Cameron. The Remain campaign focused on the positive effects of remaining a member of the EU by mostly arguing about the economic benefits of membership (Swales, 2016, p. 5).

The Leave campaign did not allow the populist party Ukip, a strong supporter of Brexit, to join the official leave campaign. Therefore, Ukip made their own campaign

‘Leave.EU’ (Ford & Goodwin, 2017, p. 23). The party Ukip, with its leader Nigel Farage, was one of the boosters of anti-EU sentiments. Arron Banks, an insurance company owner, was the main funder of Ukip and their campaign for the EU referendum. The Democratic Unionist Party also set up their own campaign to support the ‘leave’ vote, which it held in Northern Ireland and England (DCMSC, 2019, p. 64).

All the campaigns used political micro-targeting in the campaign for the EU Referendum, but Vote Leave, Leave.EU and Britain Stronger in Europe used it most extensively (Shipman, 2016, p. 437). The Leave and Remain campaigns used “highly sophisticated data-mining techniques to analyse the electoral roll, bought-in databases and ‘scraping’ information from people’s social media accounts to build a model of a campaign’s target voters” (Shipman, 2016, p. 437).

4.1.2. The UK Government

The Department for Digital, Culture, Media and Sport (DCMS) and the Home Office both compiled the White Paper ‘Online Harms’. These two departments are led by Secretaries that are part of the current UK Government, which consists of members of the Conservative Party. The main priorities of the department for DCMS are: “growing the economy;

connecting the UK; encouraging participation; sustaining excellence and promoting Britain; supporting our media; ensuring social responsibility” (DDCMS, 2017). As such, it is

concerned with the Web 2.0 and social media. The UK Government recently changed the Department’s name in July 2017, adding ‘Digital’ to it (Parliament, 2019). This emphasizes the role of the department in the digital sphere as well. The department works with 46 agencies and public bodies, from which the ICO and the DCMSC are two.

The main role of the Home Office (H.O.) is to keep the citizens of the UK safe and secure (Home Office, 2018). The priorities of this department are: “Counter terrorism; cut crime; protect vulnerable people and communities; control immigration; manage crises; deliver world-class services and promote growth; leaving the European Union (Home Office, 2018). It overseas 30 agencies and public bodies, such as the police, intelligence services and immigration bodies.

(30)

30

These two departments drafted the White Paper together because the digital online sphere touches upon both their policy areas. The White Paper passed the phase of consulting the public and is now in the phase of consultation from experts of industry, think tanks, civil society, and political parties (Wright & Javid, 2019, p. 95). The Government proposes several legislative measures in the White Paper with regards to harmful behaviour on the internet. The intent of the British government is ambitious: “the government wants the UK to be the safest place in the world to go online, and the best place to start and grow a digital business”

(Wright & Javid, 2019, p. 5). The White Paper aims at enforcing the first legislative measures in the world for oversight and accountability of tech companies, and for user-generated content on social media (Wright & Javid, 2019, p. 3).

The government published the White Paper in May 2019. The White Paper is about more topics apart from the topic of this thesis, such as internet bullying or the dark

web(Wright & Javid, 2019, p. 22).. However, it is partly about political micro-targeting, harms to free and fair elections and the right to privacy. In particular, the Government

mentions the danger that online disinformation poses to democratic values and principles, the creation of filter bubbles and the lack of transparency in online political communication (Wright & Javid, 2019, p. 5). This thesis analyses these parts of the White Paper. 4.1.3. The Digital, Culture, Media, and Sport Committee (DCMSC):

The role of the DCMSC is to monitor the policy, administration, and expenditure of the Department of DCMS and its associated agencies (Parliament, 2019). The Committee has the powers to order people to give evidence and to obtain documents sealed in another

jurisdiction (DCMSC, 2019, p. 5). The Committee has eleven members from the House of Commons. Furthermore, the DCMSC founded a Sub-Committee on Disinformation after the publication of the report about ‘disinformation and fake news’ (Parliament, 2019a). They argue that “there is a public interest in continuing our examination of the continuing threat posed by disinformation to democracies” (Parliament, 2019a).

The DCMSC published its interim report on July 29 2018 and the final report on 18 February 2019. There was an interim report because the inquiry raised concerns that they wished to address immediately (DCMSC, 2018, p. 5). In particular, the DCMSC points to the allegations made in The Observer of Cadwalladr (2017) as the main reason for starting their investigation: “This kind of evidence led us to explore the use of data analytics and

psychological profiling to target people on social media with political content, as its political impact has been profound, but largely unappreciated” (DCMSC, 2018, p. 4).

(31)

31

The DCMSC called their final report 'Disinformation and ‘fake news’. The interim and the final report are the result of an inquiry on disinformation in political elections in the UK but in other states as well (DCMSC, 2019, p. 5). It investigated possible violations of voters’ privacy, how online information might have influenced voting behaviour and to what extent foreign actors have tried to manipulate the elections. These reports are also about the relation between data breaches in British and American elections. For the most part, however, the reports are about the EU referendum. The reports are specifically about how the political campaigns hired companies like Cambridge Analytica & AggregateIQ to execute micro-targeted messages.

The DCMSC reports are the result of collaborations with governmental agencies from other countries, the questioning of relevant individuals in oral evidence session and the receiving of 170 written submissions (DCMSC, 2019, p. 8). By doing so, they spoke to representatives of political parties, data brokers, social media platforms, data analytics companies, university staff and more. Moreover, the DCMSC has the power to let people testify, so they spoke to many of people that might have breached the right to a private sphere or have threatened to damage the fairness and freeness of elections.

4.1.4. The Information Commissioner’s Office (ICO):

The ICO is the data protection authority of the UK. The current Information

Commissioner is Elizabeth Denham. The ICO functions as an independent, non-departmental public organization, which the Department for DCMS sponsors (ICO, 2016, p. 19). They protect people’s information rights and currently oversee the Data Protection Act 2018 (DPA2018), which is an implementation of the General Data Protection Regulation (GDPR). During the campaign for the EU Referendum, before the implementation of the GDPR in 2018, the ICO oversaw the Data Protection Act 1998 (DPA1998) Because the ICO is a public body and checks compliance with the law, this thesis regards them as a ‘government

organization’, which is a broad term for all public bodies relating to public services. Moreover, the ICO informs companies and organizations about the processing of personal data, contributes to public awareness of data protection rights, and has several rights to implement fines on organizations who breached the data protection law (ICO, 2018c, p. 5). Moreover, the ICO provides the option for citizens to file a complaint against an organization breaching their information rights. The ICO is mostly concerned with part one of political micro-targeting. They deal with the potential unlawful processing of data and intervene if organizations do not act accordingly.