Propaganda through ‘reflexive control’ and the mediated construction of reality

The nature of reality has been a central concern of philosophy and the social sciences, but since the proliferation of social media, psychological operations have taken on greater visibility and significance in political action. ‘Fake news’ and micro-targeted and deceptive advertising in elections and votes has brought the tenuous character of political reality to the fore. The affordances of the Internet, World Wide Web and social media have enabled users to be mobilised to varying degrees of awareness for propaganda and disinformation campaigns both as producers and spreaders of content and as generators of data for profiling and targeting. This article will argue that social media platforms and the broader political economy of the Internet create the possibilities for online interactions and targeting which enable form of political intervention focused on the destabilisation of perceptions of reality and recruit users in the construction of new politically useful realities.


Introduction
It is not new for political actors to trick, cajole or confuse populations in attempts to produce political outcomes or for leaders to use propaganda as part of their military arsenal. However, since the proliferation of social media, 'psychological operations' have taken on greater visibility and significance in political action (Briant, 2018;Singer and Brooking, 2018). The lying and denunciation of critics as 'fake news' by US President Donald Trump and the micro-targeted and deceptive social media advertising in the 2016 US presidential election and United Kingdom's EU referendum brought the tenuous character of political reality to the fore (Allcott and Gentzkow, 2017;Marwick and Lewis, 2017). Valuable work has conceptualised the still emerging political consequences of 'big data' and targeted advertising enabled by online platforms. However, this has focused on political consequences of micro-targeting individuals based on psychographic profiles and related power asymmetries involved (Tufekci, 2014) and the 'modulating controls' enabled by the categorisation of users based on inferred characteristics (Cheney-Lippold, 2011). What has so far been under-emphasised is the ways in which users are mobilised to varying degrees of awareness. Propaganda and disinformation campaigns require direct employment of workers to produce and disseminate content, which has always been the case, but the strategies prominent today also depend upon mobilisation of different categories of users and audiences (many unwittingly) both as producers and spreaders of content and as generators of data for profiling and targeting. This article will argue that social media platforms and the broader political economy of the Internet create the possibilities for online interactions and targeting which enable a form of political intervention focused on the destabilisation of perceptions of reality and recruit users in the construction of new politically useful realities. The following sections will introduce the theoretical approach taken to propaganda, reality construction and sociality online in this article. Subsequently, a new way of analysing online propaganda activities will be introduced through the concept of 'reflexive control' (borrowed from a form of Russian statecraft) which is achieved through the mobilisation of three kinds of 'agents of influence'. This analysis will then be explained through its application to several relevant cases taken from recent online propaganda strategies. The implications of this analysis for how we understand the relationship between everyday Internet users and their role in the functioning of political power and the construction of social reality will be suggested in the conclusion.

Propaganda
While propaganda was initially used to neutrally describe information dissemination, it has increasingly taken on a pejorative meaning associated with manipulation and disinformation (Welch, 2014: 4-5). Official and governmental definitions tend to position it as an activity of an adversary whereas one's own agencies conduct persuasion (Kuehl, 2014: 12). To achieve desired responses, the propagandist tends to present information in two ways; by controlling the media as a source of information and presenting information which appears to be from a credible source (Jowett and O'Donnell, 1999: 42). The former might traditionally be achieved through the direct state control of media or through using legal or other measures to pressure independent outlets to transmit a particular message or censor their output. The latter would tend to work either through establishing a trusted source of information in a foreign territory which both provides useful information for local populations and propaganda (e.g. Voice of America, BBC World Service, RT) or by secretly embedding journalists in the employ of the propagandist to spread their message. Some have suggested it is more helpful to consider that propaganda (unlike persuasion) tends to appeal less to the rational will instead seeking to bypass the autonomous decisionmaking process of the target (Stanley, 2015: 62). The distinction made by Stanley is somewhat stark but does highlight an often posed distinction between an ideal of an autonomous decision-maker and one whose decision-making capacities have been manipulated or impaired. Rather, I position propaganda, at least for the purposes of this article, as deliberate attempts to manipulate a target's ability to process and 'filter' information. Crucially, this does not position those who are successfully targeted by propagandists as more 'emotional' or 'reactive' (indeed examples are presented below when 'rational' argumentation is used for propaganda) than other populations, rather, I focus on those who are attempting to disturb the filtering process. We will see that the digital economy of the Internet itself is predicated on the disturbance of such filters.
Propaganda strategies are always shaped by their media environment which can be understood in terms of the connectivity enabling information exchange, the content and the cognitive impact of using the content all of which have been transformed by the Internet and connected devices (Kuehl, 2014: 17). This new 'network propaganda' has the same intentions as older forms (manipulation for political ends) but uses aspects of the online ecosystem which make it more susceptible to the dissemination of misleading messages (Benkler et al., 2018: 24). A growing body of work has investigated the workings of this new 'computational propaganda' (Woolley and Howard, 2016) through studies of 'trolling' (Marwick and Lewis, 2017;Sest and March, 2017), automated social media accounts ('bots') (Murthy et al., 2016;Shao et al., 2018) and 'fake news' or 'disinformation' (Bosco, 2018;Briant, 2018). This literature demonstrates the change in strategies which have been enabled by digital technologies and networks such as 'hijacking' mainstream news media with extremist and conspiratorial content (Coleman, 2014;Marwick and Lewis, 2017), using 'bots' to shift political discourse (Shao et al., 2018;Singer and Brooking, 2018), micro-targeting individuals through social media advertising and exploiting 'filter bubbles' (Bosco, 2018;Silverman and Alexander, 2016).
Although these studies clearly identify the novelty of the methods used in digital propaganda, their analysis is broadly consistent with the 20th-century 'broadcast' model and does not properly take account of the extent to which 'targets' are actively engaged in practices of manipulation. The cases presented below are, of course, not entirely without precedent as ordinary members of the public have often been used to spread propagandistic messages in a manner similar to those described in the section, 'Locally recruited agents of influence' below. For instance, the US government's 'Four Minute Men' were recruited during WWI to give short speeches mostly at cinemas (while the reels were being changed), promoting the war effort with centrally directed messages targeted at particular communities through using locals with whom they could identify (Auberbach, 2017;DeBauche, 1997: 80-84). Although there is some consistency with such examples of 'peer persuasion', the technical affordances of online networks produce systems which are to some extent self-organising, enabling forms of 'citizen engagement' which are aligned with 'official' messages but not bound by traditional centralised organisational logics (Penney, 2017: 403;Stromer-Galley, 2014: 12). However, I will suggest that these logics are open to manipulation through the structuring of encounters which enable imitative behaviours and thus for flows of ideas to be 'steered'. The contemporary saturated media environment requires individuals to 'filter' information themselves and be actively involved in assessing the veracity of content and it is increasingly at this juncture where propaganda intervenes. The objectives of propaganda are no longer strictly to manipulate but to '[reorient] a targeted individual or state to self-select or privilege certain information' (Fitzgerald and Brantly, 2017: 235). Furthermore, audiences, and paid operatives, are mobilised to spread messages themselves. I will outline my theoretical approach to these developments through engagement with existing literature on the role of digital media in the social construction of reality and its potential for 'steering' beliefs and behaviours in the following two sections.

Reality construction
Analytical approaches to the social construction of reality have a long and storied history (see Burr, 2003: 10-15), but the seminal The Social Construction of Reality (Berger and Luckmann, 1971) with its emphasis on the role of communication, everyday action and practice is especially useful as a foundation for understanding what Couldry and Hepp (2017) call 'The Mediated Construction of Reality'. They assert that while communicative technologies (from writing, through the telephone to WhatsApp) have always impacted on how we communicate, the extent of integration of digital media into our lives today has led to an unprecedented state of 'deep mediatization'. Changes to the 'media ensemble' have transformed the fundamental dynamics of social figurations and therefore the ways in which meaning is produced (Couldry and Hepp, 2017: 77). This is neither relativist nor consistent with the post-truth claims or the politicisation of scientific knowledge in contemporary populist rhetoric, rather, it is an acknowledgement that 'truths' are the product of discursive practices and material relations of power and political economy (Angermuller, 2018;Sismondo, 2018).
The integration of social media content into everyday interactions (Couldry and Mejias, 2018;Couldry and Van Dijck, 2015) and mainstream media (Lăzăroiu, 2014) has accelerated the flow between what Adoni and Mane (1984) call 'objective', 'symbolic' and 'subjective' social reality. For example, an individual can present their interpretation of events ('subjective reality') on social media (whether honestly expressed or not) which can serve as content for mainstream media outlets which then forms part of the 'symbolic reality' which is assumed by others to bear at least some relation to the 'objective reality' (of actually existing social facts). Indeed, social media content can shape 'objective reality' by becoming part of the cultural and political discussion and thus necessitate reactions from political figures thereby lending it political weight (regardless of its veracity). The distance between 'objective reality' and 'subjective reality' is further collapsed when the processes by which the former is constructed is obscured as is common in the online world (Bilić, 2015(Bilić, : 1272. The perception of 'objective reality' is coloured by media institutions which construct what Couldry (2003) calls the 'myth of the mediated centre' or the assumption that there is a central understanding of the perception of social reality around which we can coalesce. But claims to such centrality have been undermined by the Internet to be replaced by a proliferation of different 'centreings' (Bilić, 2015;Hepp and Couldry, 2010). Such 'centreings' have been widely discussed in the form of 'filter bubbles' (Bozdag and Van den Hoven, 2015;Davies, 2018;Pariser, 2011) but this existing work has tended to focus on what is obscured by the filtering process and the political consequences but there are more ontological concerns relevant to this article. 'Time' (through shared histories and archives) and 'space' (through metaphors, virtual objects and the restrictions and potentialities programmed into interfaces) are central to the construction of the online aspects of social reality (Gotved, 2006). But these essential components of reality construction have become more controlled and centralised since 'platforms' have come to dominate online interactions (Casilli and Posada, 2019;Gillespie, 2010;Srnicek, 2017). For instance, Facebook constructs 'virtual realities' through their 'social graph', which maps users' interactions with one another and 'objects' (pages, videos, groups) and enables them to produce 'brand like assemblages' which position users in relation to content which does not necessarily bear any direct relationship to real-world lives and processes (Arvidsson, 2016: 9-12).
In the following section, I will suggest that this process of mediated reality construction occurs through fundamental components of sociality and that digital media provide new tools for observing and directing these everyday actions and interactions of users. This creates new opportunities for political actors to influence opinion.

'Steering' online behaviour
I argue for an approach to digital propaganda which takes account of the nuanced ways in which the actions and interactions of users and the ways in which these associations are mobilised for commercial purposes have become an integral part of propagandistic practices. For this, I turn to a sociology of the Internet informed by Gabriel Tarde's sociology of imitation and innovation. This has been argued to be well suited to analysis of the Internet because it provides an understanding of the spread of ideas 'as waves of imitative encounters [which] contain elements of mass conformity and imitation, while at the same time demonstrating elements of individual choice, innovation and creativity' (Burgess et al., 2018(Burgess et al., : 1041. Such 'waves' are enabled through the formulation of 'publics' which aid the transmission of beliefs and desires with the activities of audiences commercially valued due to their ability to produce 'a temporary association of strangers held together by a common passion and [ . . . ] commitment to particular values' (Arvidsson and Bonini, 2015: 167). The relational, Tardean approach to sociological analysis eschews any distinction between individual and society or agency and structure. Instead focusing on 'monads' which are temporary 'wholes' (or assemblages) made up of entities 'lending' some of their aspects. This is an ontology built on 'having' rather than 'being' in the sense that individual subjects can be broken down into certain characteristics (such as of gender identity) or beliefs which they 'possess' but also 'possess them' (De Freitas, 2016;Latour, 2007).
In an era in which individual users of the Internet, and particularly social media, are defined by the 'digital traces' they leave behind, it is easier than ever to identify concrete manifestations of Tarde's ontology. Through the lens of the network an individual social media user is defined by the list of attributes associated with them and by their connections to other entities (e.g. users, pages), but crucially the user and the network are formed of the same substance rather than the latter being another level added on top of the former (Latour et al., 2012: 592-3). This means that the focus must always remain on the level of the phenomena (ideas, beliefs, desires) which flow through networks rather than on the individuals or the seemingly solid social structures. Venturini (2019) has recently applied Tarde's theory to the spreading of 'fake news' (although he prefers the term 'junk news') online. He suggests that the innovation and imitation at the heart of Tardean sociality have been accelerated and trivialised by the technical affordances and political economy of a commercialised Internet built on advertising revenue (Venturini, 2019). However, Venturini opposes the reading of this phenomenon as a coordinated and sophisticated operation designed to manipulate audiences and it is here I depart from his analysis. While I agree that the events described in this article have been enabled by a multitude of developments often only loosely related (if at all), this does not mean that the present situation has been arrived at accidentally, nor does it mean that particular actors are not seeking to mobilise or direct aspects of the network for their own purposes. As Thrift (2010) has asserted, Tarde's analysis also shows us how 'imitative processes can be consciously and carefully steered' (p. 260) by 'aiding and abetting certain aspects of continual transformation, strategically bending processes so that it "ripens" in certain directions rather than others' (p. 263). This analysis provides an ontological framework which can see the dense integration of human behaviour and consciousness with digital networks without positioning human actors as idealised, atomised, rational beings who are (or should) be resistant to attempts at external ideological influence or propagandists as all-powerful and all-seeing manipulators.
The technical affordances of online networks with their capacity for 'steering' populations have combined with the advertising-focused political economy of the Internet and social media to produce a particular relation to the mediated construction of reality described above. Below I will suggest how this situation has enabled a form of propaganda which focuses on manipulating a target's selection and privileging of information to be applied to whole populations.

Reflexive control and agents of influence
It is my assertion that the theory of 'reflexive control' as a model of propaganda when combined with the theories outlined above can reveal important aspects of how propaganda functions today. Although this theory predates digital media, as it was originally used to describe strategies of 'psychological operations' developed in Russia during the Cold War, technological developments have enabled it to be mobilised against a much broader range of targets by non-state political actors and commercial enterprises. The 'reflexive control' approach is a form of 'non-linear warfare' which is a 'means of conveying to a partner or an opponent specially prepared information to incline them to voluntarily make the predetermined decision desired by the initiator of the action' (Thomas, 2004: 237). It is thus 'a process by which one enemy transmits the reasons or bases for making decisions to another' (Lepsky cited in Thomas, 2004: 238). While there are similarities with traditional forms of propaganda, 'reflexive control' is distinct as it seeks to affect the decision-making process of targeted individuals by manipulating their perception of reality. This approach requires the profiling of individuals to understand their moral and psychological dispositions by gathering data on their biography, habits and psychology (Thomas, 2004: 242). As Thomas (2004) summarises, By definition, reflexive control occurs when the controlling organ conveys (to the objective system) motives and reasons that cause it to reach the desired decision the nature of which is maintained in strict secrecy. The decision itself must be made independently. A 'reflex' itself involves the specific process of imitating the enemy's reasoning or imitating the enemy's possible behavior and causes him [sic] to make a decision unfavorable to himself [sic]. (p. 241) Ultimately the 'controller' is trying to impose their will on the target not by telling them what to think but by influencing their decision-making practices and their perceptions of reality. This is done by manipulating the 'filter' used by the target, this is used to refer to the collection of concepts, knowledge, ideas and experience through which they make decisions and distinctions between information which is useful/useless or true/ false (Thomas, 2004: 247). Traditionally 'reflexive control' was used against military and political leaders as they were deemed to be able to make strategic decisions of benefit to the controller. Moreover, it would have been impossible to target non-prominent figures or large groups because the strategy is dependent on detailed profiles of such figures which were difficult to compile. However, today such profiles are routinely produced on all social media users enabling the reflexive control approach to be applied more widely.
'Reflexive control' is one aspect of a broader strategy developed by the USSR in the 1950s which sought to exercise overt and covert influence on events and behaviours in foreign countries and referred to as 'active measures'. These were implemented by 'agents of influence' who were divided into three types: 'fully employed operatives' infiltrating organisations and spreading messages, 'locals' recruited to the cause and 'unwitting accomplices' unaware that they were the subject of control by a foreign agent (Cull et al., 2017: 3). Contemporary online propaganda functions through the mobilisation of distributed users the actions of whom can be understood through all three categories of 'agents of influence'. The following section will outline how the first category of 'fully employed operatives' are engaged in constructing new realities or using disinformation to obscure targets' perception of reality.

'Fully employed "agents of influence"'
The first and most obvious way in which the principles of 'reflexive control' are applied to digital politics is using fully employed 'agents of influence' to plant disinformation, discredit alternative viewpoints or simply cause confusion. Perhaps the most well-known example is the Russian Internet Research Agency (IRA), which is a state-funded group that operates cyber attacks in Russian interests. They seek to manipulate social and political discourse (particularly in foreign territories) by posting messages favourable to the regime or to undermine oppositional messages or individuals. This is done both through personally creating accounts and posting messages under the guise of a fictional figure ('sock puppets') or using 'botnets' to amplify messages created by 'sock puppets' or 'real' accounts useful for their cause. In 2014, IRA employees were expected as a daily minimum to post comments on news articles 50 times, manage six Facebook accounts with at least three posts, hold two discussions in news groups and manage 10 Twitter accounts with 50 tweets per day (Singer and Brooking, 2018: 111-112). Such activities by Russian operatives formed part of the influence campaign conducted during the 2016 US presidential election which was found by the US National Intelligence Council to have used, among other means, 'paid social media users or "trolls"' (Office of the Director of National Intelligence Council, 2017: ii). A UK government report on the '[l]essons learned from the EU referendum' of 2016 found similar tactics were used to 'influence public opinion' (Public Administration and Constitutional Affairs Committee, 2017: 35). Comparable strategies have been used in China to suppress anti-state sentiment in their own population. Rather than the antagonistic approach of the IRA, the Chinese '50c Party' (named for the erroneous observation that individuals are paid US50c per social media post) are civil servants seconded to spread positive messages through 'cheerleading' messages aimed at drowning out discontent or criticisms of the state (King et al., 2017). In many cases, 'fully employed' agents use 'bots' to amplify messages on private networks (e.g. closed WhatsApp and Facebook Messenger groups) as well as openly on social media. This happened when a 3500 strong 'botnet' was used to attack Marcelo Freixo (a candidate in the 2016 mayoral election in Rio de Janeiro) with the same messages being posted more than 100 times per hour (Arnaudo, 2017: 20). Despite the seeming success of 'bots' in amplifying the efforts of employed agents (for reports on their influence see, Arnaudo, 2017;Machado et al., 2018;Ruediger et al., 2018;Woolley and Howard, 2016) China, one of the most notorious and sophisticated users of propaganda techniques, largely eschews automation. Instead, when producing social media propaganda both for foreign (using Twitter) and domestic (using Weibo) audiences, they make use of their vast and cheap human resources to employ users to personally post to fake accounts in huge numbers with 6% of reposting on news stories and 30% of accounts considered 'opinion leaders' found to be fake (Bolsover and Howard, 2019: 4).
Other countries have incorporated such activities into their military wings. In a 2012 conflict, the Israeli Defence Force and Hamas conducted a social media battle alongside the use of physical force by circulating propaganda, highjacking popular hashtags and hacking commercial Facebook pages (Singer and Brooking, 2018: 193-194). In 2015, the British Army established the '77th Brigade' which drew recruits from across all units of the army to engage in 'unconventional warfare' specifically targeting recruits with skills in social media and psychological operations (Allison, 2016;MacAskill, 2015) to use digital marketing strategies (Miller, 2018) and behavioural analysis to 'adapt behaviours of the opposing forces and adversaries' (The British Army, 2017). The Israeli private intelligence firm 'PSY Group' is known to have pitched its services to the Trump campaign, a controversial religious sect in the Netherlands and the president of Gabon (Entous and Farrow, 2019;Mazzetti et al., 2018), market their services with the slogan 'Reality is a matter of perception' (PSY Group, n.d.). They claim to be able to 'Shape Reality' (Entous and Farrow, 2019) through using 'gathered intelligence and broad expertise to build highly effective, targeted, online and offline campaigns' (PSY Group, n.d.). These examples represent cases where people have been directly employed by states or companies to produce content explicitly intended to manipulate public opinion through shaping targets' perception of reality without them knowing. Moreover, this is done through engaging in specifically digital forms of intervention, that is, generating online content or creating networks and stimulating imitative sharing practices in others in a manner which is largely indistinguishable from that of genuine users.

Locally recruited 'agents of influence'
The second category of 'agents of influence' are 'locally recruited' although their intentions may not be directly aligned with those who benefit from their actions. First are those who are sympathetic to the ideology of a political group or actor but do not appear to have a direct connection to them. After Donald Trump's election General Michael Flynn (previously Trump's national security adviser) praised the 'army of digital soldiers' who he claimed helped win the White House through manipulating social media discourse (Singer and Brooking, 2018: 176). Discussion boards such as Reddit, 4Chan and 8Chan were their main tools as these are structured such that they encourage a culture of inflammatory and extreme posting with posts ordered according to recency and engagement so often the only way to stay at or near the top (and therefore gain exposure) is to attract a high number of comments or 'upvotes' (Coleman, 2014: loc 672). 'Alt right' activists, often sympathetic to Trump, see their role as being to 'wake up' centrists, or liberal conservatives, to the realities of the oppression of middle-class White males by mainstream media and liberal democracy. They engage in sustained campaigns to recruit converts by using offensive humour and irony (often in memes) to blur the line between sincerity and parody and encouraging the imitation and repetition of posting and sharing practices they would otherwise consider to be distasteful, offensive or dangerous (Marwick and Lewis, 2017;Singer and Brooking, 2018).
Some similar strategies are used by the Chinese 'voluntary 50c army' who seek to convert others to their pro-state ideology (but unlike those discussed above are not paid agents) through ridiculing and hooking opponents with fabricated anti-state stories in order to demonstrate their gullibility (Han, 2015). Similarly, committed actors were influential in Jair Bolsonaro's presidential campaign in Brazil through the mobilisation of closed WhatsApp groups the significance of which was acknowledged by his supporters heard triumphantly chanting 'WhatsApp' after his victory (Belli, 2018;Doctorow, 2019). Volunteer group administrators would ban critics of the candidate or bombard anyone questioning his policies with fabricated stories and memes designed for virality usually produced by a dedicated group of 'influencers' (Nemer, 2018). WhatsApp groups supporting specific parties in the 2019 Indian General election are also considered to have been effective with analysis showing high levels of 'junk news' being circulated in easily shareable image formats (Narayanan et al., 2019).
A second category of 'recruited agents of influence' are people whose online activities are clearly of benefit to some political groups or actors but appear to have no direct connection, or even in some cases ideological alignment, with them. Perhaps the paradigm case is the group of Macedonian teenagers who made millions of dollars producing 'click bait' websites for an American audience in the run-up to the 2016 presidential election (Tynan, 2016). They realised that there was an increasing appetite for outrageous and conspiratorial stories which were pro-Trump and anti-Clinton, so employed other teenagers to generate new content for websites they owned. This was enabled by Facebook's advertising system which could easily be 'gamed' to drive traffic if they used sufficiently enticing headlines and content (Singer and Brooking, 2018: 118-121). These young men had no political alignment with Trump but identified aspects of the social media economy which could be exploited through the generation of 'fake news' and it just so happens that the demand among, and value of, Trump supporters was much higher than other groups (Smith and Banic, 2016). Similarly, one of the most prominent American purveyors of pro-Trump 'fake news' produces what they consider to be satire, and at one time generated around US$10,000 a month through Google ads but conceded they may have helped Trump win the 2016 election (Dewey, 2017). These entrepreneurs are exploiting the political economy of the advertising-funded Internet which places high value on sensationalist disinformation and the creation and sustaining of publics through the mobilisation of interpersonal passions (Arvidsson and Bonini, 2015: 168).

Unwitting 'agents of influence'
The spreading of ideas through social media networks is dependent on the activities of individual users unwittingly acting out the desires of political actors who use outrage to attempt to trigger responses which can be seen as bypassing autonomous decisionmaking (Stanley, 2015: 16) and constituting users as 'automatic subjects' (Till, 2019). But a Tardean sociology of the Internet shows how such imitative actions are fundamental to the social thus suggestions that lack of education, critical thinking or resilience to influence are to blame for the spread of propagandistic messages are shown to be misguided. The targeting of social media users who are likely to act as effective 'receptors' and 'spreaders' to propagandistic messages has been enabled by the demographic data and psychographic profiles on Facebook and other social media sites generate. These profiles are constructed out of the everyday imitative and innovative behaviours of users with platforms directing their 'passions' through encouraging them to identify with particular brands, ideas, issues or public figures (Arvidsson, 2016;Arvidsson and Bonini, 2015).
Guardian journalist Cadwalladr's (2017) investigation into the use of Facebook data by the Leave campaign in the UK EU referendum found that Finding 'persuadable' voters is key for any campaign and with its treasure trove of data, Cambridge Analytica could target people high in neuroticism, for example, with images of immigrants 'swamping' the country. The key is finding emotional triggers for each individual voter.
Once such profiles have been constructed, anyone who wants to influence users through advertising (whether this is to buy a product or to vote) can use tools such as Facebook's 'Lookalike Audiences'. This enables advertisers to choose a source audience (a Custom Audience created with your pixel data, your mobile app data or fans of your Page) and we identify the common qualities of the people in it (ex: demographic information or interests). Then we find people who are similar to (or 'look like') them. (Facebook Ads Help Center, n.d.) The digital media director for Trump's 2016 election campaign credited the 'Lookalike Audiences' tool with a significant role in their win (Boghani, 2018). However, it is the extensive integration of Facebook with the advertising industry which has enabled this profiling to become powerful. Former Cambridge Analytica employee Christopher Wylie described how the company used 'off-the-shelf' cultural narratives constructed by the fashion, music and other industries to match up with Facebook data to determine which users would be receptive to pro-Trump messaging. They found that users positive towards 'American heritage brands' such as Wrangler were more likely to be receptive than those with a liking for Nike or Louis Vuitton (Kansara, 2018). Facebook thus provided a way to connect the aesthetic and emotional relationship people had to these brands with their 'persuadability' through analysing their engagement with the platform. It was precisely these emotional engagements with aspects of identity (as objectified in brands) which were considered the most effective ways of targeting those users most susceptible to 'mobilization' by producing messaging which articulated threats to their identity in similar ways as in online fan cultures . Thus 'patriotic' Facebook groups were created by the IRA to falsely claim the perpetrator in the shooting of a Boston police officer as a Clinton and Black Lives Matter supporter. Similarly, Russian operatives were found to have created Facebook groups and used targeted advertising to mobilise a pro-gun rally outside an Islamic centre in Houston, Texas, and the opposition to this rally (Nadler et al., 2018: 31-34). Facebook's monetisation strategy is built on the exploitation of affective relations and social reproduction (Jarrett, 2016), enabling the activities associated with them to be used for propagandistic purposes. The same imitative online activities ('liking', 'sharing' or simply reading or watching) which enable profiles to be constructed and beliefs and desires to circulate generate data which are both of commercial value to brands and political value to propagandists.

Conclusion
It is the position of this article that political power today, when functioning through social media, is dependent on the imitative activities of large swathes of the population which are 'steered' through activating affective engagement. In some cases, people are paid to construct and maintain fake social media profiles working for military divisions such as the UK Army's '77th Brigade', government agencies such as the Russian IRA or civil servants in the '50c Party' in China. Sometimes these activities are intimately connected with automated processes and the work of 'bots' and 'botnets'. However, the majority are much more loosely (if at all) connected to state agencies. This can be seen with individuals (such as the Macedonian teenagers) identifying opportunities afforded by the economics of the web and producing 'click-bait' advertising beneficial to other political actors. Message board trolls, such as those on 4Chan, are similarly cognisant of the political economy of the Internet and social media and how this can be leveraged through strategic manipulation of virality for political rather than economic ends. But most of the reality construction can be seen in the everyday actions and interactions of people who are simply engaging in political and non-political discussion who nevertheless help to spread messages and produce psychographic profiles beneficial for viral manipulation of discourse and targeted advertising.
I suggest there are two elements of the above-described influence campaigns which make them distinctive to the digital era and have been enabled by the affordances of contemporary networked digital technologies. First, all the strategies mobilise the basic elements of sociality (as identified by Tarde) of imitation and invention and their role in enabling and directing flows of beliefs and desires. The propagation of the latter has been intensified by social media, and the Internet in general as have previous innovations in communication technologies (Barry and Thrift, 2007: 518). Furthermore, social media platforms have been especially effective in this due to the innovations which have enabled them to become so successful at profiling users, enabling their propensity to act as 'receptors', and encouraging sharing behaviours thus shaping them into effective 'spreaders' (Crockett, 2017;Thrift, 2010). The information and power asymmetries between platforms and users mean that indirect control (which does not appear like control) can be produced (Rosenblat and Stark, 2016) in a manner similar to 'reflexive control'. The same network structures which enable the extraction of value produce the possibility for users to be mobilised for propagandistic purposes.
Second, the kinds of influence which are being attempted can best be understood as a form of 'reflexive control', that is, using profiles of targets to direct messages which will destabilise their 'filter' and therefore their ability to make judgements in their own interests (Kasapoglu, 2015;Thomas, 2004). The strategies used by the nation states and political campaigns described above (directly or indirectly) attempted to destabilise truth claims and/or construct new micro-realities through targeted messaging (Cull et al., 2017; Office of the Director of National Intelligence Council, 2017; Public Administration and Constitutional Affairs Committee, 2017). Although 'reflexive control' is a Russiandeveloped approach to statecraft, it is not my position that all these examples have been directly enabled or encouraged by the Russian state (although in some cases they may have been), rather that the Russian approach to political influence is becoming increasingly significant in global politics. Today the kind of detailed profiles which were previously only obtainable through time-consuming and expensive research and spying can now be easily drawn off from social media profiles constructed from data freely given away by users. These profiles help to construct 'brand like assemblages' by enabling the positioning of users in relation to content (generated by themselves or others) and advertising based on analysis of their affective relationship (Arvidsson, 2016: 12). It is these derived, 'virtual realities' which Cambridge Analytica found to be most valuable in their targeting for the US election (Ferrier, 2018;Information Commissioner's Office, 2018). While Arvidsson (2016) questions whether Facebook's 'virtual reality' bears any direct connection to 'reality' (instead seeing it as analogous to financial derivatives), I suggest that data points which underpin it could only be constructed through enticing users into the spreading of messages and construction of useful profiles. So, while Facebook's 'social graph' might be a 'virtual reality' it is one to which our lives are increasingly aligned thus making the interventions of political (or commercial) actors in online lives more powerful. Whether Internet users are 'fully employed' (e.g. '77th Brigade'), 'locally recruited' (e.g. 4Chan 'trolls') or 'unwitting' (e.g. retweeters of IRA posts) 'agents of influence', they are contributing towards the 'mediated construction of reality' (Couldry and Hepp, 2017) by producing or spreading the materials with which we construct our understandings of the world.