Book Review

Woolley, Samuel (2023). Manufacturing Consensus: Understanding Propaganda in the Era of Automation and Anonymity. New Haven and London: Yale University Press

Anne-Mette Holmgård Sundahl

Victoria University of Wellington, Wellington, New Zealand

E-mail: [email protected]

In Manufacturing Consensus, Samuel Woolley explores the democratization of propaganda that has been facilitated by social media. Propaganda is here defined as ‘the use of politically biased information in considered attempts to manipulate or influence the opinions and actions of individuals and, more broadly, society’ (p. 4). Through anonymity and automation, social media make it possible for a bigger and more varied group of people to influence political campaigns and spread propaganda more widely. The book thus emphasizes the bottom-up creation of propaganda as a contrast to the top-down focus on state actors that has been predominant thus far. Woolley focuses on how computational tools like bots and human-driven methods like sock-puppets can be used for political purposes as examined through different interviews and conversations with people who are, one way or another, linked to this process. More specifically, the book centers on four different actors: political campaign workers, technology industry experts, journalists, and automated political influencers – each group with a dedicated chapter elaborating on their reasons and methods for using computational propaganda.

Chapter 1 introduces the scope of the study as well as the key concepts and methods. In relation to the latter, namely the case selection is worth mentioning. While Woolley provides context by discussing the propagandistic use of bots in relation to different countries from around the world, namely Ecuador, the United States, the United Kingdom, and Turkey are brought to the fore (p. 25). Some aspects of the mechanisms described in the book might thus be influenced by this case selection.

In Chapter 2, Woolley provides an overview of how the study of propaganda in the context of media has been approached in the past. The title of the book is a nod to the phrase “manufacturing consent” which was popularized by Edward S. Herman and Noam Chomsky to describe how mass media have been shaped by the elite (p. 43). Woolley builds on Herman and Chomsky’s propaganda model and their five “filters” that describe how the US media operates to consider the role of social media and propaganda (p. 48). Herman and Chomsky’s original model did not take the ‘effects of the media on the public’ into account (Herman & Chomsky, 2008: xii) – this is to some extent replicated in Woolley’s study as the author similarly focuses on how propaganda is made and less on how it is received.

Chapter 3 is the first of four chapters demonstrating extensive fieldwork and conversations with people from distinct groups ranging from well-resourced and well-organized to those from the other end of the spectrum. Chapter 3 focuses on the former as Woolley elaborates on the digital activities of state-based producers of propaganda – namely in relation to the governmental bot use, state-sponsored trolling and how states can benefit from such tools and social media strategies. This is a logical starting point given that the focus on top-down propaganda and state actors is aligned with previous research, and state actors remain the most well-resourced and thus capable of utilizing computational propaganda to the fullest.

Chapter 4 engages with the less-resourced and less-organized people who are not affiliated with campaigns, corporations, or media institutions, but who are nevertheless able to go viral and attract attention from traditional media and governments. In a way, this chapter thus forms the core of the book as it illustrates the democratizing aspect of contemporary propaganda production with social media becoming a ‘many-to-many system’ allowing people beyond the traditional elite to become politically influential (p. 84). Labelling this group of people “automated political influencers”, due to their aim of influencing politics and their similar tactics as the traditional digital influencers, Woolley sets out to elaborate on the characteristics of these individuals as well as how they use automation for political communication – and more importantly, how this bottom-up propaganda dissemination affects digital disinformation (p. 94). We learn that automated political influencers can be regular people with average jobs who run complex computational propaganda campaigns on the side. While some are motivated by money, others are trying to push a certain political agenda – motives that are often intertwined. Later, in the conclusion of the book, Woolley also alludes to a potential third motivation for becoming an automated political influencer as he explains how one bot builder was hoping that if he became famous on Twitter, he could also become famous on TV (p. 162). In this sense, the motivation for bot building rather seems to be a form of self-promotion – albeit one that presumably also would lead to more money. This brings an interesting thought to mind in relation to Woolley’s definition of propaganda. If self-promotion is indeed a key motivator for some bot builders, are they then deliberately trying to manipulate people’s opinions, or are they trying to attract attention regardless of the political outcome?

This chapter also begs the question, whether the emergence of automated political influencers can be considered as a key driver of the democratization of propaganda. Wolley states that bots allow for democratization as ‘they allow the individual to make and disseminate propaganda, rather than reserving that ability to the elite’ (p. 57). But should the automated political influencers instead be considered a new type of elite? Woolley does acknowledge the fact that some level of tech-savvy is needed to contribute to this bottom-up production and dissemination of propaganda, but the power that such skills hold is mainly elaborated on in relation to the big tech companies in Chapter 5.

Woolley further raises some interesting questions at the end of Chapter 4, reflecting on who is responsible for what follows from the spread of disinformation originating from for instance sock-puppet social media accounts. The dissemination of such can have very real consequences as witnessed in the attempts to kidnap a governor, or the storming of the Capitol Hill in Washington DC. Woolley asks whether the ones who provoked such actions by manufacturing consensus online are equally responsible for these offline criminal acts. But can the people making the bots be held accountable or were they themselves influenced? Even if the automated political influencers could and should be held accountable, what about the politicians who benefit from these digital efforts? The automated political influencers can be considered unofficial middlemen in campaigns which makes it even more difficult to trace it back to the politicians. This negatively affects the accountability of politicians. In this sense, it can make the traditional or old elite even more powerful due to the lack of transparency and following lack of ability to place responsibility.

In Chapter 5, Woolley discusses the role of the big social media companies in facilitating and spreading propaganda. Individuals specifically design bots to take advantage of social media sites’ trending algorithms in the hopes that the content will go viral and get picked up by traditional media. Woolley argues that social media have “systematic problems.” One problem is that algorithms have human biases encoded into them, another is that the tech companies behind these platforms have an incentive to increase user engagement and push viral content to create profit (p. 111). By claiming that the algorithms are “objective” and merely responding to the preferences of the users, tech companies avoid taking responsibility for providing a platform where propaganda and disinformation can spread.

Chapter 6 explores how computational propaganda has influenced the work of journalists and news organizations. Woolley highlights three learnings from his fieldwork. First, as social media is a cheap and uncomplicated way of accessing information, journalists risk inadvertently reporting on the misleading information spread by computational propagandists – thus amplifying and legitimating this content. Second, journalists might also be the target of bots in campaigns, orchestrated by state-sponsored trolls or automated political influencers, intended to silence them. Third, journalists have adopted the methods of the propagandists by using bots to connect people or disseminate their data and findings. While this does not constitute propaganda in Woolley’s definition, as it is used for the purpose of transparency and honesty, it nevertheless highlights the many uses of bots.

Bots can thus facilitate that journalists can disseminate more information wider. Whether this is unequivocally beneficial for combatting disinformation or improving the public debate more generally is, however, debatable. People who follow news on social media experience information overload which decreases what they learn (van Erkel & Van Aelst, 2020) – the question is thus whether the journalistic use of bots will exacerbate this problem. Should spreading as much information as possible be a goal in itself – or is there an increasing need for journalists to only present the key facts and help the reader interpret them?

The discussion of how journalists have participated in perpetuating manufactured consensus by computational propaganda further highlights how the effects on the practices and conditions of other actors have been underexplored in the book – namely those of the political establishment. Studies suggest that social media have changed the way politicians communicate and how they try to appeal to their voters (Bobba, 2019). However, Woolley does not elaborate on whether, and if so how, computational propaganda has affected the political communication of the “traditional” political elite. Are politicians deliberately changing their political campaigns to tailor their messages for dissemination by bots and sock-puppets? Namely simple messages or polarizing content seem more suited for creating engagement on social media. The need to attract attention appears to be a significant motivation for today’s politicians, and automated political influencers are particularly equipped to help them achieve this goal.

The conclusion of the book provides recommendations for how the problem of computational propaganda should be mitigated and identifies the future of computational propaganda. Woolley highlights the potentially concerning trend of the increasing use of encrypted chat apps that provide ideal conditions for the spread of computational propaganda. Another concerning development is geo-propaganda which is a way of using digitally gathered location data to target specific groups for propaganda.

The book provides an excellent overview of what computational propaganda is, how it can be used, and which actors are implicated in this process. Woolley’s extensive fieldwork provides a good background for the discussions and gives insight into something that, for most people, remains an area that is difficult to grasp. While certain aspects of these discussions could have been further elaborated, the book is a needed first step in understanding the so far relatively underexplored dynamics of propaganda in the era of automation and anonymity.