From the Manufacture of Consent to Dataism

One of the collateral shocks of the Great War was the collective discovery of propaganda as a tool for mobilization through the media. Marc Bloch wrote, a few years after returning from the front (2014, p. 107)¹:

It is time to open a serious inquiry into the false news of the war, because the four terrible years are already receding into the past and, sooner than one might think, the generations who lived through them will slowly begin to disappear.

Between the two World Wars, the United States became the laboratory for the manufacture of consent (Lippmann, 1922), where those whom Edward Bernays (1947) called the engineers of consent worked and were trained.

A century later, between 2016 and 2018, the discovery of state-sponsored troll factories, such as the Russian Internet Research Agency (IRA), and the Cambridge Analytica scandal, brought renewed attention to some of those themes: propaganda, which has become computational, and the massive use of data and knowledge to guide (or manipulate) public opinion.

What are the characteristics of the new consent factory, compared to the “classic period” of propaganda and “administrative” sociology?

The Manufacture of Consent and Its Critics

The protagonists of this “laboratory” were Bernays himself, Harold Lasswell, George Gallup, and Paul Lazarsfeld, who contributed to the birth of “administrative” social research (as Lazarsfeld himself defined it: 1941), which was organized with bureaucratic-corporate criteria, aimed at answering practical questions, and producing useful knowledge for institutions².

Charles Wright Mills, in his 1959 essay The Sociological Imagination, underscored the consequences of the bureaucratic ethos of this type of research and its dependence on funders. Adorno’s critiques of the use of statistics in the culture industry and of “abstracted sociology” are also well-known.

In Blumer (1956) as well, the critique of the “variable” was not (only) about quantification as such, but about the idea of engineering consent and of the public as an indistinct mass, that is, about Lippmann’s phantom public (Lippmann, 1925): the supposed “objectivity” of the variable is an artifact that obscures the interpretative processes that constitute the social world, and through which issues become meaningful to the actors.

The subsequent debate between quantitative and qualitative methods, which Mills would have likely dismissed as a dispute between academic cliques, has perhaps obscured the issue of the relationship between social sciences and power, and the role of the intellectual.

The Power of Data: Lessons from the Past

The Role of Platforms

Data are inherently technocratic in nature as they require specialized institutions to be produced, organized, and disseminated.

However, if in the twentieth century the generation, collection, and dissemination of data were almost the exclusive prerogative of the state, universities, or institutes conducting independent or commissioned research, today the weight of these institutions has diminished compared to that of platforms.

The new (real) “engineers of consent” are increasingly data scientists and actual engineers working for Meta, Google, TikTok, and X³.

Research has found itself in a position of dependence on platforms that can change the rules of access to data at any moment (Venturini & Rogers, 2019).

Computational politics, moreover, is no longer that of Cambridge Analytica: platforms—and their owners—now play an even more central role, as, with the progressive closure of APIs, they monetize data that is now largely unavailable externally, offering campaigns over which even the clients themselves have little control.

Data: From Collecting to Harvesting

The first aspect I would like to emphasize is that the attention devoted to the research process, from defining research questions to constructing variables, which characterized the “administrative research” of Lasswell and Lazarsfeld, contrasts with the epistemological assumptions of modern “dataism” (Van Dijck, 2014) and computational research practices.

This contrast can be well represented by the expressions used to indicate data collection: data gathering (or collecting) and data harvesting. Collecting implies an intentional, often ex ante, and in any case, theoretically oriented organization of data. Harvesting, on the other hand, refers to the “reaping” of data that “has been generated” in online processes: large, fast, and unstructured.

The great surveys made a fundamental contribution to the development of social theory because it was the problems (no matter whose they were) that guided the research, and the data were ‘constructed’ or ‘chosen’ in light of theory.

Today, by contrast, data is collected automatically, and problems are dictated — if not by what has already been harvested, then by what is harvestable. “Engagement,” “virality,” and “user networks” are studied because that is the data that is offered (Venturini & Rogers, 2019), accepting the platform’s ontology, and risking to miss the symbolic and relational dimension, the context that supports beliefs and fake news, or the political economy of the platforms that makes such processes possible.

This is precisely Blumer’s “lax choice of variables” (Blumer, 1956) or Adorno’s “fetishism”. The transition to this new century seems to have confirmed the most pessimistic predictions of Mills, Blumer, and Adorno.

Beyond Techniques: Social Sciences and Power

Reducing the debate from methodological to “technical” (quantity/quality) is therefore a classic example of depoliticization, or, in Adorno’s terms, of “mystification”. The question is the one posed by Lynd in 1939, namely Knowledge for What?, which José van Dijck (2014) re-proposes as: “Useful knowledge, but for whom?”.

Big data can be used to deconstruct platform processes, while we know that the most effective computational propaganda combines automation with “participant observation”: trolls, after all, perform ethnography of social networks to better insert themselves into online communities and cultures, and to create and spread effective disinformation content.

The difference, to quote Mills again, lies entirely in the “type of practicality” at whose service techniques and knowledge are placed: the “illiberal practicality” that engineers consent, or the “liberal practicality” of tools for freely understanding and debating, which was the idea of public and democracy that Blumer defended with his critique of the variable.

References