The Costs of Connection

Book Review: The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism by Nick Couldry and Ulises A. Mejias

This book is a sobering view into a very real dilemma for every human alive today. In particular, Couldry and Mejias extend a particularly thoughtful interpretation of Karl Marx’s critical view of social, cultural, and economic life to how data is used to colonize in so-called “information economies.” At this point, I think it is important to point out this is one of many books on a similar subject, e.g. We Are Data takes a more loosely post-Marxist stance that I appreciated. However, The Costs of Connection holds a unique position in the data colonialism argument in that it uses theory with a lot more care and consistency than many of the other books I have read on the subject. It was only through this delicacy in theory that my critiques below could be made with clarity. Some books are so negligent with theory it is hard to know where and how to critique them, and this gives the reader a false sense of importance through perceived “novelty” that does not really exist. So for that alone, this book should be praised as theoretically significant to existing debates.

This book is also crucially of value not just because of its theoretical consistancy, but also because it connects debates about the social use of data to some theoretical lenses that academia holds dear. In doing so, this book makes itself indispensably placed to impact one facet or another of how critical data research views its own study. There are some ideas here which, at times, made me uncomfortable with some of my own chosen professional goals. In that sense, it would be reactionary if I said this book didn’t do its job. I finished the book reflecting on my own approaches to information studies with a critical lens. This book, I think, has made me a better person and a better theorist.

Given this praise, as I stated, there are some things I would critique about this book. I think at its core, it overstates how broad its criticisms can reach with ideological consistency. There are limits in application, while difficult to account for in a single review, are necessary to develop or else this argument could run rampant and become self-defeating in application. But first, allow me to outline the main contributions of this book, and then I will discuss some points that I believe suggest theoretical flaws in need of work.


As the book suggests, this is a modest but well-argued extension of Karl Marx’s critical approach to studying the exploitation of labor, but in this case, in relation to the labor of data production. Further, it argues that such exploitation is not new, but it is an expected cultural output of capitalistic tendencies towards exploitation which have gone farther back to capitalism’s historical roots in colonialism.

The book does an excellent job of outlining clear historical connections between colonialism and the current information economy of data production and related uses to predict and manage people. It also does a great job of outlining important moments exemplifying these points, and why they must be avoided.

This book also goes a step further and attempts to show that undoing this, a decolonialist project, will be difficult but necessary unless data colonialism is to “hollow out” all social life. That is, the developments of platform technologies have always been an extension of trying to datafy people and to use that data to gain profits they otherwise could not extract from people who are individually losing purchasing power. That is, data extraction is argued to be “raw” data extracted from people. The fact that individual subjectivities are not respected and the value of this data are never really negotiated under the “social contract” of capitalism, means saying “no” to data extraction is necessary, and projects need development to redeem individual power over these kinds of extraction. The book argues that many post-human projects have attempted this, but risk the deterioration of the individual. Therefore, we need to return to respecting the agency of individuals as full subjective actors in data economy. How this will occur is not clear, however this book moves towards providing a meta-schema of how to do this within a somewhat traditional Marxist lens that has been updated to include telecommunications technology.

In doing so this book suggests a few things of crucial importance:

  • Data are never “raw” in the sense that they have already been partially produced by us regular folk before they ever appear in a spreadsheet. This data exists partially as a trace of social culture being instituted within a data colonialist project of social extraction.
  • Data extraction when applied to managerial or policing projects is inherently data colonialism, and thus problematic.
  • Anyone using data to make decisions about society or the economy is systematically involved in this problem.
  • Data which objectifies decisions or the ethics of individuals hollow out the social by making it systematically a development bent on capitalizing on data rather than supporting social cooperation and culture. That is, it mechanizes what is not so naturally mechanical, and thus criminalizes or moderates anything outside of its expected profitable mechanism.
  • The institution of using external cloud storage for data, called the “Cloud Empire”, suggests simplicity in extracting data, and holding it away from the people who created it, and this is done under the argument that this is for the people’s own good and for the value of the products they use. However, this rationale is problematic in the same way historical colonialist projects were “civilizing” “uncivilized” countries.
  • Post-humanism can be done badly in the sense that it can be appropriated into capitalistic paradigms of data colonialism, so it should be considered carefully before used if at all.

I have framed this points to tell them in the way that I tend to interpret them as accurate critical approach to considering the use of data and theories surrounding them, however they are not exactly presented as such. Some ideas here are presented in ways that I find some places to grumble. Of course, to some degree this might be a stylistic choice in how I produce my research, and of course my biases on this topic might be leading me to grumble more than I should, but I hope I avoid coming across as reactionary, but instead offer a genuine response to some limitations with this sort of social analysis. Without further ado…

Interests of Concern and Development

My main opposition to this book, and many others like it, is that it falls for a particular error in understanding computational sciences. In particular, this book makes the argument that computational description or prediction is or is causally connected to data colonialist projects. While, certainly, data colonialism as described here requires computational description or prediction to exist, the other direction is not necessarily a fair criticism. Works like Data Feminism 1 are uniquely positioned to provide actionable counterarguments. This does not undermine the point of the book, however. It simply means that more nuance is required to explain what kinds of computation actively or potentially provide tools for data colonialism. There are some very clear examples of what does actively harm the social in respect to this book’s arguments. Civility research, an area of digital communication, is actively attempting to provide tools for social policing. Platform moderation techniques are an area of social management, and also are tools formed to provide capitalists with data which conforms to these platforms’ view of good and bad. However, that does not convince me that social network analysis which describes, for example, a structural understanding of communicative flows across Twitter are actively harming or “hollowing out” the social so much as the social metaphysics implied in computational interpretation might. That is, often it is not the computation itself that is actively contributing to developing tools of colonialism.

That said, the data which the scientist uses arguably was in some way accessed through data extraction technics well within the critics of the authors, and should be a concern of consideration. However, I would argue the same thing is true of qualitative technics of interpreting history in the way that this book does. If data extraction is fundamentally problematic in that it steals society of its essential mechanisms for action outside of capital, qualitative data extraction which provides us with histories do precisely the same. The distinction between the qualitative and quantitative in this respect seems to me to be a false dichotomy. Predicting or describing people outside of the systems of culture in which they live is still data extraction regardless. In order to escape this counterpoint, I think this book has to concede that it is not data construction itself which is the problem but the systemic arrangements in which capitalism provides for the labor of producing data. I do not see this book as inconsistent with this counterpoint in many of its main points, but it does suggest otherwise in order to maintain a critical Marxist perspective.

This brings me to another point, the insistence that ethical data production is a return to the subjective point of the individual, and that its necessarily the centralization of this data which is capitalist is a point which I strongly contest. It was, in fact, the Utopian Socialism which suggested the centralization of data as a process which solves all inequities in labor production. It was F.A. Hayek who argued against Utopian Socialists to provide why an economy based on such ideas was doomed to failure.2 This omission in history seems to undermine some of the methodological practices of the authors. That is, their history has a bias in interpreting what is real anti-capitalism which omits the point that it was the Austrian School economic perspective which was largely pro-capitalist which argued this point against utopianism in the Socialist Calculation Debate in the 1940s. In this way, the authors are actually coming full-circle. They are arguing against a more socialist strain of economics which exists more closely aligned with Mechanism Design. This is why conservatism to Marxist approaches to solving these problems often never truly escape the “utopian” capitalism the authors describe as the complete hollowing out of the social.

“Utopian” capitalism and utopian socialism/Marxism absolutely agree with each other in the mechanisms of data production and management. They fundamentally exercise the same metaphysical claims about information and data: i.e. cybernetic utopias. It is the ideologies of what to do with that data once they have it which makes them historically different. In this sense, this carefully chosen historical omission constructs an ideological consistency in their critical analysis which is inaccurate. In particular, data colonialism does not diminish simply because capitalists aren’t doing the extracting. Marxists, Communists, Socialists, and many other ideologies are fully capable of this themselves and have shown so. I can hear the “Real Communism has never existed!” response here. I hear you, but that undermines the historically material analysis of the book, so the comment stands. Merely holding and acting within the virtues of an ideology which opposes capitalism historically does not stop data colonialism.

Some alternative to traditional Marxism is necessary, contrary to the authors’ arguments is indeed necessary to move to decolonize us from data extraction. In this way, I don’t find it necessary to remain conservative to humanist (as opposed to post-humanist) ontologies of agency and power. In this book, the authors suggest that moving to use Latour’s Actor-Network Theory (and by extension Tarde’s ontology) are problematic because they “risk” dehumanizing individuals. However, as I pointed out, this has already been done within historical sociological arguments which are not too far away from the authors’ ideological position. If this assessment is fair, every ideological position in question “risks” the loss of agency of the individual. I do not entirely buy the argument that there are no other ontological positions of describing and critiquing the datafication of people. This seems to be more of an stylistic choice in academic preference than an appeal to a theoretically cogent point. While we all make choices such as these in writing or bounding our writing, this could have been more tactfully either with more clear justification, or with acknowledged limitations in knowing what post-Marxist ontologies do to the status of “human”.

All of my critiques, however, are minimal points which should serve more as warnings for the limits of such a theoretically dedicated perspective of human being. There is definitely a bit of importance to each of the points the authors have made of which I have critiqued here. For example, it truly isn’t entirely clear how post-Marxist ontologies risk dehumanizing. This should be considered a serious and significant warning! That said, I am not as convinced that they are any more serious than other critiques (including my own here) of the ideologies the authors have chosen.

For example, while Hayek might be considered taboo by those untrained in (post-)Marxist traditions of economics, looking at history of economic theory developed by Philip Mirowski and Edward Nik-Khah2 3 on the subject would prove insightful. While these authors are not within the school of thought as the authors here, they are “heterodox” economic theorists and historians which would suggest a possible synthesis that develops the authors’ critique more completely. In particular, their work outlines an interesting history of how Hayek is fundamental to modern socialist movements taken seriously in economic sciences, their limitations, and their lessons about economics as a science. Even if these Mirowski and Nik-Khah are ignored in their analysis of Hayek by those who chose to stay with different academic traditions, these arguments will never be resolved unless the criticisms of Hayek, Mirowski, and Nik-Khah are answered, and therefore, the economic motivation for data colonialism will remain stable.

Using such Mirowski and Nik-Khah might extend these arguments towards a more clear notion of how data and information fundamentally are connected in similar ways to both neoclassical economics and these socialist traditions, and how they can be escaped more clearly without appealing to dogmatic ideologies and metaphysical claims. Such a synthesis could help respond to the authors’ point that there are genuine decolonialist criticisms of postcolonialism focusing too much on culture and not enough on economics. Mirowski and Nik-Khah published a chapter on the limits of accepted economic science and media theory, because they have a mutual blind spot in explaining each other.4 Making such a blind spot visible would perhaps prove fruitful. If one finds the arguments of Couldry and Mejias compelling, Mirowski and Nik-Khah will offer points of inspiration outside of the ideological traditions here, and so I recommend them as a comparative reading on the subject.


Image Attribution: Alexander O. Smith

  1. D’Ignazio, C., & Klein, L. F. (2020). Data Feminism. MIT Press. 

  2. Mirowski, P., & Nik-Khah, E. (2017). The Knowledge We Have Lost in Information: The History of Information in Modern Economics. Oxford University Press.  2

  3. Mirowski, P. (2002). Machine Dreams: Economics Becomes a Cyborg Science. Cambridge University Press. 

  4. Beverungen, A., Mirowski, P., & Nik-Khah, E. (Eds.). (2019). Markets. University of Minnesota Press. 

You might also enjoy