In an interesting new piece, Jim Thatcher, David O'Sullivan and Dillonn Mahmoudi propose that big data functions in the context of capital as “accumulation by dispossession,” which is David Harvey’s term for what Marx called “primitive accumulation,” the process by which capital adds to its wealth by taking goods from others and adding them to the system. Marx: “so-called primitive accumulation, therefore, is nothing else than the historical process of divorcing the producer from the means of production” (Capital I, 875 [I am using the Penguin edition]). Perhaps the best example of this is the one detailed by Marx: the enclosure movement in England involved the privatization of agricultural common spaces in England, such that it was no longer possible to graze sheep on lands held by the community in common; the result was that a lot of peasants, who ended up with no or inadequate amounts of private property, lost everything of value they had and became “free labor,” forced to sell themselves to the emerging factories. As Marx sums up the process:
“The spoliation of the Church’s property, the fraudulent alienation of the state domains, the theft of the common lands, the usurpation of feudal and clan property and its transformation into modern private property under circumstances of ruthless terrorism, all these things were just so many idyllic methods of primitive accumulation. They conquered the field for capitalist agriculture, incorporated the soil into capital, and created for the urban industries the necessary supplies of free and rightless proletarians” (895)
I am very sympathetic to the thesis, and there is something profoundly right about it, insofar as Thatcher et al. rely on the separation of the valued information from the person who produces it. But I also think it needs tweaking, for reasons that emerge in the paper itself: the data trail that a person leaves is generally itself without value, and only becomes valuable when aggregated with a lot of other data. In other words, as I tried to argue a while ago, data is itself without value; it is only when it becomes information that it realizes that value.
It seems to me that the accumulation processes of big data is involved in a much earlier stage, the commodification of data into information itself, which involves both the elevation of exchange value over use value, and the conversion of qualitatively different items of data into commensurable units of information. These are, to an extent, equivalent processes, as Marx notes: “as use values, commodities differ above all in quality, while as exchange values, they can only differ in quantity, and therefore do not contain an atom of use-value” (128). Still, I think it’s worth teasing the two threads apart here.
First, on the reduction of quality into quantity. In Marx, this is the conversion of labor into labor power; without this conversion of qualitatively different objects and processes of labor into something that could be measured by a consistent standard, capitalism could not function. Marx:
“The labor that forms the substance of value is equal human labor, the expenditure of identical human labor-power. The total labor-power of society, which is manifested in the values of the world of commodities, counts here as one homogeneous mass of human labor-power, although composed of innumerable individual units of labor-power. Each of these units is the same as any other, to the extent that it has the characteristic of a socially average unit of labor-power and acts as such, i.e., only needs, in order to produce a commodity, the labor time which is necessary on an average, or in other words is socially necessary. Socially necessary labor-time is the labor-time required to produce any use-value under the conditions of production normal for a given society and with the average degree of skill and intensity of labor present in that society” (129).
That is, in order to put an exchange value on something, you need a system of equivalences. Capital – and Marx is thinking of factories here – uses the average amount of time it takes to produce such a thing under current social/economic conditions. This then becomes the primary determinant of the object’s value, and the fact that one object might involve digging around in the mud and another putting together electronic components isn’t really relevant. Or, as he puts it more summatively, “what exclusively determines the magnitude of the value of any article is therefore the amount of labor socially necessary, or the labor-time socially necessary for its production” (129).
Certainly there are things to object to here, particularly with the decline of Fordism, but I want to emphasize that this conversion, as Mary Beth Mader reminds us in a somewhat different context, is an ontological one; it is no accident that Marx says that the analysis of commodities shows them to be “very strange thing[s], abounding in metaphysical subtleties and theological niceties” (163). That is, there is a fundamental difference between human labor, in the sense of making a nail, or planting a seed, and the same thing expressed as the expenditure of socially averaged time. So one of the things that capital does in converting data into information is to convert it into a 21c equivalent of labor-power: information, which can be priced. When data becomes information, i.e., “as soon as it emerges as a commodity, it changes into a thing which transcends sensuousness” (163). That is:
“The mysterious character of the commodity-form consists therefore simply in the fact that the commodity reflects the social characteristics of men’s own labor as objective characteristics of the products of labor themselves, as the socio-natural properties of these things. Hence it also reflects the social relation of the producers to the sum total of labor as a social relation between objects, a relation which exists apart from and outside the producers” (165).
Let’s take the same quote, and substitute data and information:
“The mysterious character of data commodified as information consists therefore simply in the fact that the information reflects the social characteristics of men’s own data as objective characteristics of the data itself, as the socio-natural properties of that data. Hence it also reflects the social relation of the producers of data to the sum total of data as a social relation between information, a relation which exists apart from and outside the data."
Whatever value the data had to the person is converted becomes inexpressible except in market terms of monetary value. Consider, for example, privacy. A consistent theme in the literature on privacy is that it is very hard to say what it is and why individuals value it (given the Marxian framing, I think privacy is a good topic because it circumvents, at least to an extent, Baudrillard’s critique that Marxism fetishizes use-value). As I argue elsewhere, the data industry approach to privacy is one that encourages users to treat their private information in commodified terms, as something they might exchange for some other consumer product. But this commodified version of privacy makes it impossible to understand what the privacy meant to the person in question: is it dignitary? Does it preserve her relation to her father, as in the case of the pregnant teenager who received pregnancy-related advertising from Target – before she told her parents the news? Even more significantly, this process makes the social value of privacy completely invisible.
This conversion of use to exchange value does, therefore, seem like a place to talk about primitive accumulation, because individuals may very well lose things of importance to them in the process of data collection. For example, the mysterious processes of credit (and other) scoring systems, which aggregate vast amounts of data about people, might make it impossible to obtain financing to buy a house, when previously one could have done so. Absent the protections of the GINA, information that one has the BRCA1/2 genetic mutation that increases the risk of breast cancer might make the individual in question uninsurable, when previously she could have obtained coverage. Even the credit ratings of one’s friends have predictive value.
The other thing that commodification does, strongly suggested in the Marx passage on commodity fetishism, is to deracinate the data, separating it from whatever context it originally occurs. Helen Nissenbaum has written about this in the context of privacy, noting that information cannot be understood without reference to the context in which it is disclosed, and that sometimes privacy harms occur precisely when information is taken out of context. This is a general feature of commodification, and I have argued elsewhere that the commodification of locally produced foods into trademarked geographical indicators can affect the social relations of production in ways that both obscure the actual origins and production of the food, and dispossess members of the cultures in question of their own variant versions of the good. But there are other examples as well: as Shannon Winnubst details at length, the branding of “coolness” is above all a process of deracination, where a resistant African-American social practice becomes a formal expression of a personal brand, an expression that can be made without any reference to the social struggles behind it: in the 1950s, coolness was a form of black resistance politics. In this way, politics becomes invisible, and the entire social field becomes legible only in economic terms (I take it this is the core of any successful definition of neoliberalism).
One way all of this manifests itself, quite insidiously I think, is on the topic of “revealed preferences.” Recall that on a Hayekian account, the virtue of markets is that they show what people prefer, based on what they buy. This account is too simplistic, of course: I might prefer to buy a Porsche but end up buying a Kia instead, because I can’t afford the former. Thus my “preference” will be coded as Kia over Porsche. But that’s clearly not the case – it’s that I had to fit my preferences into my budget; that information isn’t visible in the purchase decision. On a first glance, big data seems to be designed to solve this problem: if we aggregate enough data about me, then my real preferences will become clear. But of course there’s problems here too, not least of which is that it will be very difficult to tease those out except by correlating my purchase with people otherwise similarly situated to me (and it can’t be by socio-economic status, because they’ll all buy cheaper cars for the same reason), and the product will have to be some sort of risk-preference hybrid: the model says there’s a 52% chance that I actually prefer a Porsche. But if I never buy one, then this prediction is suspect. How can you account for a negative? And, even if we arrive at the conclusion that there is a 52% chance I prefer a Porsche, based on some marvelous discovery about people “like me,” then there’s still a pair of problems. The obvious one is that it’s not clear how this information is useful, except insofar as Porsche might want to quit advertising in venues where people like me might be found.
The other problem is that we now have to know how the determination of “like me” is supposed to be generated. What we definitely know is that it will have to exclude any number of important characteristics of me the concrete individual – this is Marx’s point about the conversion of labor to labor-power, and Mader’s point about the ontological difference of norms. In that sense, the deracination of relevant information makes my preferences inscrutable, because they are no longer about me. They are about a profile, and profiles and representations always are incomplete. All of this undermines the putative epistemic value of markets.
In terms of primitive accumulation, I think what all of this means is that one thing I am deprived of – in other words, an example of accumulation by dispossession internal to the commodification of data into information – is the capacity to form and express my actual preferences. And that is significant because the more our “preferences” are restricted to those things that we can buy, the more our lives are determined by market logic – and the more things that exceed market logic become invisible. Or, worse (and in a Foucauldian, not particularly Marxist, vein), we only attach value and ourselves to things with market value. As Julie Cohen notes, the problem goes well beyond a simple invocation of context, where even scholars such as Nissenbaum “appear content to address context while holding the self constant, thereby ignoring the problem of emerging subjectivity and its relationship to contextual change” (19). If my subjectivity is limited to market choices, then my emergence as homo economicus becomes a lot more likely. I am dispossessed of different ways of living - not just my ability to actualize them, but to even understand or formulate them.