Donald MacKenzie is a professor of sociology at the University of Edinburgh. He successively worked on nuclear weapons computing systems, on high-frequency trading algorithms, and his current projects are related to online advertising. After giving a talk at the SKAPE Centre entitled “Materiality, Policy and Power in Advertising Markets: Ad Servers and Header Bidding”, he accepted to do this interview for the SKAPE Centre’s blog. During the interview, Professor Donald MacKenzie evokes his project on the material political economy of online advertising and its scientific and policy implications.
Maxence Dutilleul: You are well known for your work on High Frequency Trading (HFT) and online advertising is a topic that you have tackled very recently. What led you to study this topic and what is the link with finance?
Donald MacKenzie: I basically came to the point with my HFT work where it was time to wrap it up and I was starting to sort of scan for other areas. I went to a talk by Paul Edwards, who is one of the people who has contributed to the literature on infrastructures, where he talk about online advertising. And nearly everything that he said I agreed with except for when he said that it’s a system that operated in microseconds (millionths of a second), and I already knew that was not so. Online advertising is a system that operates in milliseconds, in other words in thousandths of a second, not millionths of a second. Just that planted the idea that it might be interesting to take something that at some level looked similar to HFT (which operates in nanoseconds: billionths of a second) but was actually significantly different. It is really that excellent talk by Paul that got me starting to think about it. And then by very good fortune, one of my HFT interviewees had moved over to work in digital advertising and very helpfully introduced me to the field.
Maxence Dutilleul: You have a very original approach of digitalisation. Digitalisation can be thought as very virtual, but you actually emphasise its materiality by developing the notion of “material political economy”. Can you explain this notion to the people who don’t know your work?
Donald MacKenzie: This is, of course, one single idea, not three ideas, but the easiest way to explain is to go through the three words. First of all, “material”. As any computer scientist will tell you, virtuality is a far from cheap material creation. To make systems virtual is a tour de force of material configuration. So, not all the time, but at least some of the time it is worth enquiring into the material foundations of computer systems. This is very much the case in the world of HFT. If you are operating at nanosecond speed, you have to consider a computer as an assemblage of silicon, metal and plastic through which electrical signals pass. The job is to get them to pass as quickly as possible. The mechanism of transmission of information between different data centres becomes hugely important and a fibre-optic cable is too slow. Thus, you have to transmit through the atmosphere using, for example, microwaves. So that is “material”.
Then, “political”. I would say two things. One is that I am using the notion of material politics drawn from Actor-Network Theory (ANT), specifically from the work of John Law and Annemarie Mol. It refers to the different ways in which the material world and the material practices can be organised, and which ways become real and which get sidelined often has a broadly political aspect. But I also do not want to ignore other senses of politics. In my book on HFT, I trace the connections into the formal political system in the US, the structure of Senate committees and the like. In the case of online advertising, legal politics, for example, is important and, particularly, the interpretation of antitrust law is significant.
And finally, “economy”. In the case of finance, it seems to be almost a redundant thing, because finance is obviously about economy. But I particularly use the term to flag up the importance of routine money-making through things like fees. It is natural when you look at finance to think about the making or the losing of very large sums of money. Those are the newsworthy episodes. But actually, the financial system is a routine money-making machine. Sometimes I think that people in social sciences like myself, social sciences like sociology, have tended slightly to avert our eyes from the mundane grubby making of money and to be honest, I did that myself when I started to work on finance. And I have always admired the work of Olivier Godechot because he always kept firmly in mind the issue of bonuses and salaries and how people were remunerated and the consequences of that for socio-economic inequality more generally. So the economy in the notion of material political economy is designed to point in that direction and, as I have said, specifically to point to the importance of routine money making.
Maxence Dutilleul: You are a professor of sociology, but you speak very much of political economy. What is your position towards this very specific tradition and is it a reference to historical materialism?
Donald MacKenzie: Of course there are many flavours of political economy. I am not using the term political economy as a kind of polite way of saying Marxism, so to speak. I would not regard myself as being a Marxist even though I have been influenced by the Marxist tradition. For example, I have never been able to convince myself of the validity or usefulness of the labour theory of value. What I will point to, in relation to Marx, is volume I of Capital and in particular the history of the machine and the mechanisation of work. That is a wonderful exploration of the material politics of technology. I am actually being quite specific in how I use the notion of material political economy because all three words are important to me. What I am simply trying to do is to say – and there is nothing novel in saying this, but it is still important to say – that there are many different ways of doing politics. The organisation of the material world and the material practices is one way of doing politics.
Maxence Dutilleul: You have also mentioned the ANT and in the meantime, you speak about the social struggles and I know that you draw out some things from the field theory in the Bourdieusian sense. What is your position in the clash between these two theoretical traditions and what is the role of materiality in it?
Donald MacKenzie: That is an interesting and a complicated question. I don’t have to tell you that there was Parisian academic micropolitics in the clash between Bourdieu’s field theory and ANT. In particular, there is a kind of shocking ad hominem attack on Bruno Latour from Pierre Bourdieu. So you know the two schools were in tension for reasons that were in part local. But I don’t think one could accuse Bourdieu himself of ignoring materiality because, after all, he worked on the Algerian house. It is more the way in which field theory has gotten adopted in wider sociology. One of the central tenets of field theory is that there are different positions in the field, and some are more advantageous than others. The natural way for the sociologist to approach that is to think of the positions as socio-economic roles and, of course, that is very important. But if you think about HFT, there is a sense in which the advantageous position may be a spatial position rather than a socio-economic role. If you are transmitting signals from one data centre to another via microwaves, you want your microwave towers to be along the geodesic route between the two data centres because that is the shortest route on the surface of the earth. That means that particular physical spatial locations become very important. That, in a sense, captures the fact that positions can be quite literally spatial positions and not just socio-economic roles and that is sort of a long-winded way of saying that we need to keep both in mind.
For sure, there are ontological differences between Bourdieu’s field theory and ANT, but sometimes I think that one does not have to be obsessed with these sorts of things. If you think about theory as a sense of empirical sensitivity, two positions that are, in a sense, ontologically different can still productively point you to important empirical phenomena.
Maxence Dutilleul: I feel that there is a sense of materiality as a limitation to virtuality…
Donald Mackenzie: What I would actually point to, in terms of virtuality and materiality, comes from the history of computing. One of the important developments was what is called “virtual memory”: if you have a multi-user computer system, you do not simply allocate fixed memory space to each user, but you set it up so that from the user’s point of view, it looks like they have infinite memory. They do not have to worry about the physical bits on a spinning disk. So it’s virtual in that sense. But creating virtual memory is a material achievement, you have to configure the material world in sophisticated ways to make that happen. Typically if you take any phenomenon that you think of as virtual, at least in the world of computing, and you dig into it, you find that there are specialized and often expensive material configurations that are making virtuality possible.
Maxence Dutilleul: And speaking of material and mundane limitations to virtuality, what can you tell us about the resistance against online advertising?
Donald MacKenzie: There is resistance to online advertising of a very mundane kind. Users can install ad-blockers whose goal is to prevent any ads at all from being shown to them. There is also the politics of privacy, which is a resistance to data sharing of an illegitimate kind. And this is a very important thread in the history of online advertising. The case of header bidding is rather different because, to a substantial extent, in those two previous domains, ad blocking and privacy, the battle is typically an insider/outsider battle, whereas header bidding is a conflict within the world of online advertising about how the crucial auction systems should be organised.
Maxence Dutilleul: How do you see the link between these very mundane resistances and the formal state regulation of privacy?
Donald MacKenzie: One can argue that moves within the industry, particularly the moves by Apple have probably actually been more consequential than regulatory efforts such as the GDPR because those efforts easily become a sort of policy of consent. Anybody who uses computer systems in Europe is driven mad by the need to accept cookies every time. Of course, it would be an unusual person who does not just take the quick and easy option of accepting cookies because the other option typically takes you a few more seconds. Thus, Apple’s do-not-track initiative is given force by the fact that the designs of apps need to be in Apple’s app store. And Apple’s app store is an estate so to speak, to use a term from the middle ages: it sets the rules and even big and powerful corporations such as Facebook have to follow the rules if they want to be in the App Store. When you are presented with one of Apple’s screens, it is designed in such a way that it is just as quick and easy to say “No, I do not want to be tracked” as it is to say “Yes, I am ok with being tracked”. It is not like the consent for cookies which is a kind of asymmetric kind of process. And it has real effects! If you click “Do not track” it does not mean that you will not be tracked, you will be tracked by other means. But it still translates into significant constraints on the extent to which advertisers can track what you’re doing, less I think in the targeting of advertising than in the measuring of whether or not advertising is being successful. It is arguable that an initiative such as that within the industry, that is backed by what is, in essence, a monopoly kind of power that Apple has, has got a bit more force than government regulation.
Maxence Dutilleul: And how would a policy of materiality look like in your point of view?
Donald MacKenzie: It is difficult to say in the abstract because I am struggling to think of a very clear and convincing example of a fully successful policy of materiality in domains of the kind we are discussing. However, let me actually revert for a second from advertising to high-frequency trading in finance, where there is a very interesting proposal that regulators could impose that you should shift from continuous trading (which underlies nanosecond speed races) to very frequent, but still “single-point-in-time” auctions. This has been proposed by the University of Chicago economist Eric Budish. The rationale of the proposal is that there is waste in the HFT system because of what are essentially zero-sum speed races between participants. They must spend money on the technology to be fast, but there is no sort of net additional benefit because the increments in speed are tiny and not meaningful in any broader economic sense. So Budish’s proposal is designed to eliminate – or at least reduce – that form of waste. And that is something that, if they so chose, regulators could impose. I do not actually think that they will. It may actually curiously happen as a result of Google getting involved in the financial system. As I mentioned in the talk they have invested in the Chicago Mercantile Exchange (CME), which is the very heart of the world of HFT, and the CME’s systems are going to be moved into Google’s cloud services. This is a horrendously difficult issue from the viewpoint of a software engineer because tiny speed increments are hugely important in HFT, and how can you achieve any kind of guarantee, as it were, of equal speed, when speed is measured in nanoseconds? It is pretty tough. So I think it is not at all impossible that Google software engineers will see the virtue from a software design viewpoint of Budish’s scheme.
Donald MacKenzie is a professor of sociology at the University of Edinburgh. His research explores the intersection betweek the STS and economic sociology. He has had a considerable influence over STS in the last decades. He has received numerous distinctions for his research, the most recent one being the Leonardo da Vinci medal awarded by the Society for the History of Technology. His latest book, Trading at the Speed of Light, was published in 2021 by Princeton University Press.
Maxence Dutilleul is a visiting student researcher at SKAPE under the mentorship of Nathan Coombs and Donald MacKenzie. He is also affiliated with the ENS Paris-Saclay. His research relates to money and central banking in Europe and Western Africa.
Paul Edwards is a professor emeritus of Information and History at the University of Michigan and WJ Perry Fellow at Stanford University. Il a notamment contribué à développer l’idée d’infrastructure épistémique et travaillé sur divers sujets liés au cyber, à l’expertise et à la gouvernance.
See for instance: John Law, Annemarie Mol, Globalisation in practice: On the politics of boiling pigswill, Geoforum, Volume 39, Issue 1, 2008, Pages 133-143, ISSN 0016-7185.
See Donald MacKenzie, Trading at the speed of light: how ultrafast algorithms are transforming financial markets, Princeton University Press, May 2021.
See Pierre Bourdieu, Science de la science et réflexivité : cours du Collège de France, 2000-2001, Raisons d’agir, 2001, pages 21-24 and 47-66. Bourdieu very toughly criticises what he calls the “new sociology of science” for being fully relativist and for not taking into account the specificities of the “field of science”.
General data protection regulation
Eric Budish is Paul G. McDermott Professor of Economics and Entrepreneurship at Chicago Booth. See for instance, Eric B Budish, Peter Cramton, John J Shim, “The high-frequency trading arms race : Frequent batch auctions as a market design response”, Quarterly Journal of Economics, Volume 130, Issue 4, 2015, Pages 1547-1621.
Cover photo credit: Photo by Google.