Software tunnels through the rags ‘n refuse
Text of talk given at the Platform Politics conference in Cambridge 13.05.11
It took New York police officer, William Barker two hours to find Homer Collyer dead in his apartment in March 1947. Barker had to crawl through a window into a second-story bedroom, burrow his way through newspaper bundles, empty cardboard boxes lashed together with rope, the frame of a baby carriage, a rake, and old umbrellas tied together, folding beds and chairs, half a sewing machine, boxes and parts of a wine press. For the next two days police continued to search the house, literally finding their way through 25,000 books, a horse’s jawbone, a Steinway piano, an early X-ray machine, baby carriages, a doll carriage, rusted bicycles, old food, potato peelers, a collection of guns, glass chandeliers, bowling balls, camera equipment, the folding top of a horse-drawn carriage, a sawhorse, three dressmaking dummies, painted portraits, human organs pickled in jars, the chassis of a Model T Ford, tapestries, hundreds of yards of unused silks and fabric, clocks, 14 pianos (both grand and upright), a clavichord, two organs, banjos, violins, bugles, accordions, a gramophone and records, and countless bundles of newspapers and magazines. Oh and 130 tons of garbage.
Rodinsky’s room was also piled high with material. While it was not as overwhelming as the Collyers’, when the door to 19 Princelet Street in London’s Spitalfields was opened again in 1980 after over 11 years the redevelopers were met with material stuff: newspapers, books and papers, gramophone records, clothes and an A-Z marked with obscure journeys into the London suburbs, scraps of paper and sweet wrappers, all covered with indecipherable scribblings in many languages as well as a half-finished cup of tea and a pot of porridge still on the stove. What followed was another detective story as Rachel Lichtenstein pieced together the life and disappearance of David Rodinsky and Iain Sinclair traced his wanderings across London from the material objects he left behind.
What unites these two stories is the way in which the Collyer brothers and David Rodinsky were positioned or even recreated as governmental subjects through their material objects, the rags ‘n refuse they collected, hoarded or archived. They became targets of police reports, medical and mental health professionals as well as journalists, artists and writers who read their lives from their stuff and positioned them as subjects.
Every twenty minutes Facebook adds more stuff to its collection:
- 1 million links
- 1.4 million event invites
- 1.9 million friends requests accepted
- 2.7 million photos, 1.3 million of which are tagged
- 2.7 million messages sent
- 1.89 million status updates
- 1,6 million wall posts
- 10.2 million comments
This digital stuff is housed in at least 9 leased data centres or server farms, each around 35,000 square feet. Facebook is currently building its own centre which will be 307,000 square feet. These quaintly named ‘farms’ house 60,000 servers and cost in the order of $50m a year to run.
Google is notoriously secretive about its hoard of data. What we do know is that it spent $757 million on its seven data centers in the third quarter of 2010 and that those centres process twenty petabytes of data a day. Google’s hoard, like Facebook’s includes our digital detritus – our email messages, our YouTube videos, our Picasa pictures and Blogger postings as well as 1 trillion cached webpages. Those farms also house the digital footprints we leave as we use Google’s services – our logins, IP addresses, search terms and histories, maybe our creditcard details in Google checkout and records of the ads we clicked, the times and journies we made.
Like the Collyers and Rodinsky, Facebook and Google hoard digital objects but unlike those real-world hoarders, the digital recluses also generate new data, new digital objects as they work. Their algorithms burrow through that data like a police patrolman or a researcher, tracing clues, forming connections, building pictures, but unlike those real-world investigators, Facebook and Google’s algorithms create new data objects – connections between data files, between friends, searches and adverts, between activities and objects. And that new data is fed back into the archive, ready to be searched, found and connected again.
What is important to note is that those data connections are also governmental data objects. Just as the Collyers’ and Rodinsky’s rags ‘n refuse became pieces in constructing their subjectivity for media, law and social service systems, so the digital detritus we leave for Facebook and Google, and that they in turn generate from that rags ‘n refuse, construct us as data objects and targets, ‘friends’ or demographics, healthcare risks or subversives. This goes beyond the issue of privacy of individual data objects to a wider field of governmentality through data trails and software-generated connections and subject positions. Even if our personal data is never released, even if we remain ‘anonymous’, the unhuman software patrolmen that burrow through the digital archives create a picture of us as part of a social graph or an aggregated search community. Whether these data subject positions are ever sold on to advertisers or insurance companies or subpoenaed by the state, they remain our social CV, our digital subjectivity. Whether those objects and traces are ever seen by human eyes is irrelevant, they remain data connections and data objects.
As an example, Facebook is rolling out facial recognition where a photo added to the hoard will be processed, and suggested names or tags presented to the user who, on clicking on one, will of course create another data object just as she does when clicking on a Like button as Anne and Carolyn discussed yesterday. Only yesterday it was announced that Facebook will ‘allow’ users to tag Pages effectively tagging brands and objects. These data objects and trails, the photograph, the record of searching for, tagging or liking the photograph, can be seen as material in the sense which Jane Bennett talks of “vibrant matter”, “quasi agents or forces with trajectories, propensities, or tendencies of their own”. Here the magnetic traces on the data storage media that Matthew Kirschenbaum quite literally dissects in Mechanisms, whether what we commonly understand to be an object (e.g. the image file-object) or what we could call a weird object (the Like trace-object, the tag or tagging object/event), are in Bruno Latour’s terms, actants, doing things in the world.
I want to present some tentative thoughts towards an account of these data objects.
Graham Harman’s object-oriented philosophy draws on Bruno Latour and Alfred North Whitehead, positioning both as “philosophers of concrete, individual entities”. The problem for Harman is that they do not go far enough. For Latour objects derive their power and presence from their relations or alliances. For Whitehead they are moments of becoming. For Harman any move away from a strict actualist focus on the object to either advocating a second realm of objects (the “eternal objects” of Whitehead) or a realm of potentiality beneath objects (the “plasma” as Latour speaks of it in Reassambling the Social) is a mistake. For Harman there are objects. That is it. Change happens in the world not as objects become and perish or enter new relations but as they connect. Connections are different than relations because whereas the latter happens outside the object – in a network, a “quasi-plenum” or realm of potential, the former is a matter of objects. There is no need in this framework for the object to perish or for the relations to be pushed to an outside. There is no need to take one’s eye off the object-ball. Rather the flux or mesh of objects (the assemblages, media ecologies, networks, hyperobjects or whatever other term we use) can be addressed as a matter of the objects themselves.
To bring this back to the digital hoard: The data-objects (the photos or credit card details); the data-mined objects (the Friends connection or clickthrough trail) and the datamining objects (the algorithms burrowing through and creating new data), all of these can be seen as objects circulating in and through Facebook and Google’s archive-hoards. All three actualists would perhaps see those files, database entries and software agents as objects, entities in the world. Latour might see them as constituted by their relations with other actants in the network: hardware servers, other software, engineers and lawyers, company business plans and competition legislation. Whitehead might see them as a series of occasions, discrete instants of becoming and perishing, as occasions of data connection.
Harman however would see them as objects that are not “exhausted by their relations to other objects”, that withdraw from view and have an existence outside of their connections with other actants. This is his fourfold object of real object, real qualities, sensual objects and sensual qualities. Where Latour puts the emphasis on the network (relations) as what gives the Facebook wall photo or an algorithm its presence and its power and Whitehead would stress the transience of the Google image search, Harman would put the emphasis on these objects, as more than their relations, contexts and becomings.
Some argue that Harman sees objects in isolation only, that he somehow refuses to entertain the idea that they connect. Far from it. Harman’s whole philosophy is built on trying to understand how objects connect. He just wants to understand that connection as a matter of objects not of an exterior relationality. For Harman, objects encounter each other in the heart of another object. A real object encounters a sensual object in the heart of a new real object.
We can draw the way that the photo on someone’s Wall and the Facebook facial recognition algorithm connect at the level of objects rather than by recourse to some meta-framework of network or capitalism or globalisation. And this perhaps allows a new form of politics. We can talk of two objects: the face-recognition algorithm and the photo uploaded to a user’s Wall and Facebook’s hoard. These objects could be seen as related within a field of network politics, info-capitalism and so governmentality. Alternatively they can be seen as connecting – however briefly – within the ‘molten core’ of an object we could call the “Facebook image data object”, the key to Facebook’s business plan, the site of governmental surveillance or subjectivity and the target of critical politics.
In this perspective the real algorithm can never encounter the whole photo-object, it connects with only a dimension of that object, with a sensual object. Its connection in the moment of posting on the Wall does not exhaust what that photo or database-object is or does – the photo data-object has a position as object before and after, it is a site of other connections and workings and it is being constantly reconstituted as new searches are conducted around it, new datamine connections made or the data copied or moved from one server farm to another. At the same time the real photo-object encounters the facial-recognition algorithm (within the Facebook software). In fact it encounters an intentional image of the algorithm-object, a particular instantiation. It encounters a moment of running, a particular position of the code. It does not encounter the complete complex reality of that algorithm in terms of its history, the political, legal and business battles around its creation, or continuing work, its nature and other connections. It does not need to. It needs to encounter what is necessary to establish the database data point – the clue the governmental patrolman needs to establish subjectivity or Facebook needs to targets ads. This encounter happens not in some extra space or context, in Latour’s plasma or Whitehead’s space of becoming or even just between objects in trials of strength. It happens within another object, what we might call the “Facebook image data object”, a real and specific instantiation of governmental power.
In one sense this is a form of nested objects but it is important to emphasise that these are not nested in any hierarchical let alone value-laden sense. There is no sense in which objects connecting with other objects should be seen as leading to a foundational macro or micro object. This model not only refuses to leave the object but also refuses to find the single object. There is no Facebook-object or Surveillance-object or Capitalism-object that acts like ‘context’ or ‘relation’ as foundation for all connections. Nor is there some machine code-object or electrical charge-object that can stand in for a founding object or fundamental particle.
Latour would at some level agree. He would of course insist that networks whether capitalism or the Twittersphere should be approached via objects or actants. The difference is that for Harman that can and should be done at the scale of objects not relations. Similarly Whitehead would perhaps argue for a focus on specific events not the abstract. But again the difference is that for Whitehead, as Harman says: ‘actual entities “perpetually perish”’. They do not lie behind their accidents, qualities, and relations like dormant substrata, but are ‘devoid of all indetermination’. For Harman this is ontologically problematic. For me it closes off potential for mapping objects, their connections and the spaces for exploit.
The advantage of a non-relational object-oriented approach is (of course) fourfold: It not only allows an escape from macro/micro-reductionism but it also provides a way of escaping the problem of the subject. It allows us to talk of essences and technological determinism without a sneer; and finally it enables us to open up Exploits for intervention.
Firstly this perspective escapes correlationism, Quentin Meillassoux’s term for the tendency to see everything in terms of the human-world connection. From this perspective there is no world without the human nor human without the world. It is this separation (yet partnering) of subject and object that drags us away from focusing on objects, their connections and their working. In terms of data-objects, correlationism demands we address images, algorithms and the Facebook database in terms of the humans using or at least thinking about them. At the very least this means it becomes difficult to explore machine vision systems such as face-recognition where computers ‘see’, ‘file’ and ‘analyse’ with no human intervention, a situation an object-oriented approach could happily discuss in terms of a photo object connecting within face-recognition object within a surveillance-image-evidence object.
Secondly, an object-oriented approach allows us an account of ‘essence’ that does not close off debate, connections, change and power. For Harman: “there is no avoiding a concept of essence”. That does not mean that essences are eternal or natural. “To defend essence is not to conspire in a sinister plot by the Party of Reaction. It is nothing more than to insist that objects are not exhausted by the relations to other objects” he says. What we experience as essence is the outcome (or emanation as Harman calls it) of the tension between the object and its qualities. There are things about a table, a photograph or even an algorithm that are ‘necessary’ for it to be that table, photo or software that works. But these qualities are not identical with the object. They do not exhaust it. This is significant because it means we can talk of seemingly insubstantial data-objects such as searches or click throughs as things. We can say: “yes there is a data-mined object” and then trace its connections within objects. We can use that essence as a space for Exploit. An object-oriented essence is a starting point not an end.
Even more controversially perhaps, this rescuing of ‘essence’ allows a similar embracing of ‘technological determinism’. As Geoffrey Winthrop-Young puts it: “to label someone a technodeterminist is a bit like saying that he enjoys strangling cute puppies”. An irreductionist, object-oriented reading of essence however allows us to say: “yes technology determines”. The issue become how that determination is drawn in terms of causality or what DeLanda calls “catalysis” for instance. But leaving that debate to one side, the issue is that again an object-centred approach can explore determinations as connections within objects rather than as reflections of something more basic, foundational or powerful. It allows us to proudly and openly say that the connection between an image-file-object and the Facebook algorithm (within the Facebook image-object) does things.
Finally, the escape from the subject, from the context, the relation, the continuum and the occasion – the focus on the object – allows a space for what Alexander Galloway and Eugene Thacker call the Exploit and it is here where an object-oriented approach to software and media meets politics. An approach to the computational/governmental space based on objects not networks or relations, changes the focus of struggle and change.
For Galloway and Thacker, struggle operates at the level of objects – in their case protocol. Struggle “must not be anthropomorphic (the gesture, the strike); it must be unhuman (the swarm, the flood)”. A virus does not fight a system, it overwhelms it. That struggle must be seen not as resistance but as “hypertrophy”.
Viruses or distributed denial of service (DDOS) attacks do not resist software they push it until it breaks. They clog up the server with too many requests, overloads, spam. But a DDOS attack can be seen as working not by simply overwhelming a network but by reconnecting objects (the http protocol, server requests, customers details etc) within the target object – in the case of recent Anonymous actions for instance, objects such as the PayPal or Amazon S3 object. Here an object-oriented approach of seeing and working with objects connecting within objects, rather than a field of relations, open up political potential.
In a more constructive example perhaps, a focus on the code object not the whole Internet allowed the connecting of objects within the Apache server (object). This software object can be seen, and used as a model, as a reconfiguration of objects whereby new possibilities for server-client relations were released. The hackers who brought objects together as they created the (open source) code for the Apache server were working with and through objects in the creation of a new object. This Is object oriented programming as object oriented politics. A more recent example would be the Open Source Distributed Micro-blogging service thimbl.net, an attempt to connect software and social objects in a new configuration. The whole ecosystem of APIs is an equally interesting object-oriented space. Perhaps even sites and services that allow non coders to connect and mashup data and data objects whether that is Yahoo Pipes or WordPress, or even in it’s own way Facebook itself are object spaces.
An object-oriented approach allows one to see all the objects in play at the same scale in the computational/governmental topology. Here the photos I upload, the protocols that encode them, the data trails I leave, the proprietary iPad I create them on as well as the algorithms that position them and me – the whole governmental mix, are objects connecting within objects. The aim is not to trace relations external to those objects but connections within them. To move from understanding objects in terms of their relations is not to deny connections. Rather it is to place those connections – those governmental tunnels through the rags ‘n refuse – front and centre, because they are issues of objects not issues of plasma or potential. Object-oriented approaches to the governmental mesh of the hoard allow us to deal with the unhuman objects of media and to address the connections that are made and can be made.
To return finally to the Facebook and Google hoard-archives and the unhuman patrolmen who burrow through our rags ‘n refuse, generating governmental positions as they go, an object-oriented approach to the Exploit offers new hope. Remaining true to a focus on objects and a flat ontology, rejecting relations as necessary to objects, it becomes possible to see how the data objects we willingly or unwillingly assign to Web 2.0 hoards are connected within those archives with others within governmental objects – the search-record object, the surveillance-object, the friend-object. These can be the target of Exploit. These are what can be reconfigured or realigned through new connections developed by new algorithms or software objects. The hoards may not be ours, the patrolmen burrowing through them may not be us, but that doesn’t mean we can’t find new ways through the rubbish.
The read-along with pictures version is available here (1.8MB pdf) and if you’re really keen you can see the video below… the advantage being you can see and hear my fellow panelists including the fine Mr Jackson.