In 1979, the French philosopher Jean-Francois Lyotard submitted to the Quebec government a document that would later be published as The Postmodern Condition: A Report on Knowledge. Lyotard’s report closed with a call for the opening of public access to the databases and knowledge banks on which future decision making would depend. If knowing things was increasingly the function of the research and development wings of giant corporations such as IBM, Lyotard’s concern was that capitalist techno-science had essentially only one way of knowing what it knew, which was to subject knowledge to the test of performativity. For the corporate knower, knowledge was valid to the extent that it made the system of capitalist accumulation function more efficiently. To liberate knowledge from this single goal or telos, the process of formulating knowledge would have to be taken up by multiple actors pursuing different, perhaps even incompatible, ends: opened out to a dissensus without a single overarching regulative norm.

Decades later, Tim Jordan’s Information Politics makes a similar call for open access to the data held by giant corporations, distinguishing between a privatised conception of data as exploitable ‘personal information’ linked to individual identity, and an alternative model of ‘simultaneous complete use’ in which rules mediating between contending claims to ownership need no longer apply. Under the latter model, the question would not be whether I or Google was the rightful owner of the profile Google held of my browsing activity, but how that information, contextualised by all of the other data accumulated by Google, might be made available for multiple purposes. How have the stakes of this political demand changed in the decades since Lyotard’s report?

The (possibly alarmist) claim recently surfaced on social media that it was only a matter of time before some enterprising hacker managed to connect the records held by porn sites of their users’ browsing histories to the individual identities of those users, creating considerable opportunities for individual blackmail or general mischief. My personal reaction to this scenario – oh god please no – was balanced by a tranquil sense that a great many people would be in the same boat, and that the likely social impact of mass disclosure was difficult to anticipate. It might be horrific and hilarious in about equal measure. However, sites such as Pornhub already occasionally release their own statistical analyses, showing which US states evince the greatest interest in teenagers, spanking, interracial couples and so on. Public access to their – suitably anonymised – access logs might yield much of sociological interest.

It’s surprising that Jordan doesn’t discuss anonymisation, as it offers a way of managing the tension between information’s inherent capacity for simultaneous complete use, and the desire of individual users of information systems for privacy, in particular freedom from government surveillance. Jordan wants to argue that the model of simultaneous complete use is liberatory, but ends up as a result having to construe our ‘fetishistic’ attachment to the information that tracks and profiles our online identities as something to be overcome, rather than as a vulnerability over which we might need to take great care. The chapter in Information Politics on ‘hactivism,’ while strong on the libertarian politics of early hacker culture, Wikileaks and the Arab Spring, has nothing to say on the subject of doxxing and other forms of highly personalised online harassment. The same exploits which enable the enterprising cracker to access and anonymously disseminate classified government information are used to raid the cloud backup accounts of female celebrities for nude photographs. That in itself does not discredit the use of such techniques for more ethical purposes, but it does suggest that a stronger basis for justification is needed than the rather abstract injunction that it’s good to share. Among hackers themselves, the ‘white hat’ / ‘black hat’ distinction is both significant and highly contested – this, arguably, is where the real politics of the situation resides.

In order to build out a framework for thinking about information politics, Jordan starts by sketching a kind of metaphysics of information, a toolbox of concepts which are then applied to a range of case studies. Information itself is characterised as ‘difference that moves,’ an interestingly open-ended formulation reminiscent of Derrida’s isolation of the ‘grapheme’ or ‘iterable mark’ as the operator of a kind of general textuality. I almost said ‘unit’ rather than ‘operator,’ but Derrida’s point about the grapheme is that it never comes to rest in a unitary identity: it is not just this mark on a page, but the repeatability of the gesture of marking what it marks. A similar problem arises in Jordan’s reference to the ‘bit’ as a unit of information. If information is ‘difference that moves,’ then its collection into repositories of data, buckets of bits, must be seen as a kind of arrest of movement, a reification. Yet it is data in this sense, information organised in a way that makes it repeatedly traversable by algorithms, that is held by databases; and the patterns of organisation that support what Jordan calls ‘platforms’ are what computer science knows as data structures. If information is difference that moves, then data is information caught in a kind of holding pattern, ‘given’ according to a particular way of arranging its ‘givenness.’

Jordan observes that software platforms convert events – a mouse-click, the submission of a form containing details that will establish a user’s profile within the system – into a kind of information specific to each platform. Who I am to (or ‘on’) Facebook is an aggregate of my interactions with Facebook’s platform. Without a theory of data, however, the nature of what has been captured and represented there remains difficult to describe with any specificity. Facebook ‘knows’ all kinds of things about me – but what is the ontology of its archive? What kinds of things is it able to infer on the basis of that knowledge, and how?

Information Politics attempts to deal with the ontology of data and operations on data through the concept of ‘recursion,’ which it treats as a general name for the re-processing of information to produce new information. When Amazon analyses my purchases, and those of other people who have purchased similar things to me, in order to make recommendations about other things I might like to buy, its computers are using a variety of machine-learning techniques to make inferences that serve the ultimate purpose of increasing Amazon’s share of the market – its performativity. If the recommendations engine succeeds in driving up sales, it receives still more information about purchases which can be fed back into the model. Amazon’s recommendations engine is in this sense a component in a cybernetic or reflexively self-regulating system. By describing the feedback loops that regulate such a system as ‘recursive,’ Jordan aims to draw attention to the possibility that the improvement in the overall efficiency of the system promised by such regulation might come at the cost of an ‘exponential’ increase in the quantity of information it must manage.

I think there is a confusion here, however, and it is exemplified by Jordan’s statement that Google performs ‘searches on searches’ in order to optimise the behaviour of its search engine. The expression ‘searches on searches’ suggests a potentially endless accumulation of information, since one could always have searches on searches on searches, and so on. But here we need to consider the types of data and the algorithms involved. The operation that finds search results in Google’s index based on user-supplied phrases or keywords is one type of algorithm; the operations that Google then applies to the history of users’ interactions with the search engine are of other kinds, with other goals. They are not ‘searches’ in the same sense that the original operation was a ‘search.’ Rather than a runaway process of generating information endlessly and circularly out of information, what we have is the extraction of one type of data from another through a process of automated inferential reasoning. There is no direct ‘recursion,’ because each stage accepts the outputs of a distinct prior stage as inputs, rather than feeding back into itself.

An efficient cybernetic system does not recursively produce unbounded amounts of information, but rather systematically reduces the amount of information it needs to consider through heuristics and pattern recognition. What Google and Amazon are always trying to get better at is throwing away information, distinguishing with the least effort and the greatest cogency between relevant and irrelevant data. The political question here is, what counts as ‘relevant’? What criteria are admissible (and which is it technologically feasible to apply)?

To Lyotard’s call for open access to the contents of corporate data banks, Jordan adds that the ‘recursions’ practiced by corporations on their data should also be opened to public scrutiny, re-use and redefinition. In this, I am certain that he is right. Machine-learning is typically directed towards some goal supplied from without: the teleology of its algorithms is not inherent in the algorithms themselves, but in their ‘training’ towards some specific end. But to intervene politically in this area, it is more necessary than ever to understand the ontology of data, its constraints and affordances. The information politics of the present and near-future is less about property rights than it is about the expanding powers, and constitutive limitations, of Bayesian inference.

Dominic Fox is a writer and programmer living and working in London. He is the author of Cold World and blogs at www.thelastinstance.com.

Tweet

Find a Local Independent Bookshop from which to Purchase this Book

Please enter a postcode or a city in the UK