In this short post I want to think about the limits of computation, not the limits theoretically of the application or theorisation of computation itself, but actually the limits to which computation within a particular context should be contained. This is necessarily a normative position, but what I am trying to explore is the limit at which computation, which can have great advantages to a process, institution or organisation, starts to undermine or corrode the way in which a group, institution or organisation is understood, functions or how it creates a shared set of meanings. Here though, I will limit myself to thinking about the theorisation of this, rather than its methodological implications, and how we might begin to develop a politics of computation that is able to test and articulate these limits and understand the development of a set of critical approaches which are also a politicisation of algorithms and of data.
By computational creep I am interested in the development of computation as a process rather than an outcome or thing (Ross 2017: 14). This notion of “creep” has been usefully identified by Ross in relation to extreme political movements that take place by what he calls “positive intermingling”. I think that this is a useful way to think of the way in which computationalism, and here I do not merely mean the idea that consciousness in modelled on computation (e.g. see Golumbia 2009), but more broadly as a set of ideas and style of thought which argues that computational approaches are by their very nature superior to other ways of thinking and doing (Berry 2011, 2014). This is also related to the notion that anything that has not been “disrupted” by computation is, by definition, inferior in some sense, or is latent material awaiting its eventual disruption or reinvention through the application of computation. I would like to argue that this process of computational creep takes six stages:
- Visionary-computational: Computation suggested as a solution to existing system or informal process. These discourses are articulated with very little critical attention to the detail of making computational systems or the problems they create. Usually, as Golumbia (2017) explains, these draw on metaphysics of information and computation that bear little relation to material reality of the eventual or existing computational systems. It is here, in particular, that the taken-for-grantness of the improvements of computation are uncritically deployed, usually with little resistance.
- Proto-computational: One-off prototypes developed to create notional efficiencies, manage processes, or to ease reporting and aggregation of data. Often there is a discourse associated with the idea that this creates “new ways of seeing” that enable patterns to be identified which were previously missed. These systems often do not meet the required needs but these early failures, rather than taken as questioning the computational, serve to justify more computation, often more radically implemented with greater change being called for in relation to making the computational work.
- Micro-computational: A wider justification for small scale projects to implement computational microsystems. These often are complemented by the discursive rationalisation of informal processes or the justification of these systems due to the greater insight they produce. This is where a decision has been taken to begin computational development, sometimes at a lightweight scale, but nonetheless, the language of computation both technically and as metaphor starts to be deployed more earnestly as justification.
- Meso-computational: Medium-scale systems created which draw from or supplement the existing minimal computation already in process. This discourse is often manifest in multiple, sometime co-exisiting and incompatible computations, differing ways of thinking about algorithms as a solution to problems, and multiple and competing data acquisition and storage practices. At this stage the computational is beyond question, it is taken as a priori that a computational system is required, and where there are failures, more computation and more social change to facilitate it are demanded.
- Macro-computational: Large-scale investment to manage what has become a complex informational and computational ecology. This discourse is often associated with attempts to create interoperability through mediating systems or provision for new interfaces for legacy computational systems. At this stage, computation is now seen as as source of innovation and disruption that rationalises the social processes and helps manage and control individuals. These are taken to be a good in and of themselves to avoid mistakes, bad behaviour, poor social outcomes or suchlike. The computational is now essentially metaphysical in its justificatory deployment and the suggestion that computation might be making things worse is usually met with derision.
- Infra-computational: Calls for overhaul of and/or replacement of major components of the systems, perhaps with a platform, and the rationalisation of social practices through user interface design, hierarchical group controls over data, and centralised data stores. This discourse is often accompanied by large scale data tracking, monitoring and control over individual work and practices. This is where the the notion of top-view, that is the idea of management information systems (MIS), data analytics, large-scale Big Data pattern-matching and control through algorithmic intervention are often reinforced. In this phase a system of data requires free movement of the data through a system through an open definition (e.g. open data, open access, open knowledge), which allows standardisation and sharability of data entities, and therefore of further processing and softwarization. This phase often serves as an imaginary and is therefore not necessarily ever completed, its failures serving as further justification for new infrastructures and new systems to replace earlier failed versions.
This line of thinking draws on the work of David Golumbia, particularly the notion of Matryoshka Dolls that he takes from the work of Phillip Mirowski. This is in relation to the notion of multiple levels or shells of ideas, that form a system of thinking, but which is itself not necessarily coherent as such, nor lacking in contradiction, particularly at different layers of the shells. This “Mirowski calls the ‘’Russian doll’ approach to the integration of research and praxis in the modern world'” (Golumbia 2017: 5). Golumbia makes links between this way of thinking about neoliberalism as a style of thinking that utilises this multi-layered aspect and technolibertarianism, but here I want to think about computational approaches more broadly, that is as instrumental rational techniques of organisation. In other words, I want to point to the way in which computation is implemented, usually in a small scale way, within an institutional context, and which acts as an entry-point for further rationalisation and computation. This early opening creates the opportunity for more intensive computation which is implicated in a bricolage fashion, that is that, at least initially, there is not a systematic attempt to replace an existing system, but over time, and with the addition to and accretion of computational partialities, calls become greater for the overhaul of what is now a tangled and somewhat contradictory series of micro-computationalisms, into a more broad computational system or platform. Eventually this leads to a macro- or infra-computational environment which can be described as functioning as algorithmic governmentality, but which remains ever unfinished with inconsistencies, bugs and irrationalities throughout the system (see Berns and Rouvroy 2013). The key point is that in all stages of computationally adapting an existing process, there are multiple overlapping and sometimes contradictory processes in operation, even in large-scale computation.
Here I think that Golumbia’s discussion of the “sacred myths among the digerati” is very important here, as it is this set of myths that are unquestioned especially early on in the development of a computational project. Especially at what I am calling the visionary-computational and proto-computational phases, but equally throughout the growth in computational penetration. Some of these myths include: claims of efficiency, the notion of cost savings, the idea of communications improvement, and the safeguarding corporate or group memory. In other words, before a computerisation project is started, these justifications are already being mobilised in order to justify it, without any critical attention to where these a priori claims originate and their likely truth content.
This use of computation is not just limited to standardised systems, of course, and by which I mean instrumental-rational systems that are converted from a paper-based process into a software-based process. Indeed, computation is increasingly being deployed in a cultural and sociological capacity, so for example to manage individuals and their psychological and physical well-being, to manage or shape culture through interventions and monitoring, and the capacity to work together, as teams and groups, and hence to shape particular kinds of subjectivity. Here there are questions more generally for automation and the creation of what we might call human-free technical systems, but also more generally for the conditions of possibility for what Bernard Stiegler calls the Automatic Society (Stiegler 2015). It is also related to the notion of digital and computational systems in areas not previously thought of as amenable to computation, for example in the humanities, as is represented by the growth of digital humanities (Berry 2012, Berry and Fagerjord 2017).
That is to say, that “the world of the digital is everywhere structured by these fictionalist equivocations over the meanings of central terms, equivocations that derive an enormous part of their power from the appearance that they refer to technological and so material and so metaphysical reality” (Golumbia 2017: 34). Of course, the reality is that these claims are often unexamined and uncritically accepted, even when they are corrosive in their implementations. Where these computationalisms are disseminated and their creep goes beyond social and cultural norms, it is right that we ask: how much computation can a particular social group or institution stand, and what should be the response to it? (See Berry 2014: 193 for a discussion in relation to democracy). It should certainly be the case that we must move beyond accepting a partial success of computation to imply that more computation is by necessity better. So by critiquing computational creep, through the notion of the structure of the Russian doll in relation to computational processes of justification and implementation, together with the metaphysical a priori claims for the superiority of computational systems, we are better able to develop a means of containment or algorithmic criticism. Thus through a critical theory that provides a ground for normative responses to the unchecked growth of computations across multiple aspects of our lives and society we can look to the possibilities of computation without seeing it as necessarily inevitable or deterministic of our social life (see Berry 2014).
 The title “Against the Computational Creep” is reference to the very compelling book Against the Fascist Creep by Alexander Reid Ross. The intention is not to make an equivalence between fascism and computation, rather I am more interested in the concept of the “creep” which Ross explains involves small scale, gradual use of particular techniques, the importation of ways of thinking or the use of a form of entryism. In this article, of course, the notion of the computational creep is therefore referring to the piecemeal use of computation, or the importation of computational practices and metaphors into a previously non-computational arena or sphere, and the resultant change in the ways of doing, ways of seeing and ways of being that this computational softwarization tends to produce.
Berns, T. and Rouvroy, A. (2013) Gouvernementalité algorithmique et perspectives d’émancipation : le disparate comme condition d’individuation par la relation?, accessed 14/12/2016, https://works.bepress.com/antoinette_rouvroy/47/download/
Berry, D. M. (2011) The Philosophy of Software: Code and Mediation in the Digital Age, London: Palgrave Macmillan.
Berry, D. M. (2012) Understanding Digital Humanities, Basingstoke: Palgrave.
Berry, D. M. (2014) Critical Theory and the Digital, New York: Bloomsbury
Berry, D. M. and Fagerjord, A. (2017) Digital Humanities: Knowledge and Critique in a Digital Age, Cambridge: Polity.
Golumbia, D. (2009) The Cultural Logic of Computation, Harvard University Press.
Golumbia, D. (2017) Mirowski as Critic of the Digital, boundary 2 symposium, “Neoliberalism, Its Ontology and Genealogy: The Work and Context of Philip Mirowski”, University of Pittsburgh, March 16-17, 2017
Stiegler, B. (2016) The Automatic Society, Cambridge: Polity.