Do algorithms think? Do buildings speculate? In his review of Luciana Parisi’s recent book, Contagious Architecture, Jeremy Lecomte considers her claim that parametric architecture is a mode of algorithmic computation that should be understood as speculative thought
Contagious Architecture digs deeply into the logic of computation. Dealing with digital, interactive and parametric architecture, it explores the way algorithms transform our digital environment. Almost ten years after Abstract Sex (2004), Luciana Parisi continues to seek the transformations prompted by the ingression of computational technologies into almost all aspects of contemporary culture. But by simultaneously broadening the empirical and theoretical scope of her analysis, she also gets closer to examining the nature of algorithmic logic.
While making decisive and original contributions to the understanding of computation, Parisi’s excellent book is, nevertheless, both dense and difficult. Ranging from computation and digital metaphysics to algorithmic objects, aesthetics and architecture, the book multiplies points of theoretical intervention. Parisi articulates highly abstract points and precise examples at an incredible speed. But despite her pedagogic efforts, it is not always clear whether she is iterating aspects and instances of the same argument, or exposing many different and overlapping ones.1 Still, it seems possible to argue that her book follows a central argument: the idea that computation should no longer only be read in terms of finite sets of instructions, but rather as constituting an entirely new mode of speculative thought.
This ambitious and radical idea is related to a diverse set of theoretical discussions. This might discourage readers who are unfamiliar with the many specialist areas her book covers. From information theory to parametric architecture, from Félix Guattari’s diagrammatics to the speculative and natural philosophy of Alfred North Whitehead, and from contemporary debates around Object Oriented Ontology to materialist debates about thought and computation, it sometimes feels like only those already versed in these subjects will be truly able to follow her and understand the importance of her contribution. Divided into three long chapters preceded by a relatively short but clear introduction, this book is a demanding read, but it develops many very interesting ideas.
While it would seem reductive to trim her book down to parametric architecture, it nevertheless seems the most tangible and prevalent subject matter. Recently elevated to the status of a new global style by Patrick Schumacher, parametric architecture generally indexes the fluid, smooth and organic forms that architects like Frank Gehry, Zaha Hadid, Kas Oosterhuis, Greg Lynn and Francois Roche (to which Parisi has devoted several sections of her book) have been conceiving since the late ’90s, with the help of computers. Within the field of architecture, parametricism can be said to define design practices using modelling software to generate forms from the differential encoding of different parameters – from cost estimation and energy usage, to circulation flows, noise, geological constraints and geographical conditions. Although imbued with many problematic aspects when seen in relation to the general political and economic environment in which it intervenes, parametric architecture presents important innovations in terms of form and conception. And it is mainly at this level that Parisi engages her discussion.
Although it deals quite extensively with parametric architecture, Parisi’s book does not look like any that would be more traditionally associated with the field of architecture. It sometimes seems that Parisi employs the term ‘architecture’ metaphorically, constantly moving from actual architectural objects and practices to the logic of computation and algorithms themselves. But the book as a whole engages this problem head on, suggesting that it is in fact precisely the dissolution of that distinction that needs to be addressed.2 Arguing that parametric architecture produces new types of objects and new spatial conditions (morphogenetic forms that emerge from locations that change over time), she does not only explore new architectural forms, but examines their digital nature. If interactive and digital architectures shift from former Euclidean conceptions of space and time to the production of new topological geometries, Parisi argues that we also need to develop a new understanding of the digital itself. Largely drawing on the more recent debates in information theory, Parisi argues that contemporary means of computation enter new speculative territories.
What Is Computation?
Defining her move from parametric architecture to the logic of computation, the first chapter argues that
a consideration of algorithmic architecture can assist us in understanding algorithms as actual objects: as spatiotemporal data structures that are internally conditioned by infinities as incomputable entities.3
Contrary to the understanding of computation in terms of pure instrumental rationalism, Parisi develops the idea that it needs to be conceived in terms of speculative reason. She starts by recalling and opposing the two main ways in which algorithms have been conceived. On the one hand that they correspond to a set of finite instructions (first-order cybernetics); on the other that they are able to adapt and vary according to external stimulations and changing conditions (second-order cybernetics). According to her, both conceptions remain unsatisfactory because, she argues, this alternative is unable to solve ‘the problem of calculating infinite series of probabilities in a manner that also includes the probability of incomputability’.4 Criticising axiomatic absolutisation, Parisi argues that what computation does can only be understood by directly engaging with the problem of the incomputable not in opposition to but within computation. According to her, the ubiquity of computation in the organisation of our contemporary environment should not be understood in all encompassing terms, but rather in relation to its incompleteness:
Against the metacomputational view of a universe contained in simpler axioms, I will argue that incomputable limits are truly intrinsic to computation.
According to Parisi, one must acknowledge the fact that
ontological complexity or chaotic incompleteness does not emerge from order, but is rather the unconditional condition […] of procedural calculations.5
Parisi addresses contingency within rather than outside algorithmic processing. And this is why questions of change, probabilities and potentialities play such an important role in her book. Largely drawing on Gregory Chaitin’s algorithmic theory of information, she argues that change and potential are not exterior to a computational logic that would not be affected by incompressible and uncomputable data, but are rather inherent to the necessary incomplete nature of computation. Far from considering that parametric architecture is the incarnation of contemporary procedures of control and capitalist driven modes of evaluation and valorisation, Parisi argues that it can be taken as an interesting example of ‘the actual existence of spatiotemporal data structures infected with incomputable quantities.’ According to her it does not only model existing structures, but rather becomes ‘a metamodel of algorithmic objects, which are conceived here as discrete entities imbued with infinity.’6
Image: Greg Lynn, Embryological House, 1997-2001
From this standpoint, it seems quite difficult to understand why Parisi needs to introduce a categorisation of digital and computational processes in terms of algorithmic objects. Why objects, one might retort? And just as in the case of Graham Harman’s work, the answer is not obvious.
Her discussion of Harman’s philosophy makes it clear: Parisi does not share his interest for substantial unities. While Harman argues against process oriented philosophies (such as Whitehead’s) and other fragmentary conceptions that, according to him, are unable to account for the individual essence of an object, Parisi engages more closely with Whitehead to argue that the formalism and geometry produced by algorithmic architecture are actually evolving structures of relations. Yet, neither does Parisi want to see everything vanishing into open-ended series of processes. Although her book is all about change and potentialities, she also wants to stress the fact that, in her view, algorithmic entities are both abstract and actual. They are not virtual, but full actualities that do not only relate to other actual entities but process them in various different ways.
Following this, Parisi distinguishes her views from both metacomputational approaches (which, according to her, consider that ‘finite sets of algorithms, or mathematical axioms, generate infinitely complex structures’) and interactive conceptions that replace these finite sets with ‘processes relying on external inputs and temporal variations’. According to her, both overlook the fact that ‘algorithms are actual entities imbued with infinity.’7 Her conception of algorithmic objects therefore seeks to resist reducing algorithms to finite quantities and sets of instructions and, at the same time, to evolutionary interactive models based on sensors, processors and feedback mechanisms. But it is not always easy to see whether these discrete objects she talks about characterise the architectural objects produced by digital and parametric models or the units (notably cellular automata) allowing those models to function and be designed.
Grounded in Whitehead’s philosophy, her conception appears in fact to be very close to that which Greg Lynn develops in terms of ‘objectiles’ and ‘blobs’. Employing those terms, Lynn argues that objects are no longer closed entities designed according to exact coordinates in space. But rather curved surfaces that can only be defined in terms of vectors and transformations. According to Parisi these objects are spatiotemporal events. When she talks about space events, it would seem that she plays Whitehead against Harman in order to understand actual architectural objects and the digital objects used in object-oriented programming under the same occasionalist conception. But she also frequently refers back to the notion of object, arguing that they are ‘always more than one and less than many’; always more than one closed unity and less than an indefinite series of open processes. It is thus difficult to understand why she keeps emphasising objects, instead of dropping this notion altogether and talking about occasions and events, as is the case in most of Whitehead’s philosophy.8
This question appears even more complicated to solve when considering what seems, in fact, to be the central argument of the book: the contention that far from being mere instrumental axioms and sets of instructions (be they bound to subtle interactive mechanisms), algorithmic objects deploy an original mode of thought.
This argument presents significant contributions to the development of a non-anthropocentric definition of thought. But while her definition of algorithmic processes beyond finite axiomatics and interactive forms seems decisive, her contention that they would display modes of rationality that can be characterised in terms of speculative thought seems far more problematic.
Parisi discusses this in relation to Whitehead’s notion of ‘prehension’. For Whitehead this notion describes how actual entities may relate to other entities by means of selection, inclusion, exclusion, and transformation.9 Parisi extends this to a
process by which an actual entity confronts infinite data through the physical and/or conceptual selection, evaluation, inclusion, exclusion, and transformation of data, and by which it thereby invests and reprograms the actual field of potentiality.10
The problem is that her deep and thorough engagement with Whitehead’s philosophy and contemporary debates around neuroarchitecture do not really then explain the jump she makes from the definition of algorithmic processes (in terms of prehension), to what she first calls speculative computation and then speculative thought.11 Although the whole book may be well summarised by this relation according to which Parisi sees algorithmic architecture as ‘one example of the way in which computation builds the present through the prehension of infinite data’, we are left with the problem of knowing why and how algorithmic architecture would thus not only be a case of speculative computation but further, a case of speculative thought ‘exposing reason, logic, and calculation to the power of the incomputable’ remains totally open.12
The first chapter concludes on a beautiful and really interesting description of Whitehead’s understanding of speculative thought. This, it appears, has nothing to do with Hegel’s idea of speculation. Insisting on both novelty and theoretical orientation, Whitehead’s conception moves away from both practical and formal ones, and directs reason to the question of final causation. His problem is not critical reason, but rather instrumental reason. To this he opposes those abstract processes that select and evaluate present and past conditions according not only to future possibilities, but also to final causes. This conception contends that ‘reason provides the judgement by which novelty passes into realisation, into fact.’13 And while Whitehead argues that the function of reason hence precedes direct observation and any synthesis of empirical data, and that it is in fact grounded in abstract schemes according to which the significance of novel facts and observations can be attested and evaluated, it is still unclear how this leads Parisi to suggest that ‘abstract schemes that involve the automatic selection, inclusion and exclusion of infinite amounts of data’ can be defined in terms of speculative thought.
Control And Change
Due to its ongoing discussion of computational randomness, incompressible data, uncomputable quantities and digital contingencies, Parisi’s book appears to be all about potential change and unpredictable novelties. This problem runs through all examples she deals with, and largely determines the way she engages with the authors and theoretical arguments that she mobilises.
Her argument that algorithmic architecture may display one example of the way computation cannot be seen to be substituting pure instrumental calculation for the more complex mechanisms of rationality but should rather be seen to develop new modes of speculative thought, is bound to the distinction she makes between topology and mereotopology.14 Mereotopology, she argues, can be seen to annex another approach to extension. One that would lie in between what Gilles Deleuze and Félix Guattari used to understand in terms of striated space on the one hand (the space of social stratifications, hierarchies and other metric co-ordinates) and smooth space (the more fluid and continuous space of topological deformations) on the other. One that would account for the many processes that organise their reciprocal relation: relations between parts and wholes, between discrete parts and their boundaries. She argues that,
While the gridlike architecture of striated space places discrete unities at the centre of a design made of points connected by lines, the topological curves of smooth space (produced by algorithmic or parametric architecture) start from the generative power of a point, the meshing and folding of which becomes the condition for the emergence of a new form.15
Although she explicitly states that she is not concerned with criticising parametricism, neither for its formal hypostasis nor for its relation to market driven neoliberal conceptions of space and social relations, she contends that mereotopology potentially escapes the logic of topological control that parametric architecture shares with neoliberal politics.
Doing so, she argues that ‘parametricism is not abstract enough to meet the possibilities offered by a radical formalism.’16 Instead of criticising parametric architecture from the outside, from the standpoint of its collusion with principles, strategies and functions that are all bound to the support of capitalism’s development in times of crisis, Parisi argues that it is on the level of its formalism (the level of algorithmic abstraction) that parametricism should be challenged. Contrary to the topological architecture of smooth (but preemptive) control displayed by parametric architecture, Parisi contends that the mereotopological approach developed by Whitehead between 1916 and 1929
contributes to describe the existence of an asymmetry between topological formalism and what [she refers to] here as space events. These space events are actual architectures of relation that define what cannot be reduced to topological control: the probability of chaos, the unilateral indetermination of data within computation and at another level the unleashing of unlived reality into urban design.17
It is interesting to see that instead of speaking about principles, Parisi moves to the modes of organisation and the technologies of conception themselves. Computation cannot hence be reduced to the neoliberal politics that infused parametric architecture. Again, far from addressing computation from the outside, it is on the level of computation and within its logic that Parisi seeks to distinguish between preemptive technologies of control, and an aesthetics that would be more open to contingencies, potentialities and unpredictable change. Doing so, however, Parisi seems to also replay the rather traditional divide between what is reducible to control and what is not. By concentrating on this unpredictability, on randomness and incomputable quantities within computation, she also reduces topology to the way parametric architecture understands it. It seems that, while it can be argued that the problem with parametric architecture is that it is not yet abstract enough, it should also be argued that the emphasis it puts on form and smooth continuities (not to speak about its obsession with biological forms) tends to overlook other questions – ones that, contrary to what Parisi argues in her book, cannot be separated from the analysis of the more general political and economic ecology to which it belongs. By addressing the computational logic at stake in parametric architecture, Parisi lays bare the organisational logics that shape our environment. But by not addressing the more general ecology of relations (notably between architecture and the city) to which this organisational logics are bound, she also risks reducing the question of politics and control to a formal and logical problem.
Thanks to Inigo Wilkins for his reading of an earlier version of this text.
Jeremy Lecomte is currently a PhD candidate at the Manchester Architecture Research Centre. He is also a founding member and co-director of the international research laboratory Glass Bead and of the associated journal, released in 2014
Luciana Parisi, Contagious Architecture: Computation, Aesthetics and Space, Cambridge: MIT Press, 2013.
1 This is exemplified by the glossary, situated at the end of the book, that proposes very interesting and compact definitions of the main notions she uses. But it is also a rather long list, presenting concepts that sometimes overlap in ways that become even more difficult to follow in the book.
2 This appears clearly in the general definition Parisi gives of architecture in her glossary, where she argues that ‘In a general sense, this term (architecture) refers to the matrix or structure whose general arrangement of relations produces space and time.’ Parisi, p.258.
3 Parisi, ibid., p.83
4 Ibid., p.13.
5 Ibid., p.21.
6 Ibid., p.43.
7 Ibid., p.3.
8 Ibid., pp.47,60,62.
9 See A. N. Whitehead, The Concept of Nature, Amherst, NY: Prometheus Books, 2004; Process and Reality: An Essay in Cosmology, New York: Free Press, 1978; and I. Stengers, Penser avec Whitehead: Une libre et sauvage creation de concepts, Paris: Seuil, 2002.
10 Ibid., pp.265-266.
11 Ibid., p.9.
12 Ibid., p.71.
13 Ibid., p.73.
14 Ibid., p.264.
15 Ibid., p.xi.
16 Ibid., p.86.
17 Ibid., p.88.