Summary of Part 2, Chapter 2: The Regulative Function of Culture in the Relation between Man and the World of Technical Objects. Current Problems.
In this chapter Simondon makes his case for mechanology as a corrective to the historical and cultural failure to properly make sense of the relationship between humans and machines. He starts back again with the Encyclopedists: they did not understand machines as fully automata; instead, they still thought of them as assemblages of devices, in other words, at the level of the element. This led to a false sense of, and belief in, continuous progress (because it is possible to see such continuous progress in the saturation/concretization of elements; if they had been thinking at the level of individuals or ensembles (I think he would say), progress would not only appear “serrated” in S’s terminology, it would also not be misunderstood as some mystical process that happens by itself).
In any event, it is the way change happens at the level of the element in the 18th century, which characterizes that period’s mix of euphoria and anxiety regarding technology. Euphoria arises from the experience of continuous improvement ongoing during that time; anxiety in this period regards “those transformations that provoke a break within the rhythms of everyday life, making the old habitual gestures useless” (130). S delineates an interesting distinction between tools and instruments, “if by tool one understands the technical object enabling one to prolong and arm the body in order to accomplish a gesture, and by instrument the technical object that enables one to prolong and adapt the body in order to achieve better perception.” He gives an interesting discussion on how many tools will be both tool and instrument, for instance a hammer also gives feedback to the user on the resistance and movement of the nail being driven; his point is though that in this case the hammer is still primarily a tool, since its quality as an instrument is subordinate to its use as a tool; he holds that this is still the case even when a mason uses a hammer to tap a wall to get a sense of its composition. In contrast, telescopes, microscopes, etc. are instruments, pure and simple.
With the advent of “complete technical individuals” in the 19th century, the previous anxiety with technological change becomes much more acute, as there are now machines which replace humans.
It is not necessarily through its size that the factory distinguishes itself from the craftsman’s workshop, but through the change in relation between the technical object and the human being: the factory is a technical ensemble that is comprised of automatic machines, whose activity is parallel to that of human activity; the factory uses true technical individuals, whereas, in the workshop, it is man who lends his individuality to the accomplishment of technical actions. (131)
The progress of the nineteenth century can no longer be experienced by the individual, because it is no longer centralized with the individual as the center of command and perception in the adapted action. (132)
The notion of progress thus “splits in two,” as humans lose their earlier “kinesthetic” contact with technology, and alongside the sense of progress exists a growing anxiety due to the disconnection with technology and its growing incomprehensibility of scale.
Progress is henceforth thought of as cosmic, at the level of its overall results. It is thought abstractly, intellectually, in a doctrinal manner. Progress is no longer thought by craftsmen, but by mathematicians, who conceive of progress as man taking possession of nature.
“The individual who thinks progress is not the same individual as the one who works,” S argues – note that, in contrast to, e.g., Braverman, or Bookchin, who made the same historical observation, S attributes this differentiation between the thinker and the worker to the effects of the societal experience of this stage of technological development (viz., of the technical individual), rather than to the social or economic order per se. For all S’s disavowal of having any dialectic going on in his account of history, his model does feel like it has the somewhat dissociated clockwork effect of an idealist dialectic, in which stages just somehow follow each other (his invocation of context, experience, etc. being too a priori to be properly termed materialist, imho).
S in fact goes on to argue that his account provides a deeper understanding of alienation than that of the Marxist concept, which is, in S’s view, superficial, merely “juridical and economic”:
Beneath this juridical and economic relation exists an even more profound relation, that of the continuity between the human individual and the technical individual, or of the discontinuity between these two beings. … The alienation of man in relation to the machine does not only have a socio-economic sense; it also has a physio-psychological sense; the machine no longer prolongs the corporeal schema, neither for workers, nor for those who possess the machines. (133)
He goes on to state that bankers, etc., are just as alienated as anyone else, despite not being exploited for their labor – and quickly dismisses Hegel’s master-slave dialectic as an explanation for this – his ultimate point being, basically, that everyone is alienated and has only a partial understanding of contemporary technology and its relation to the human. It could be quite easily demonstrated that Simondon is not accurately representing or engaging with the full elaboration of the process of alienation in the Economic and Philosophic Manuscripts (and the first kind, alienation from nature, would pretty much cover the “machinic alienation” or whatever you might call it, that he is trying to outline; he could also be said to be describing what Marx would see as the role of automation as an aspect of real subsumption). But his agenda is actually, once again, to explain away any given subject-position within society as partial and alienated, thus showing the need for a new mechanological perspective.
The perspectives of both labor and capital are “late” (presumably meaning “outdated”) with regard to the modern technical individual; and the “dialogue” or struggle between the two is “false because it is of the past” (134). [To the extent that there is a validity to the stages of technical development and awareness that S elaborates, his error is the presentist one, in supposing that each stage completely displaces or supplants the previous stages, instead of layering over and interacting with the previous stages complexly]. [And this is an aspect of how his account reads like an old-fashioned, simplistic idealist dialectic (like, say, Stirner’s (sorry, Max!)) instead of, say, more nuanced Bakhtinian dialogism).]
S concludes this discussion with a much clearer exposition of his concept of finality (which I confess I was a bit confused by in the previous chapter). It is, basically, instrumentalism, and it shapes and limits the perspectives of both the worker and the capitalist. They understand machines, thus, in terms of the purpose for which they are put to work; this [external condition?] prevents them from understanding the “internal coherence” of the machine, and thus its true nature. The development of which understanding will, of course, be the goal of the mechanologist.
He goes on to discuss the ideal form of coupling between human and machine as equals, in other words with the human “not merely as a being who directs or utilizes it through the incorporation of ensembles, or as a being who serves it by supplying matter and elements. … There is an inter-individual coupling between man and machine when the same self-regulating functions are better and more subtly accomplished by the man-machine couple than by man or machine alone” (135).
To illustrate this, he posits a difference between how the memories of machines and humans work, and how they can work together combining their distinct strengths. Machines can only record; their memory does not even strictly speaking contain forms (because this would require an awareness of these forms), “but merely a translation of forms, by means of an encoding in a spatial or temporal distribution” (136). Humans are required to perceive the forms recorded in machine memory. This indifference to form is a strength of machine memory, in that it allows it to record “elements without order;” human memory, in contrast, requires a sense of order in order to remember. Also, though machine memory has a certain plasticity, this is the plasticity of being able to be written and erased. Human memory, less reliably “monomorphic” and reliable than machine memory, nevertheless also has the ability to infer and interpret, aka the “plasticity of integration” (137), and thus is able to draw on experience and memory to make predictions and fill in gaps in its knowledge. Thus, the proper context for the “coupling” of human and machine memory is those complex procedures in which both are needed.
This leads to a reiteration of S’s insistence that “Despite appearances, it is, on the contrary, the truly automatic machine that least replaces man” (139); this is still true a priori, because (in previous chapters) he has defined automata as lacking any “margin of indeterminacy” or openness (and machines which do have such a margin of indeterminacy are not “automata”); human interlocutors are thus necessary in any operation more complex than pushing a button to start and end an automatic process.
It becomes a reasonable question as to whether the development of machine learning has led this aspect of Simondon’s thought to become outdated, with potential consequences for his entire model of ideal human-machine interaction. Surely, Simondon is aware that automata can involve sensors and actuators, and thus be open to outside information; it is just that they are limited in their ability to respond. Thus, an air-conditioner can turn on and off in response to ambient temperatures, but it cannot turn itself off because the water-drain line is backing up – unless such a capacity has been built into the machine. A human observer, in contrast, needs no previous specific programming to go “oh shit, the water is backing up” and take some action in response. [Though it still seems to me, that the difference between closed/automatic system and open/ad hoc system should be a continuum, not a binary as S treats it.]
Briefly reviewing a few recent articles which discuss machine learning in a theoretical context informed by Simondon, we can see Rantala and Muilu (2023: 8) asserting that machine learning does have a “margin of indeterminacy” but that learning machines are still limited by their programming in terms of their ability to respond. Haworth (2020), discussing the “possibility of independently creative machines,” argues that the very idea of these machines as “independently creative” is based on the “fantasy of absolute autonomy” whereby we imagine ourselves as sovereign subjects instead of as parts of complex human-machine ensembles (and then, in a nightmarish vision, transfer this autonomy to the uncanny action of machines, instead of recognizing that they, as well, are more accurately understood as also embedded in such ensembles). Haworth thus seems to follow the Simondonian line of dismissing “the Robot” as a nonsensical figment, instead of addressing the question directly as to whether machine learning can or could render learning machines independent of any need for human interaction.
In any case, machines need humans as servants, technicians, or organizers; the self-regulation of automata is not enough for the machine to comprehend “the whole of the milieu,” for which both human and machine are required (139-40).
S criticizes the “autocratic philosophy” of the technocrats, who seek to use machines as slaves; the human should be at the same level as the machine, not an inferior, nor a superior. He embarks on a discussion of the limitations of 19th century understanding of machines; “The nineteenth century could produce only a technological technocratic philosophy because it discovered engines and not regulations” (i.e., feedback/information theory). He discusses examples of 19th century technology in which there is no distinction between the energy channel, and the information channel; understanding the different needs of these, and developing distinct channels, is the key aspect of progress in 20th century technics; this changes even the concept of efficiency, which is different for the flow of information than it is for energy used in production, motive force, etc.
This leads him into an interesting discussion of information, which requires that its channel of transmisson be capable of variability – consistent order, always the same, cannot transmit new information. Thus, information bears a resemblance to chance, yet it must, ultimately, be distinguishable from both order/form and pure chance, as a sort of intermediate entity (150). The distinction between form and information is linked to that between the machine and the human:
There is, in effect, an important gap between the living thing and the machine, and consequently between man and machine, which comes from the fact that the living thing needs information, while the machine essentially uses forms, and is so to speak constituted with forms. …
The human individual thus appears as having to convert the forms deposited into machines into information; the operating of machines does not give rise to information, but is simply an assemblage and a modification of forms; the functioning of a machine has no sense, and cannot give rise to true information signals for another machine; a living being is required as mediator in order to interpret a given functioning in terms of information, and in order to convert it into the forms for another machine.
[Returning us again to the question as to whether this stark opposition is still valid, and/or whether this is a useful way to define “automata.”]
He goes now into reiterating the difference between his view and that of the cyberneticians, based on the progress of his discussion to this point. “The machine is a deposited fixed human gesture that has become a stereotypy and the power to restart” (151). The cyberneticists overemphasize the analogy between machines and living organisms, but the truth is that the former “neither nourishes itself , nor perceives, nor rests,” like an actual living organism. He continues for several pages with a discussion of the distinction (from Bergson with some amendments) between open and closed machines, the former allowing for some margin of indeterminacy, and the latter are true automata, per his definition.
He returns to the key concept of transduction, with the example of a continuous relay that converts (transduces) potential into actual energy; information is also linked to this moment of transduction: “It is during the course of this passage from potential to actual that information comes into play; information is the condition of actualization” (155). The concept of transducer is expanded to “a regulative function in all machines having a certain margin of localized indeterminacy in their functioning;” in turn, humans, and all living creatures, are also transducers, as convertors/modulators of potential into actual energy. This capacity as transducer is part of what ensures a particular role for humans in the human/machine assemblage:
It is in fact very easy to construct machines that ensure a much greater accumulation of energy compared to that which man can accumulate in his body; it is equally possible to use artificial systems that constitute effectors that are superior to those of the human body. But it is very difficult to construct transducers comparable to the living thing. (156)
In fact the “transducers” found in machines are not actually fully “transducers,” according to Simondon’s definition [arguably, this is a result of his practice of constructing definitions from presumed essences; cf. my earlier criticisms of his definition of “automata”], because of the role of information; machines must be given information, while living things can give themselves information [this seems to be relevant to the example I gave above with the air conditioner]. Machines can only approach problems according to the way they have been programmed; they cannot “solve” problems because this involves that extra, human, step of inference/transduction:
To solve a problem is to be able to step over it, to be capable of recasting the forms that are given within the problem and in which it consists. The solution of real problems is a vital function presupposing a recurrent mode of action that cannot exist in the machine: the recurrence of the future with respect to the present, of the virtual with respect to the actual. There is no true virtuality in a machine; the machine cannot reform its forms in order to solve a problem.
Machines can generate information but they cannot understand it unless it is presented or “given” to them, and this requires the human as a “witness,” transducing this information and representing the machines to each other (157). S concludes this section with a discussion of culture’s current inability to think correctly about the human-machine relation; “culture is unjust toward the machine” (158), and this is illustrated by comparison to cultural stereotypes of foreigners, etc., which are the product of limited familiarity and experience; with greater familiarity and experience these stereotypes can be unlearned, and a better understanding achieved.
Finally S turns to his main point, which is the conditions giving rise to an improved cultural understanding of the relationship between humans and machines, a la mechanology:
The advent of the conditions allowing man to see the technical relation functioning in an objective way is the prime condition for the incorporation of the knowledge of technical reality and of the values implied by its existence into culture. Now, these conditions are realized in the technical ensembles employing machines that have a sufficient degree of indeterminacy. For man, the action of having to intervene as a mediator in this relation between machines grants him a situation of independence in which he can acquire a cultural vision of technical realities. … Only a situation in which there is a concrete link with machines and a responsibility toward them, but which is liberated vis-à-vis each one taken individually, can provide this serenity of having technical awareness. (159)
This perspective will not be achieved from a practical use of machines (governed by an instrumental “finality”), nor from the partial perspectives from below (viz., the workers) or above (owners, overseers, etc.):
It is rather difficult for a worker to know technicity through the aspects and modalities of his daily work on a machine. It is also difficult for a man who is the owner of machines and who considers them productive capital to know their essential technicity. It is the mediator of the relation between machines alone who can discover this particular form of wisdom. (160)
However, there is not yet a “social place” or role corresponding to this mediating perspective; it would be that of the production planning engineer, except that this role is also, like those of the owners and workers, governed by the limitations of “finality.” So what is needed is—surprise!--a “psychologist” or “sociologist” of machines, “what we might call a mechanologist.”
He ends with a discussion of the relation of this mechanology to its precursor, cybernetics. Cybernetics is clearly a first step, full of promise, but hampered by several limitations, which is why it needs to be transcended. In passing, S criticizes Wiener’s simplistic opposition of information to noise [S having, in this chapter, identified information as being, rather, intermediate between form and chance] and for his faith in homeostasis [as opposed to S’s serrated evolution]. More importantly, S takes issue with Wiener’s pessimism, as W has mistakenly, and unsuccessfully, been trying to get cybernetic understanding into the minds of the powerful. S sagely advises:
For it is difficult to make philosophers kings and kings philosophers. It often happens that philosophers who have become kings cease to be philosophers. The true mediation between technics and power cannot be individual. It can be realized only through the mediation of culture. For there is something that allows man to govern: the culture he has received; it is this culture that gives him significations and values; it is culture that governs man, even if this man in turn governs other men and machines. (161)
The power of culture comes from the “great mass” of the governed; power in this model flows upward, not downward from the elites. [This may sound at first like an almost anarchist or democratic sentiment, but it is rather that of the enlightened elite, who recognize the source of their power; cf. Ruskin.]
In a time when the development of technics was poor, the elaboration of culture by governed men was enough for the government to think the problems of the group as a whole: because it went from human group to human group via the government, the recurrence of causality and information was complete and accomplished. But this is no longer true: the basis of culture is still exclusively human; it is elaborated by the group of men; however, having gone through government, it returns and applies itself to the human group on the one hand and to machines on the other: machines are ruled by a culture that has not been elaborated according to them, and from which they are absent; this culture is inadequate for them and does not represent them. (162)
So culture, as the source of power and of understanding, fails to prepare us for the technical reality of our age because it has not caught up. Machines have yet to be properly “represented” in culture the way humans are (e.g., in literature, as S discusses). Thus, the task of the mechanologist is to transform culture by means of this more accurate representation, and to do this they need to understand the essence of technicity, not via “inductive study” [which led the cyberneticists astray] but by “a direct examination of technicity according to a genetic method that must be attempted, by employing a philosophical method” (163).
Haworth, Michael, (2020) “Automating Art: Gilbert Simondon and the Possibility of Independently Creative Machines.” Journal of Aesthetics and Phenomenology, 7:1, 17-32.
Rantala, Juho; and Mirka Muilu, (2023) “Simondon, Control, and the Digital Domain.” Theory, Culture & Society. https://journals.sagepub.com/doi/full/10.1177/02632764231201337