Berry and DeCock, Computational Porosity

David M. Berry, and Christian DeCock (2026) “Computational Porosity: Benjamin, Lācis and Algorithmic Life.” Controversies of AI Society. https://doi.org/10.54337/aau.add.scai-11425

This interesting conference paper which will hopefully be further developed. The authors start off with a sophisticated discussion of Benjamin and Lacis’s concept of porosity, noting it as a product of the encounter between Naples and an exoticizing European gaze which then uses the concept to problematize the assumed arrangement of space in the North and elsewhere. Drawing on Jameson and Adorno, they also note the role the concept played in the development of Benjamin’s thought, particularly his concept of denkbild or “figure of thought.” B and DeC also point out the temporal, not just spatial, interpenetration of porosity. They tie the concept further to the Brechtian concepts (which Benjamin was influenced by) of estrangement (Verfremdung) and refunctioning (Unfunktionierung). [These specific aspects of Benjamin’s thought on porosity are not specifically returned to in the later discussion of computational porosity.]

"We argue that Benjamin and Lācis’s concept of porosity can be used to help understand how computational architectures structure contemporary social relations." (32)

The crucial difference is that computational porosity operates not through stone and concrete but through the material substrate of processors, networks and algorithms that increasingly mediate social existence. This includes the proliferation of enterprise software, algorithmic management systems, and platform-mediated labour that restructure how work is coordinated, controlled, and experienced in organisations.

They deploy the concept of computational porosity two ways:

1. “as a descriptive concept which helps understand how discretisation as a practice within computation is giving way to diffusion techniques”

2. “as a critical concept in the sense given by Benjamin and Lācis who saw it as an alternative to bourgeois ways of organising the lifeworld.”

Just as Naples resisted the rationalised planning of modern cities, computational porosity challenges organizational boundaries and hierarchies. In platform organizations, the distinction between employee and contractor, workplace and home, working time and leisure time becomes increasingly porous. Uber drivers, for instance, exist in a deliberately porous space where they are neither fully independent nor fully employed, where the car becomes simultaneously private property and workplace, where algorithms interpenetrate with human decision-making about when and where to work. (33)

The office diffuses into domestic home spaces and synchronous and asynchronous communication blur together making corporate surveillance and individual autonomy clash through activity monitoring software and flexible scheduling.

... computational systems create fluid boundaries between local and cloud processing, between human and machine cognition, and between private data and public circulation. The physical permeability [Benjamin and Lacis] identified in Naples’ buildings finds its contemporary parallel in the technical permeability of computational systems that allow data and processing to flow across previously distinct spheres and across planetary networks.

When we issue a voice command to ChatGPT or another LLM, the computation flows seamlessly between device, data centre and cloud, creating what appears as a unified interaction but which actually traverses across multiple computational domains. This technical arrangement mirrors the interpenetration of spaces that Benjamin and Lācis observed in Naples, though now operating through digital rather than architectural forms. Similarly, the diffusion processes that many AI systems now implement, make all cultural works diffuse and hybrid within the latent spaces of their neural networks, a process Berry (2025) calls diffusionisation. (33-4)

In a footnote:

The idea that porosity is now also an instrumental process, actuated through computational techniques for the diffusionisation of the lifeworld, raises interesting questions about how a practice of resistance can be integrated into the system. However, we want to suggest that porosity, as Benjamin and Lācis deploy it, points to the excess that cannot be captured fully, even when turned into a computational function. Thereby, computational porosity creates unforeseen lines of flight and potentials for resistance in social and political practice.

While computational porosity describes the broader phenomenon of interpenetrating boundaries between human and machine agencies, diffusionisation represents a specific technical manifestation of this porosity within AI systems. Through diffusion models, cultural artefacts are not simply stored or processed but become porous themselves as their features, styles, and meanings blur and intermingle within the latent spaces of neural networks. This technical process of diffusionisation thus intensifies the porosity Benjamin and Lācis observed in Naples’ architecture, as it operates not just on the level of infrastructure but on the very substance of cultural production itself. (34)

They describe using Google's “Smart Compose:”

As we compose, our thought processes become intertwined with algorithmic suggestions in ways that go beyond simple automation. The system learns from aggregate patterns of communication across millions of users, creating a kind of collective linguistic porosity where individual expression becomes mediated through statistically derived patterns.

This example thus layers all, or at least many of, the kinds of porosity they talk about (spatial, boundary-blurring, temporal, social, "diffusionist"). They also discuss agential porosity, “where human and machine decision-making become so entangled that attributing responsibility becomes difficult” (35). Through computational porosity, agency is distributed through [the assemblage] of human and non-human, with no clear [figure] in which it can be located. “This computational porosity obscures accountability whilst intensifying control and will create a number of difficulties unless reflexively understood.” They further discuss variations such as playful coding, and “workaround cultures” in which workers try to game the algorithms they are being controlled by.

Just as Neapolitans used architectural porosity to evade official functions and create alternative uses, workers develop tactics to game algorithmic management systems, exploit platform vulnerabilities, or repurpose enterprise software for unintended purposes. For example, call centre workers might share strategies for maximising metrics whilst minimising actual work, Deliveroo riders might use geographic quirks in the algorithm to secure better-paying orders, and remote workers might use mouse or keyboard automation to simulate work activity to evade surveillance software. These practices reveal the porous character of seemingly rigid computational management systems.

However, “The same flexibility that enables worker resistance also enables platforms to externalise costs, avoid employment obligations, and intensify exploitation through the blurring of work and non-work time” (35-6).

However, computational porosity is not merely analogous to architectural porosity. Rather, it represents an intensification and acceleration of the interpenetration of spaces and practices that Benjamin and Lācis observed. Contemporary computational systems do not simply enable movement between defined spheres but actively blur the boundaries between them. When we interact with AI systems or social media platforms, increasingly human and algorithmic agencies are diffused in complex ways. The “theatrical” dimension they identified in Naples’ architecture becomes literalised in computational systems that transform every interaction into a performance that can be captured. (36)

The concept of “explainability,” which Berry advocates in other writings, would create “epistemic porosity, where technical knowledge and democratic oversight must somehow coexist and interpenetrate” [it would be interesting to explore the connections between this concept and "legibility" per Enfield, et al.] Algorithmic management is another example of “temporal porosity” between past hiring decisions (e.g. encoded in training data), present applications, and future workforce composition.” There does not appear to be a set number of ways in which they want to discuss kinds of “porosity,” as they keep adding more, then circling back and revisiting ones discussed previously [perhaps one could argue there is a “porosity” to this mode of discussion.] It would be nice to have a set, clear list or overview paragraph of the forms or relations which computational porosity takes [not, of course, that Benjamin and Lacis bothered with anything of the sort], and how these tie back to their initial discussion of B&L’s porosity.

 [Whereas in my 2019 article I had looked at porosity primarily in terms of the relative openness or closedness of different spaces to interaction with each other, B&DeC seem more interested in how it creates mingled productions, blurred categories, “dissolved boundaries,” and recondite traces of (unevenly) distributed/delegated agency; this concern is likely linked to the project of “explainability” (which they do state in their conclusion); they are more interested in the politics of discursive articulation than in the politics of spatial articulation].

There are also possibilities for resistance: “For example, in adversarial machine learning, researchers and activists can deliberately exploit the porous boundaries of AI systems to reveal their limitations and biases. This recalls Benjamin’s (1930) attention to how Naples’ street urchins used the city’s new underground to subvert the purpose of this technology with playful chaos” (37). Apps like Signal “create deliberate impermeability within otherwise porous systems;” they give other examples workers’ collectives, unions, using apps.

“The European Union’s AI Act and similar regulatory frameworks create new porous spaces between technical systems and collective governance, opening possibilities for workers to contest how algorithms organise their labour” (38). The authors find parallels between use of silicon computing, and the tuff stone of Naples.

Whilst computational systems create new forms of algorithmic governmentality and platform capitalism, their porous character potentially generates possibilities for alternative social arrangements; a “chance to correct the incapacity of peoples to order their relationships to one another in accord with the relationship they possess to nature through their technology” (39, quoting Benjamin)

The key question then becomes how to mobilise computational porosity towards democratic ends. Just as Naples’ citizens used the city’s porous spaces to create autonomous zones and informal economies, we might identify how computational porosity enables new forms of collective organisation and resistance. For instance, the porous boundaries between local and cloud computing could support decentralised infrastructure projects that prioritise community control over corporate profit. The diffusional character of contemporary AI systems might be redirected towards collective knowledge production rather than data extractivism.

The conclusion turns more specifically to the subject of AI:

we can see generative AI’s outputs as a form of involuntary surrealism as they often contain unexpected juxtapositions, distortions, and a Verfremdung-­effect that can either enlighten or mislead, depending on context. Just as the Surrealists collaged disparate elements to jolt consciousness, AI often unwittingly collages fact and fiction.

Large language models, trawling through billions of data points and recombining them, might surface hidden cultural obsessions or biases in strange new forms. Indeed, image generators trained on internet data often produce biased or stereotyped images, spuriously classifying people by race, gender, sexuality, and personality .... When these biases appear blatantly in AI outputs, they can become an estranging mirror held up to society’s prejudices. It makes visible what is often obscured in polished human-made media, the deep-set biases in our collective imaginary. Thus, AI’s remix aesthetic can become a tool for critique, a way to see the “dream wishes” of society laid out unsparingly, much as Benjamin read the arcades of Paris as the dream wishes of the 19th century. (40)

Benjamin had seen that contemporary media and technology could be used for both fascism and freedom. B&DeC note that much current discourse on AI focuses on fears related to “boundary violations” between the human and the simulated. Such anxieties over borders have long been weaponized by fascism, and a better ground for progressive politics is needed. 

The question becomes not just how to maintain boundaries, but how to cultivate forms of porosity that enable flourishing rather than domination. Indeed, porosity functions dialectically in workplace struggles as it simultaneously enables new forms of worker coordination and new modes of managerial control. Workers will need to increasingly engage in collective reverse-engineering of opaque systems, sharing knowledge about how algorithms calculate work, predict demand, or evaluate performance. A critical concept of porosity must therefore resist managerial appropriation by foregrounding questions of power, exploitation, and resistance.

They turn to the question of “explainable forms of life” in the algorithmic age as a political, not just technical, issue. "This requires new institutional arrangements and technical practices that enable collective deliberation about how computational systems shape social life" (41).

The Revolution of Everyday Life, Chapter 14

Summary of Chapter 14: The organisation of appearances

Vaneigem’s summary:

The organisation of appearances is a system for protecting the facts. A racket. It represents the facts in a mediated reality to prevent them emerging in unmediated form. Unitary power organised appearances as myth. Fragmentary power organises appearances as spectacle. Challenged, the coherence of myth became the myth of coherence. Magnified by history, the incoherence of the spectacle turns into the spectacle of incoherence (e.g., pop art, a contemporary form of consumable putrefaction, is also an expression of the contemporary putrefaction of consumption) (1). The poverty of ‘the drama’ as a literary genre goes hand in hand with the colonisation of social space by theatrical attitudes. Enfeebled on the stage, theatre battens on to everyday life and attempts to dramatise everyday behaviour. Lived experience is poured into the moulds of roles. The job of perfecting roles has been turned over to experts (2). (123)

He begins with a Nkee quote from Ecce Homo: the “ideal world is a lie invented to deprive reality of its value, its meaning, its truth...” V adds, “And it is true that man lies because in a world governed by lies he cannot do otherwise: he is falsehood himself, he is trapped in his own falsehood."

All the same, nobody lies groaning under the yoke of inauthenticity twenty-four hours a day. There are always a few radical thinkers in whom a truthful light shines briefly through the lie of words; and by the same token there are very few alienations which are not shattered every day for an instant, for an hour, for the space of a dream, by subjective refusal. Words are never completely in the thrall of Power, and no one is ever completely unaware of what is destroying him. When these moments of truth are extended they will turn out to have been the tip of the iceberg of subjectivity destined to sink the Titanic of the lie. (124)

The bourgeoisie killed God and myth, but the replacement they created, the lie of the ideal, is weak and unsustainable.

Revolution was the bourgeoisie’s finest invention. It is also the running noose which will help it take its leap into oblivion.

Fascism is in a way a consistent response to this hopeless predicament. It is like an aesthete dreaming of dragging the whole world down with him into the abyss, lucid as to the death of his class but a sophist when he announces the inevitability of universal annihilation. Today this mise en scẽne of death chosen and refused lies at the core of the spectacle of incoherence. 

There is, however, an important difference between myth and its fragmented, desanctified avatar, the spectacle, with respect to the way each resists the criticism of facts. The varying importance assumed in unitary systems by artisans, merchants and bankers explains the continual oscillation in these societies between the coherence of myth and the myth of coherence. With the triumph of the bourgeoisie something very different happens: by introducing history into the armoury of appearances, the bourgeois revolution historicises appearance and thus makes the progression from the incoherence of the spectacle to the spectacle of incoherence inevitable. (124-5)

Thus, in the old unitary society there was an oscillation between the “coherence of myth” and the “myth of coherence;” this is replaced in the society of the spectacle with a progression from the “incoherence of the spectacle” to the “spectacle of incoherence.” V explains the first:

In unitary societies, whenever the merchant class, with its disrespect for tradition, threatened to deconsecrate values, the coherence of myth would give way to the myth of coherence. What does this mean? What had formerly been taken for granted had suddenly to be vigorously reasserted. Loud professions of faith were heard where previously faith was so automatic as to need no stating, and respect for the great had to be preserved through recourse to the principle of absolute monarchy.

Under pressure from the bourgeoisie, this oscillation finally pulls the unitary order apart:

There comes a time when the myth of coherence is so undermined by the criticism of facts that it cannot mutate back into a coherent myth. Appearance, that mirror in which men hide their own choices from themselves, shatters into a thousand pieces and falls into the public realm of individual supply and demand. The demise of appearances means the end of hierarchical power, that façade ‘with nothing behind it’.

None of the fragmentary ideologies of the spectacular era can more than momentarily replace this lost hierarchy, so “Eventually the decomposition of the spectacle entails the resort to the spectacle of decomposition” (126).

V now makes a very interesting argument about drama and technology:

The development of the drama as a literary genre cannot but throw light on the question of the organisation of appearances. After all, a play is the simplest form of the organisation of appearances, and a prototype for all more sophisticated forms. As religious plays designed to reveal the mystery of transcendence to men, the earliest theatrical forms were indeed the organisation of appearances of their time. And the process of secularisation of the theatre supplied the models for later, spectacular stage management. Aside from the machinery of war, all machines of ancient times originated in the needs of the theatre. The crane, the pulley and other hydraulic devices started out as theatrical paraphernalia; it was only much later that they revolutionised production relations. It is a striking fact that no matter how far we go back in time the domination of the earth and of men seems to depend on techniques which serve the purposes not only of work but also of illusion.

[This connects with the concept of “wonder” in the work of Aristotle, and the inventions of Hero of Alexandria; how unlike Vaneigem to have not included religion with theater and war as origins of illusion, on account of Hero’s hydraulic temple doors... though arguably, this was also “drama.”]

The birth of tragedy was already a narrowing of the arena in which primitive men and gods had held their cosmic dialogue. It meant a distancing, a putting in parentheses, of magical participation. This was now organised in accordance with a refraction of the principles of initiation, and no longer involved the rites themselves. What emerged was a spectaculum, a thing seen, while the gradual relegation of the gods to the role of mere props presaged their eventual eviction from the social scene as a whole.

Tragedy in turn is superseded by drama, and by comedy. In drama, “human society replaced the gods on stage” (127).

The cliché which likens life to a drama seems to evoke a fact so obvious as to need no discussion. So widespread is the confusion between play-acting and life that it does not even occur to us to wonder why it exists. Yet what is 'natural' about the fact that I stop being myself a hundred times a day and slip into the skin of people whose concerns and importance I have really not the slightest desire to know about?

This is not quite the same as being an actor in a play, because that actor retains a self to return to at the end of the performance.

The roles we play in everyday life, on the other hand, soak into the individual, preventing him from being what he really is and what really wants to be. They are nuclei of alienation embedded in the flesh of direct experience. The function of such stereotypes is to dictate to each person on an individual – even an ‘intimate’ – level the same things which ideology imposes collectively.

V talks about the increased personalization of the lies/spectacle (his example is television but obviously social media is an even better illustration). These do not work through the dissemination of ideas, but rather of gestures, portrayed on the screen and imitated by viewers.

What we have here is a school of gesture, a lesson in dramatic art in which a particular facial expression or motion of the hand supplies thousands of viewers with a supposedly adequate way of expressing particular feelings, wishes, and so on. Thus the still rudimentary technology of the image teaches the individual to model his existential attitudes on the complete portraits of him assembled by the psychosociologists. His most personal tics and idiosyncrasies become the means by which Power integrates him into its schemata. The poverty of everyday life reaches it nadir by being choreographed in this way. Just as the passivity of the consumer is an active passivity, so the passivity of the spectator lies in his ability to assimilate roles and play them according to official norms. (128)

Individuals acting out roles, personalities, mannerisms, etc. which they have consumed through media are thus being “choreographed” by the Spectacle.

We thus see the return of the original conception of theatre, of general participation in the mystery of divinity. But, thanks to technology, this now occurs on a higher level, and by the same token embodies possibilities of transcendence unavailable in ancient times.

As we have seen, the technical reproduction of magical relationships such as religious faith or identification resulted eventually in the dissolution of magic. Coupled with the demise of the great ideologies, this development precipitated the chaos of stereotypes and roles. Hence the new demands placed upon the spectacle. (129)

News stories are just assembled cliches, divided up into stereotypical categories (“Crimes of Passion, Political Affairs, Business Section, From the Police Blotter, Eating Out, etc., etc.”). There is the constant, meaningless juxtaposition of news items (cf. Latour): “The husband who kills his wife’s lover competes for attention with the Pope on his deathbed, and Mick Jagger’s underpants are on a par with Mao’s cap. It’s all one, everything is equivalent to everything else, in the perpetual spectacle of incoherence.” “The spectacle has to be everywhere, so it becomes diluted and self-contradictory.”

The spectacle’s degeneration is in the nature of things, and the dead weight which enforces passivity is bound to lighten. Roles are eroded by the resistance put up by lived experience, and spontaneity will eventually lance the abscess of inauthenticity and pseudo-activity. (130)

Seeing Like a State, Chapter 5

Summary of Chapter 5: The Revolutionary Party: A Plan and a Diagnosis

The “plan” is Lenin’s high modernist revolutionism, and the “diagnosis” is the response of two of his critics, Luxemburg and Kollantay. Scott sets up the stakes:

Lenin’s design for the construction of the revolution was in many ways comparable to Le Corbusier’s design for the construction of the modern city. Both were complex endeavors that had to be entrusted to the professionalism and scientific insight of a trained cadre with full power to see the plan through. And just as Le Corbusier and Lenin shared a broadly comparable high modernism, so Jane Jacobs’s perspective was shared by Rosa Luxemburg and Aleksandra Kollontay, who opposed Lenin’s politics. Jacobs doubted both the possibility and the desirability of the centrally planned city, and Luxemburg and Kollontay doubted the possibility and desirability of a revolution planned from above by the vanguard party. (147)

Scott draws from several key writings which reveal Lenin as a High Modernist. He discusses the idea of the party as a vanguard, and the implications of the term “mass:”

Although the terms became standard in socialist parlance, they are heavy with implications. Nothing better conveys the impression of mere quantity and number without order than the word “masses.” Once the rank and file are so labeled, it is clear that what they chiefly add to the revolutionary process are their weight in numbers and the kind of brute force they can represent if firmly directed. The impression conveyed is of a huge, formless, milling crowd without any cohesion—without a history, without ideas, without a plan of action. Lenin was all too aware, of course, that the working class does have its own history and values, but this history and these values will lead the working class in the wrong direction unless they are replaced by the historical analysis and advanced revolutionary theory of scientific socialism. (150)

Thus the vanguard party not only is essential to the tactical cohesion of the masses but also must literally do their thinking for them” (151).

This is the core of Lenin’s case against spontaneity. There are only two ideologies: bourgeois and socialist. Given the pervasiveness and historical power of bourgeois ideology, the spontaneous development of the working class will always lead to the triumph of bourgeois ideology. In Lenin’s memorable formulation, ‘the working class, exclusively by its own effort, is able to develop only trade-union consciousness.’ Social democratic consciousness, in contrast, must come from outside, that is, from the socialist intelligentsia.

Lenin draws on the metaphors of army, classroom, and also factory owners/executives [do we sense the disciplinary archipelago?]; and also architecture:

What the party has is the blueprint of the entire new structure, which its scientific insight has made possible. The role of the workers is to follow that part of the blueprint allotted to them in the confidence that the architects of revolution know what they are doing. (152)

It is surely a great paradox of What Is to Be Done? that Lenin takes a subject—promoting revolution—that is inseparable from popular anger, violence, and the determination of new political ends and transforms it into a discourse on technical specialization, hierarchy, and the efficient and predictable organization of means. Politics miraculously disappears from within the revolutionary ranks and is left to the elite of the vanguard party, much as industrial engineers might discuss, among themselves, how to lay out a factory floor.

Lenin’s revolutionary elite are “professionals,” not unlike those working for capitalist high modernism:

Just as Le Corbusier imagines that the public will acquiesce to the knowledge and calculations of the master architect, so Lenin is confident that a sensible worker will want to place himself under the authority of professional revolutionists. (153)

... the principle behind Lenin’s concerns ... stems from the sharply delineated, functional roles that the party and the working class each played. Class consciousness, in the final analysis, is an objective truth carried solely by the ideologically enlightened who direct the vanguard party. (154)

Lenin also uses a contagion metaphor, as in the party has to prevent the spread of dangerous ideas and practices among the proletariat, which could derail its (i.e., the vanguard’s) plan for revolution.

If we consider the vanguard party, as Lenin did, to be a machine for bringing about the revolution, then we see that the vanguard party’s relation to the working class is not much different from a capitalist entrepreneur’s relation to the working class. The working class is necessary to production; its members must be trained and instructed, and the efficient organization of their work must be left to professional specialists. The ends of the revolutionist and the capitalist are, of course, utterly different, but the problem of means that confronts each is similar and is similarly resolved. (156)

Lenin and Le Corbusier, notwithstanding the great disparity in their training and purpose, shared some basic elements of the high modernist outlook. While the scientific pretensions of each may seem implausible to us, they both believed in the existence of a master science that served as the claim to authority of a small planning elite. (157)

Both, of course, had the improvement of the human condition as their ultimate goal, and both attempted to attain it with methods that were profoundly hierarchical and authoritarian.

The actual revolutions of 1917 did not correspond with the plan Lenin had laid out in What Is to Be Done? “The high-modernist scheme for revolution was no more borne out in practice than were high-modernist plans for Brasilia and Chandigarh borne out in practice.” Obviously, Lenin and the Bolsheviks did not bring the revolution about; [to the contrary, they showed up late and derailed it]. “What Lenin did succeed brilliantly in doing was in capturing the revolution once it was an accomplished fact” (158).

What followed in the years until 1921 is best described as the reconquest, now by the fledgling Bolshevik state, of Russia. The reconquest was not simply a civil war against the ‘Whites’; it was also a war against the autonomous forces that had seized local power in the revolution. (159)

The model for the vanguard party depicted so sharply in What Is to Be Done? is an impressive example of executive command and control. Applied to the actual revolutionary process, however, it is a pipe dream, bearing hardly any relation to the facts. Where the model is descriptively accurate, alas, is in the exercise of state authority after the revolutionary seizure of power.

What What Is to Be Done? was in fact used for, was the creation of an official mythology after the fact, based on the glorious leadership of Lenin, etc.

It is perfectly natural for leaders and generals to exaggerate their influence on events; that is the way the world looks from where they sit, and it is rarely in the interest of their subordinates to contradict their picture.

After seizing state power, the victors have a powerful interest in moving the revolution out of the streets and into the museums and schoolbooks as quickly as possible, lest the people decide to repeat the experience. (160)

This altered story is told in such a way as to give it a sense of inevitability, thus diminishing both the actual contingency of events, and the spontaneous actions that played a large role:

It is exceptionally rare to find any historical account that stresses the contingencies. The very exercise of producing an account of a past event virtually requires an often counterfactual neatness and coherence. Anyone who has ever read a newspaper account of an event in which he or she participated will recognize this phenomenon. (390n37)

S next turns to the Lenin of State and Revolution, from 1917. Lenin here echoes the egalitarian positions of the anarchists and of Luxemburg (and Marx, as S also points out), but only strategically, as he wants to undermine the Kerensky regime; nevertheless his “high-modemist convictions still pervade the text” (161). Lenin was a fan of Fordism/Taylorism; he argued that capitalist infrastructure, bureaucracy, and technology simplify administration sufficiently, so that anyone can do it with basic enough training:

the great majority of functions of the old ‘state power’ have become so simplified and can be reduced to such simple operations of registration, filing, and checking that they would be quite within the reach of every literate person, and it will be possible to perform them for working men’s wages, which circumstance can (and must) strip those functions of every shadow of privilege and every appearance of official grandeur. (Lenin, quoted on p. 162)

The next source is Lenin’s Agrarian Question; more high modernist stuff, celebrating the inevitability of a centralized factory system imposed by capitalism (which is then to be taken over by the revolutionary proletariat). Lenin is a fan of electrification, in the sense of a centralized network that delivered power and efficiency everywhere: “The way that electricity worked was very much the way that Lenin hoped the power of the socialist state would work” (167).

Rosa Luxemburg (168ff) offered a very different view of revolution. “Luxemburg differed most sharply with Lenin in her relative faith in the autonomous creativity of the working class” (168). Luxemburg was nevertheless still a vanguardist, like Lenin:

Neither Lenin nor Luxemburg had what might be called a sociology of the party. That is, it did not occur to them that the intelligentsia of the party might have interests that did not coincide with the workers’ interests, however defined. They were quick to see a sociology of trade-union bureaucracies but not a sociology of the revolutionary Marxist party. (169)

Luxemburg, like Lenin, also used the metaphor of the party as like a factory manager, issuing orders which the worker should follow, because the manager/party sees the big picture, which the worker does not. However, this supremacy is not taken as far as in Lenin’s thought: “For Luxemburg, the party might well be more farsighted than the workers, but it would nevertheless be constantly surprised and taught new lessons by those whom it presumed to lead.” “Eschewing military, engineering, and factory parallels, she wrote more frequently of growth, development, experience, and learning.” Illustrating this is Luxemburg’s view on the role of strikes, standing in contrast to Lenin’s top-down view:

A strike or a revolution was not simply an end toward which tactics and command ought to be directed; the process leading to it was at the same time shaping the character of the proletariat. How the revolution was made mattered as much as whether it was made at all, for the process itself had heavy consequences. (170)

Per Scott, Luxemburg found Lenin’s “hierarchical logic” to be “both utterly unrealistic and morally distasteful.” ... “even if such discipline were conceivable, by imposing it the party would deprive itself of the independent, creative force of a proletariat that was, after all, the subject of the revolution.”

Lenin comes across as a rigid schoolmaster with quite definite lessons to convey—a schoolmaster who senses the unruliness of his pupils and wants desperately to keep them in line for their own good. Luxemburg sees that unruliness as well, but she takes it for a sign of vitality, a potentially valuable resource; she fears that an overly strict schoolmaster will destroy the pupils’ enthusiasm and leave a sullen, dispirited classroom where nothing is really learned. (171)

A footnote provides insight into Luxemburg’s [enlightenment] view of the transformation of proletarian consciousness through political action:

“...[The awakening of revolutionary energy could be effected] only by an insight into all the fearful seriousness, all the complexity of the tasks involved, only as a result of political maturity and independence of spirit, only as a result of a capacity for critical judgment on the part of the masses, which capacity was systematically killed by the social democracy for decades under various pretexts.” (Luxemburg, quoted in 393n79)

The mass strike, then, was not a tactical invention of the vanguard party to be used at the appropriate moment. It was, rather, the "living pulse-beat of the revolution and at the same time its most powerful driving-wheel, . . . the phenomenal form of the proletarian struggle in the revolution.” (172)

Luxemburg’s understanding of the revolutionary process, curiously enough, provided a better description of how Lenin and the Bolsheviks came to power than did the utopian scenario in What Is to Be Done?

Luxemburg argued that the Bolsheviks’ “dictatorial methods and their mistrust of the proletariat made for bad educational policy. It thwarted the development of the mature, independent working class that was necessary to the revolution and to the creation of socialism” (173).

Thus she attacked both the German and Russian revolutionists for substituting the ego of the vanguard party for the ego of the proletariat—a substitution that ignored the fact that the objective was to create a self-conscious workers’ movement, not just to use the proletariat as instruments.

Luxemburg wrote that the “collective ego of the working class” had to make its own mistakes to “learn the historical dialectic by itself,” not just be led by the vanguard party.

She believed that Lenin and Trotsky had completely corrupted a proper understanding of the dictatorship of the proletariat. To her, it meant rule by the whole proletariat, which required the broadest political freedoms for all workers (though not for enemy classes) so that they could bring their influence and wisdom to bear on the building of socialism. It did not mean, as Lenin and Trotsky assumed, that a small circle of party leaders would exercise dictatorial power merely in the name of the proletariat. (173-4)

L and L had fundamentally opposing understandings of socialism:

Lenin proceeded as if the road to socialism were already mapped out in detail and the task of the party were to use the iron discipline of the party apparatus to make sure that the revolutionary movement kept to that road. Luxemburg, on the contrary, believed that the future of socialism was to be discovered and worked out in a genuine collaboration between workers and their revolutionary state. … The openness that characterized a socialist future was not a shortcoming but rather a sign of its superiority, as a dialectical process, over the cut-and-dried formulas of utopian socialism. (174)

Scott notes how similar her view is to Malatesta’s. Her criticisms of Lenin were very prescient:

Unless the working class as a whole participated in the political process, she added ominously, “socialism will be decreed from behind a few official desks by a dozen intellectuals.”

Scott now turns to the writings of Alexandra Kollontay in support of the short-lived Worker’s Opposition within the Bolshevik party; her arguments are similar to those made by Luxemburg; only the working class could create actual communism, it could not be imposed from above by bureaucrats

Her experience of being patronized and condescended to as a representative of the women’s section seems directly tied to her accusation that the party was also treating the workers as infants rather than as autonomous, creative adults. (176)

Summarizing the link between the viewpoints of Lenin and Le Corbusier: “Each of these schemes implies a single, unitary answer discoverable by specialists and hence a command center, which can, or ought to, impose the correct solution” (178). In contrast, Scott finds in Luxemburg’s and Kollontay’s bottom-up visions of revolution the theme of mētis, which will be the ultimate lesson of this book:

Kollontay's point of departure, like Luxemburg’s, is an assumption about what kinds of tasks are the making of revolutions and the creating of new forms of production. For both of them, such tasks are voyages in uncharted waters. There may be some rules of thumb, but there can be no blueprints or battle plans drawn up in advance; the numerous unknowns in the equation make a one-step solution inconceivable. In more technical language, such goals can be approached only by a stochastic process of successive approximations, trial and error, experiment, and learning through experience. The kind of knowledge required in such endeavors is not deductive knowledge from first principles but rather what Greeks of the classical period called mētis, a concept to which we shall return. Usually translated, inadequately, as “cunning,” mētis is better understood as the kind of knowledge that can be acquired only by long practice at similar but rarely identical tasks, which requires constant adaptation to changing circumstances. (177-8)

Course in General Linguistics, Part 2, Chapters 1-3

Summary of Part Two, Synchronic Linguistics: Chapters 1-3

Chapter 1: General Observations

In these three initial chapters Saussure sets up some basic problems facing synchronic linguistics and starts to move toward solutions. In this first chapter he tackles the issue of how to study language synchronically, despite the fact that we know it is constantly changing. Thus he posits for static linguistics the study of the linguistic state:

In practice, a linguistic state occupies not a point in time, but a period of time of varying length, during which the sum total of changes occuring is minimal. It may be ten years, a generation, a century, or even longer. (99)

But since languages are always changing, however minimally, studying a linguistic state amounts in practice to ignoring unimportant changes. (100)

[It must be longer than a “point” in time, because langue has to take form as parole through [speech events] which exist in some kind of relations and contrasts with each other.] It is interesting that he seeks to ground this linguistic state as an actually existing, rather than merely an idealized or posited, “ideal type” condition, because it seems the latter would work just as well for his purposes.]

Chapter 2: Concrete Entities of a Language

Static linguistics aims to study real, “concrete entities,” not mere abstractions, but just how are these to be identified and delimited? “Any linguistic entity exists only in virtue of the association between signal and signification” (10); delineating sections out of a mere sequence of sounds, a solely phonetic transcription, would not suffice: “a sequence of sounds is a linguistic sequence only if it is the bearer of an idea.” [This argument receives a series of footnoted objections from translator Roy Harris, who calls it “an unfortunate way of putting the point”]. Saussure’s point, of course, is that the “concrete entity” has to have that dual quality of signal and signification:

This unified duality has often been compared with that of the human being, comprising body and soul. But the parallel is unsatisfactory. A better one would be with chemical compounds, such as water. Water is a combination of hydrogen and oxygen; but taken separately neither element has any of the properties of water. (102)

[This is an interesting analogy in that it refuses the potential philosophical dualism of seeing sound and concept as existing in two different realms, one physical and the other abstract or ethereal, like body and “soul”; instead they both exist, like the elements hydrogen and oxygen, in the same reality or realm; reminiscent of D&G seeing concepts, objects, etc. all as “bodies” in the Stoic sense.]

“A linguistic entity is not ultimately defined until it is delimited,” but how is this to be achieved? The definition S gives of the unit is “a segment of sound which is, as distinct from what precedes and follows in the spoken sequence, the signal of a certain concept” (emphases original). He then discusses some practical difficulties of such an approach (102-5).

Chapter 3: Identities, Realities, Values

Given these difficulties of discerning units, how can we ensure that “the fundamental concepts of static linguistics are directly based upon, or even merge with, the concept of a linguistic unit?” (106) [i.e., with the essence of how language actually works]. S explores this with the concepts of synchronic identities, synchronic realities, and synchronic value.

How can we understand the synchronic identity of a word across different spoken instances, when it can change in intonation, contextual meaning, and so on? He compares it to the way we talk about trains, such as “the 8:45 from Geneva” though on different days this is a different train, not the same one; or a city street that has been demolished and rebuilt, yet remains the “same” street. “How is it that a street can be reconstructed entirely and still be the same? Because it is not a purely material structure” (107). Indeed the material parts can all be replaced but that other, non-material part must endure, for the street to maintain its “identity.” The solution is to see the instances of the word, the train, the street, etc. as [tokens]:

Every time I utter the word Messieurs (‘Gentlemen’), I renew its material being; it is a new act of phonation and a new psychological act.

Turning to synchronic reality, S argues that the traditional division of language into parts of speech (nouns, verbs, etc.) is actually problematic and imprecise; he illustrates with the French expression ces gants son bon marché. The blame lies in the old-school, pre-scientific grammarians who invented these concepts:

Linguistics is always working with concepts originally introduced by the grammarians. It is unclear whether or not these concepts really reflect constituent features of linguistic structure. But how can we find out? And if they are illusory, what realities can we put in their place?

The solution is that both synchronic identity and synchronic reality depend ultimately on synchronic value, which the unit possesses by virtue of its place in the overall system of signification, which will be Saussure’s way of tackling static linguistics [thus giving birth to the entire phenomenon of structuralism].

Thus it can be seen that in semiological systems, such as languages, where the elements keep one another in a state of equilibrium in accordance with fixed rules, the notion of identity and value merge.

That is why in the final analysis the notion of value covers units, concrete entities, and realities. (109).

Enfield: Legibility and Agency

N.J. Enfield (2026) “Legibility and Agency.” American Anthropologist 128(1):39-51

In this very interesting article Enfield argues that “legibility is foundational to human agency” (39). He casts this as a rebuttal to Scott (1998) which is perhaps not really necessary. His larger point is that legibility is “both agency-increasing and agency-decreasing” (48) depending on social and institutional context.

He begins with an interesting example of Mrkaa boatmen in Laos reading the surfaces of rivers to avoid hazards, and ties this into literature on other culturally and practically situated practices of “reading” water. Such learned skills can be seen as “tuning:”

Whether one is navigating the streams of Mrkaa or the swells of Micronesia, whether one’s readings are direct from nature or mediated by instruments, whether one cites qualitative descriptions of numeric measures, people become tuned to reading the cues around them. This tuning is an adaptation that opens channels for flow, whether it be the flow of attention, reasoning, or response and action. This flow defines the agent’s relation with their environment. The relation is called legibility. (40-1).

Enfield’s beef with Scott is not really about the kind of “High Modern” state [and capitalist] legibility that Scott was critiquing, but more with the way that the term “legibility” is broadly used today in ways derived from Scott’s critique, without including the “high modern” or “state” or “topdown” kind of modifier that would distinguish it from the more general phenomenon which Enfield wants to use the term for: “Beyond this narrow idea of a tool of state control, the mechanism of legibility has far deeper significance for human action and interpretation in social systems” (41), at all scales of human agency and interaction. [And Enfield does, I think, recognize that this contradicts rather the spirit and the polemical intent of Scott’s book, than the explicit content, because Scott in turn includes various caveats regarding local kinds of knowledge and also the nuanced position of local elites (relevant to Enfield’s argument later in this paper).]

Enfield posits two related aspects of agency: 1) flexibility, control over action, planning, anticipation, etc.; and 2) accountability, “comprising others’ evaluations of their actions, and the rights and duties that their actions presuppose, create, and invoke;” in other words, the social and ethical significance made by the social group of the agent’s ability to act.

People use categorization to simplify their environment and their range of responses to it; E makes three points about the relation between agency and categorization (each of which are clearly related to both flexibility and accountability):

1. Categories are “sieves for decision-making and action.”

2. Categorization “sets public coordinates for social agency.”

3. Categorization coordinates “behavior, action, and understanding in groups,” so the interpretations and actions of the group’s members will converge [thus making their actions in turn more legible for the purpose of accountability].

Enfield turns first to first-order legibility, in the relation between an agent and their environment, drawing on Ingold, Gibson, Kockelman, Bourdieu, etc. Legibility is not merely a “property” of the environment: “it cannot be defined without also referring to an agent who reads it” (42) [and a context and purpose for which they are doing so]. “When this relationship achieves an enduring equilibrium for a group of agents, this gives rise to a legibility regime.”

In addition to learned legibility (attuning your attention to found signs in the environment) there is also constructed legibility (attuned to constructed signs produced by tools, instruments, etc.). Enfield also distinguishes first-order legibility regimes as to different “sources:” ancestral, goal derived, expertise derived.

He turns next to second-order legibility, “manipulating the environment to guide others’ action,” and more generally, anticipating the actions and interpretations of said others. As the perceptions and actions of other people are among the things we are most attuned to in any environment, second-order legibility is “foundational to the logic of communication.” Enfield discusses several ways that humans, honeybees, etc. do this; legibility is thus “an organizer of activity” in Lynch’s words (quoted on page 43).

In second-order legibility, we construct signs that organize others’ activities in a mechanism that is fundamental to the agency of social interaction. It builds on first-order legibility (the capacity to read an environment) by reading others’ capacity to read and then heeding that reading in the design of other-oriented action. (43)

[This is thus made possible by the three points E emphasized about categorization, and is also linked to the accountability of the others whose actions we are anticipating.] Enfield summarizes his argument:

Our goal here is to understand legibility as a resource for agency in the broadest sense, and thus to recontextualize a widely held framing of legibility as a modern imposition of large-scale states and technologies. I argue that legibility is inherent to all frameworks for situated action, from mētis to census, and that the legibility-agency relationship is foundational to social coordination and, in turn, to sociocultural institutions.

Enfield now sets about this recontextualization by discussing legibility in a series of social contexts; first off, the reading of signs to establish deviance.

Following the logic of second-order legibility, being socialized in a community means becoming tuned to how others will read your behavior and then being able to adjust as needed. … we create signs of ourselves that heed others’ patterns of assessment and nudge them as needed, but in large part, our self-presentation is a way to avoid attracting attention or response at all.

Thus, beat cops patrolled neighborhoods looking for “suspicious” characters; the trick was to anticipate this legibility regime and not arouse suspicion. Other examples include visitors to Mrkaa houses failing to understand the rules of seating location and thus standing out as outsiders.

Enfield then turns to the “high modernist” legibility famously critiqued by Scott in Seeing Like A State. He notes that

Critiques of high-modernist legibility are not critiques of legibility per se, but of the methods of states as agents of large-scale social order. … The problem is the supplanting of old, functionally evolved legibility regimes with hastily arranged, cartoonishly simple, and inflexibly quantitative new ones. (45)

[Though it is not in fact the hastiness, cartoonishness, or inflexible quantitativeness of such imposed legibility regimes that are central to Scott’s critique, it is rather that legibility regimes with these qualities can nevertheless better serve the State’s interests than locally developed and situated regimes, and so will often be adopted.] E argues further:

High-modernist legibility is not fundamentally different from any dynamic coordination device that sets terms for members of a social group who would wish to understand, influence, and align with each other. Whether it is enabling or constraining depends on who uses it and how.

He follows this with an interesting discussion of semiotic interception, in which signs are “intercepted” or [rearticulated/reterritorialized] to have means other than intended; an example is a king who is tricked into lowering his head on entrance through a low doorway, and thus “bowing” to the owner of the house.

The ever-present possibility of semiotic interception means that, to be agentive, people must anticipate and imagine potential secondary readings of their actions, some with significant consequences. Semiotic interception occurs when we are read in ways we did not foresee or intend. This ignorance of one’s own legibility is a de-agentivising force. It creates conditions for flow piracy in the domain of social action, a key factor in the exertion of political power. (46)

["Flow piracy" is a great term, apparently borrowed from hydraulics? which is unfortunately not returned to or fleshed out, though "pirates" are mentioned again later.]

Two phenomena linked to semiotic interception are preference capture, “when legibility thematizes and exploits an established preference,” and preference installation, “when legibility creates and imposes a new preference.”

The concept of semiotic interception is then expanded on with an account of how witchcraft accusations are used by the relatively powerful to take advantage of the relatively powerless, e.g., by local landowners to seize the property of a widow, by [articulating] an accident or sickness as a sign of witchcraft. Enfield primarily emphasizes how semiotic interception is made use of by the powerful; the question of course arises, isn’t it also used by those with less power, to resist, challenge, or redirect the powerful? It would be interesting to see this explored in greater depth. Enfield’s own examples include the trickster fooling the king, and people under surveillance fooling the police, specifically through semiotic interception.

Turning back to Scott, he takes issue with a passage in which Scott had argued that cadastral maps served the purposes of outsiders, not locals, because locals already “know” whose land is whose:

But if it were true that “everyone knows” whose land is whose, there would be no land disputes at the village level. Of course there are such disputes and they are all too often resolved in favor of the powerful. (48)

We would like to think that no idyllic “meadow by the river” would be sullied by such disputes. But without the publicly warrantable accountability that certain forms of legibility can offer, we are at the mercy of the locally powerful: landowners, resource-holders, lords, big men, et cetera. In these contexts, the words “everybody knows who holds the meadow by the river” are the words of a pirate, gangster, or corrupt village chief. Societal structurelessness may seem ideal when juxtaposed against the excesses of an overbearing, malevolent state with its half-empty glass of tyrannical legibility. But structurelessness is also “a smokescreen for the strong or the lucky to establish unquestioned hegemony over others.” [quoting Freeman, 1972]

This is a great and valid point, and even opens up some great further potential avenues for exploration (e.g. written constitutions and laws as legibility regimes, in order to make state power and process more transparent/contestable, a la Hammurabi, etc.). But first there seem to be some conceptual mashings-together which need to be cleaned up. First off, it is not clear that for Scott the opposite of State-imposed legibility is some complete lack of legibility; iirc there is a point where he describes complex urban street systems as “legible” to their inhabitants, though not to outsiders. Showing that achieved or learned legibility is not just State-imposed but an aspect of communication in all societies is a great point and a contribution, but it is needlessly limited by being posed as a rebuttal to Scott, who seems likely to have agreed. Second, “structurelessness” here is Freeman’s word, not Scott’s, and she is criticizing informality, not horizontality, per se (cf. Cohen, 2021, 12-13). Scott is not arguing for informality, against formality; he is arguing against centralizing systems of power-through-legibility which in turn leads him to posit the frequent inferiority of such systems in serving the needs of local people, compared to local and traditional systems. Those local systems can be formal or informal, and Scott frequently makes note of the fact that they are not necessarily lacking in exploitation and inequality—they simply will tend to be less exploitative and unequal than State-imposed systems. (For example, Scott’s “everybody knows who holds the meadow by the river,” evokes some locally legible agreement or understanding, a “publicly warrantable accountability,” which would not necessarily exclude formal mechanisms for maintaining and enforcing this understanding (I am thinking of the Andean villagers in Rappaport 1990); the fact that someone “holds” the meadow means that someone else is excluded from using it, so we are not talking about an idealized “structureless” commune of some sort). The general lesson that the “weapons of the weak” can often be the weapons of the strong is an important one, and it would be good to see it explored through Enfield’s concepts at greater length, but a polemical stance unhelpfully simplifies what is necessarily a nuanced subject.

Cohen, Yves (2021): “Horizontality in the 2010s: Social Movements, Collective Activities, Social Fabric, and Conviviality”, Mecila Working Paper Series, No. 40, São Paulo: The Maria Sibylla Merian Centre Conviviality-Inequality in Latin America, http://dx.doi.org/10.46877/ cohen.2021.40

Freeman, Jo (1972) “The Tyranny of Structurelessness” Berkeley Journal of Sociology. 17: 154-64.

Rappaport, Joanna (1990) The Politics of Memory: Native historical interpretation in the Columbian Andes. Cambridge University Press, New York.

Scott, James C. (1998) Seeing Like A State: How Certain Schemes to Improve the Human Condition Have Failed. Yale University Press, New Haven, CT.

Personal Knowledge, Preface and Chapter 1

 

Michael Polanyi (2005 [1958/1962]) Personal Knowledge: Towards a Post-Critical Philosophy. Routledge, London.

Summary of Preface

Michael Polanyi [aka “the other Polanyi”] states that his book is “primarily an enquiry into the nature and justification of scientific knowledge” (iv). He rejects the “ideal of scientific detachment” which has had a “destructive influence,” particularly in certain sciences, and seeks to “establish an alternative ideal of knowledge, quite generally.” “Personal knowledge,” the title of the book, might seem contradictory, as aiming at both the particular and subjective, and the objective and universal; but this will be resolved by “modifying the conception of knowing.”

He will build off Gestalt psychology, despite the fact that scientists “have run away from the philosophic implications of gestalt.”

Skilful knowing and doing is performed by subordinating a set of particulars, as clues or tools, to the shaping of a skilful achievement, whether practical or theoretical. We may then be said to become ‘subsidiarily aware’ of these particulars within our ‘focal awareness’ of the coherent entity that we achieve. Clues and tools are things used as such and not observed in themselves. They are made to function as extensions of our bodily equipment and this involves a certain change of our own being. Acts of comprehension are to this extent irreversible, and also non-critical. For we cannot possess any fixed framework within which the re-shaping of our hitherto fixed framework could be critically tested.

The “personal knowledge” gained through such participation with the world is not merely “subjective,” it is both personal and objective:

Comprehension is neither an arbitrary act nor a passive experience, but a responsible act claiming universal validity. Such knowing is indeed objective in the sense of establishing contact with a hidden reality; a contact that is defined as the condition for anticipating an indeterminate range of yet unknown (and perhaps yet inconceivable) true implications.

In addition to cut-off subjectivity, P also wants to challenge the idea of abstract, detached objectivity:

into every act of knowing there enters a passionate contribution of the person knowing what is being known, and that this coefficient is no mere imperfection but a vital component of his knowledge. (v)

Part One: The Art of Knowing

Summary of Chapter 1: Objectivity

Polanyi starts off by asking, what is the lesson of the Copernican revolution? The standard answer is that the lesson is basically that we (humans) are not the center of the universe, and need to understand our actual, puny place in it.

What precisely does this mean? In a full ‘main feature’ film, recapitulating faithfully the complete history of the universe, the rise of human beings from the first beginnings of man to the achievements of the twentieth century would flash by in a single second. Alternatively, if we decided to examine the universe objectively in the sense of paying equal attention to portions of equal mass, this would result in a lifelong preoccupation with interstellar dust, relieved only at brief intervals by a survey of incandescent masses of hydrogen—not in a thousand million lifetimes would the turn come to give man even a second’s notice.

However, P replies, it is impossible for us to actually think this way, and nobody, even scientists and philosophers, actually does so.

For, as human beings, we must inevitably see the universe from a centre lying within ourselves and speak about it in terms of a human language shaped by the exigencies of human intercourse. Any attempt rigorously to eliminate our human perspective from our picture of the world must lead to absurdity.

The actual lesson of the Copernican revolution was that Copernicus sought “intellectual satisfaction” over the self-importance of the previous model; he “gave preference to man’s delight in abstract theory, at the price of rejecting the evidence of our senses.” His system is every bit as anthropocentric as the Ptolemaic one, “the difference being merely that it preferred to satisfy a different human affection.”

If we call Copernicus’s conclusions more “objective” this means we think objectivity means privileging theory over sensory experience; “So that, the theory being placed like a screen between our senses and the things of which our senses otherwise would have gained a more immediate impression, we would rely increasingly on theoretical guidance for the interpretation of our experience, and would correspondingly reduce the status of our raw impressions to that of dubious and possibly misleading appearances” (3).

P lists three “sound reasons for thus considering theoretical knowledge as more objective than immediate experience”:

1. “A theory is something other than myself.” P gives the example of a map. “A theory on which I rely is therefore objective knowledge in so far as it is not I, but the theory, which is proved right or wrong when I use such knowledge.” [And thus, can be corrected and built on over time, fine-tuned through collective effort, unlike (or at least moreso than) immediate individual perception.]

2. A theory has its own structure and so “cannot be led astray by my personal illusions.”

3. “Since the formal affirmations of a theory are unaffected by the state of the person accepting it, theories may be constructed without regard to one’s normal approach to experience.” Unlike the Ptolemaic, the Copernican model holds true from any perspective in the universe.

Thus, when we claim greater objectivity for the Copernican theory, we do imply that its excellence is, not a matter of personal taste on our part, but an inherent quality deserving universal acceptance by rational creatures. We abandon the cruder anthropocentrism of our senses—but only in favour of a more ambitious anthropocentrism of our reason. In doing so, we claim the capacity to formulate ideas which command respect in their own right, by their very rationality, and which have in this sense an objective standing.

Copernicus’s theory then led on to many further discoveries which were made possible by it.

The intellectual satisfaction which the heliocentric system originally provided, and which gained acceptance for it, proved to be the token of a deeper significance unknown to its originator. Unknown but not entirely unsuspected; for those who whole-heartedly embraced the Copernican system at an early stage committed themselves thereby to the expectation of an indefinite range of possible future confirmations of the theory, and this expectation was essential to their belief in the superior rationality and objective validity of the system. (4)

A “theory which we acclaim as rational in itself is thereby accredited with prophetic powers” in that we expect it to lead to further discoveries and confirmations. This, and not the smallness of humanity, is the real lesson of the Copernican revolution:

It inspires us, on the contrary, with the hope of overcoming the appalling disabilities of our bodily existence, even to the point of conceiving a rational idea of the universe which can authoritatively speak for itself. It is not a counsel of self-effacement, but the very reverse—a call to the Pygmalion in the mind of man.

Polanyi now traces two lineages of scientific and philosophical thought beginning with the ancient Greeks. Copernicus is in the first, Pythagorean lineage, which finds the truth of reality in numbers, and thus stands for the fundamentality of theory over observation. After Copernicus, Kepler and Galileo are also in this lineage, which tends towards the mystical and rapturous. The Pythagoreans and their Renaissance successors are mistaken, but they are definitely on to something, in Polanyi’s estimation, in the way they express and value an emotional attachment to knowledge.

The other, less mystical and more materialist line, is the older Ionian one, the atomist Democritus being the avatar. The dominance of this line in European science led to the emergence of the mechanistic worldview, which separated pure mathematics, as an abstract realm, from empirical, and thus more “real,” reality. This leads ultimately to 19^th Century positivism, embodied in the perspective of Ernst Mach:

Scientific theory, according to Mach, is merely a convenient summary of experience. Its purpose is to save time and trouble in recording observations. It is the most economical adaptation of thought to facts, and just as external to the facts as a map, a timetable, or a telephone directory; indeed, this conception of scientific theory would include a timetable or a telephone directory among scientific theories. (8)

The mechanistic worldview, in turn, was replaced by relativity; P makes a big deal about Einstein not responding to the Michelson-Morley experiment, the point of which is that he was inspired not by empirical problems (as in the standard account which Polanyi is challenging), but by his own reflections and reason. After the acceptance of relativism as the dominant worldview, later experiments by Miller which supported the theory of aether were ignored, not on empirical but theoretical grounds. (Polanyi’s point here is not that Miller was correct and should therefore have been listened to; he is instead pointing out that he was ignored because his findings conflicted with what had become established theory, scientists being more interested in verification of laws than discovery.)

[At this point his argument is very reminiscent of Kuhn (not to mention Foucault), and there is a complicated relationship between Polanyi and Kuhn. Kuhn’s Structure of Scientific Revolutions came out in the same year as the revised version of this book. Kuhn cites Polanyi, and Polanyi expressed support for Kuhn’s position, but there is a more recent literature teasing out the similarities and differences between Polanyi’s and Kuhn’s (more “relativistic”) arguments.]

By virtue of the triumph of relativity over the previous mechanistic worldview,

modern physics has demonstrated the power of the human mind to discover and exhibit a rationality which governs nature, before ever approaching the field of experience in which previously discovered mathematical harmonies were to be revealed as empirical facts.

Thus relativity has restored, up to a point, the blend of geometry and physics which Pythagorean thought had first naïvely taken for granted. (15)

Polanyi seems to have two main points to make in this chapter: 1) fight against the separation of theory and observation into separate realms, with observation seen as more real and primary, and theory/reason as existing only in the mind, not also in nature; and 2) emphasize the value of a sense of wonder and beauty in science (particularly its theoretical side), which had been dismissed by the empiricists as “mystical.”

We cannot truly account for our acceptance of such theories without endorsing our acknowledgement of a beauty that exhilarates and a profundity that entrances us. Yet the prevailing conception of science, based on the disjunction of subjectivity and objectivity, seeks—and must seek at all costs—to eliminate from science such passionate, personal, human appraisals of theories, or at least to minimize their function to that of a negligible by-play. (16)

This is the myth of “objectivity" as mere empiricism: “This conception, stemming from a craving rooted in the very depths of our culture, would be shattered if the intuition of rationality in nature had to be acknowledged as a justifiable and indeed essential part of scientific theory.”

P castigates the use of “simplicity,” “symmetry,” and “economy” as euphemisms for “rationality.” He promises to elucidate his theory of personal knowledge as a corrective to all this nonsense.

We shall find Personal Knowledge manifested in the appreciation of probability and of order in the exact sciences, and see it at work even more extensively in the way the descriptive sciences rely on skills and connoisseurship. At all these points the act of knowing includes an appraisal; and this personal coefficient, which shapes all factual knowledge, bridges in doing so the disjunction between subjectivity and objectivity. It implies the claim that man can transcend his own subjectivity by striving passionately to fulfil his personal obligations to universal standards. (17)

 

The Eye of the Master, Introduction

 

Matteo Pasquinelli (2023) The Eye of the Master: A Social History of Artificial Intelligence. Verso, New York.

Summary of Introduction: AI as Division of Labor

Pasquinelli starts out with some apt quotes from Marx and Gramsci, and indicates the stakes and intent of the book with this great point:

In the twentieth century, few would have ever defined a truck driver as a ‘cognitive worker’, an intellectual. In the early twenty-first, however, the application of artificial intelligence (AI) in self-driving vehicles, among other artefacts, has changed the perception of manual skills such as driving, revealing how the most valuable component of work in general has never been just manual, but has always been cognitive and cooperative as well. Thanks to AI research – we must acknowledge it – truck drivers have reached the pantheon of intelligentsia. It is a paradox – a bitter political revelation – that the most zealous development of automation has shown how much ‘intelligence’ is expressed by activities and jobs that are usually deemed manual and unskilled, an aspect that has often been neglected by labour organisation as much as critical theory. (12-13)

He notes Sennet and others who have in fact recognized that “making is thinking;” his point resonates with the beginning Gramsci quote, to the effect that everyone is an intellectual. In this context he will take on the standard ideological account of AI and “intelligence:”

What is AI? A dominant view describes it as the quest ‘to solve intelligence’ – a solution supposedly to be found in the secret logic of the mind or in the deep physiology of the brain, such as in its complex neural networks. In this book I argue, to the contrary, that the inner code of AI is constituted not by the imitation of biological intelligence but by the intelligence of labour and social relations. Today, it should be evident that AI is a project to capture the knowledge expressed through individual and collective behaviours and encode it into algorithmic models to automate the most diverse tasks: from image recognition and object manipulation to language translation and decision-making. As in a typical effect of ideology, the ‘solution’ to the enigma of AI is in front of our eyes, but nobody can see it – nor does anybody want to. (13)

He gives the example of the self-driving car:

If the skill of driving can be translated into an algorithmic model to begin with, it is because driving is a logical activity – because, ultimately, all labour is logic. (14)

What, then, is the relationship between labour, rules, and automation, i.e., the invention of new technologies? This entanglement is the core problem of AI which this book seeks to explore.

The title of the book is from a quote from Engels:

‘The economical development of our actual society tends more and more to concentrate, to socialise production into immense establishments which cannot any longer be managed by single capitalists. All the trash of “the eye of the master”, and the wonders it does, turns into sheer nonsense as soon as an undertaking reaches a certain size. Imagine “the eye of the master” of the London and North Western Railway! But what the master cannot do the workman, the wages- paid servants of the Company, can do, and do it successfully. Thus the capitalist can no longer lay claim to his profits as “wages of supervision”, as he supervises nothing.’ (220n8; from Friedrich Engels, ‘Social Classes: Necessary and Superfluous’, Labour Standard, 6 August 1881.)

He will trace the history of AI more broadly through the developments of Babbage, Taylor, etc.

This book follows these analytical studies of the labour process through the industrial age up to the rise of AI, aiming to show how the ‘intelligence’ of technological innovation has often originated from the imitation of these abstract diagrams of human praxis and collective behaviours. (15)

Here is a thesis:

When industrial machines such as looms and lathes were invented, in fact, it was not thanks to the solitary genius of an engineer but through the imitation of the collective diagram of labour: by capturing the patterns of hand movements and tools, the subdued creativity of workers’ know-how, and turning them into mechanical artefacts. … this book argues that the most sophisticated ‘intelligent’ machines have also emerged by imitating the outline of the collective division of labour. In the course of this book, this theory of technological development is renamed the labour theory of automation, or labour theory of the machine, which I then extend to the study of contemporary AI and generalise into a labour theory of machine intelligence. (15-6)

[Wouldn’t this latter in fact be a subset of the former, not a generalization?]

Following Marx, P says the “master” is not in fact an individual but “an integrated power made up of ‘the science, the gigantic natural forces, and the mass of the social labour embodied in the system of machinery’” (16).

All of society has become a “digital factory” mediated by computer tech, social media, etc. “It is not difficult to see AI nowadays as a further centralisation of digital society and the orchestration of the division of labour throughout society.”

The thesis that the design of computation and ‘intelligent machines’ follow the schema of the division of labour is not heretical but receives confirmation from the founding theories of computer science, which have inherited a subtext of colonial fantasy and class division from the industrial age.

He gives an example of such fantasy from Turing, of his imagined Automatic Computing Engine, with humans divided into its “masters” who control it, and “servants” who it controls as its sensory organs; [aka those above and below the algorithm]; Turing argued that both classes of humans would be progressively replaced, though the masters would have more power to resist this:

Turing’s vision is contradicted today by the army of ‘ghost workers’ from the Global South, who, as Mary Gray and Siddharth Suri have documented, are removed from sight to let the show of machine autonomy go on. Paradoxically for Turing, AI came to replace mostly masters, that is managers, rather than servants – workers are needed (and always will be) to produce data and value for the voracious pipelines of AI and its global monopolies, and, on the other hand, to provide the maintenance of such a mega-machine under the form of content filtering, security checks, evaluation and non-stop optimisation. (17)

As gender studies scholars Neda Atanasoski and Kalindi Vora have pointed out, the dreams of full automation and AI such as Turing’s are not neutral but are historically grounded on the ‘surrogate humanity’ of enslaved servants, proletarians, and women that have made possible, through their invisible labour, the universalistic ideal of the free and autonomous (white) subject.

Writing a history of AI in the current predicament means reckoning with a vast ideological construct: among the ranks of Silicon Valley companies and also hi-tech universities, propaganda about the almighty power of AI is the norm and sometimes even repeats the folklore of machines achieving ‘superhuman intelligence’ and ‘self-awareness’.

Mythologies of technological autonomy and machine intelligence are nothing new: since the industrial age, they have existed to mystify the role of workers and subaltern classes. (17-18)

He quotes Simon Schaffer:

‘To make machines look intelligent it was necessary that the sources of their power, the labour force which surrounded and ran them, be rendered invisible.’ (Schaffer, quoted on page 18)

In addition to these ideological “speculative narratives” offered of AI by Silicon Valley futurists, there are also “technical histories” which voice corporate perspectives and “rarely consider the historical contexts and social implications of automation, and draw a linear history of mathematical achievements which reinforces technological determinism.”

But there are also critical histories, of “the social implications of AI from the standpoint of workers, communities, minorities, and society as a whole,” aka “critical AI studies.”

Within the expanding landscape of critical works, this book’s concern is to illuminate the social genealogy of AI and, importantly, the standpoint – the social classes – from which AI has been pursued as a vision of the world and epistemology. (19)

[Eg. AI as a class project.] Many histories trace AI and related computer techs to WWII and the cold war, but it goes back further than this, and is tied most directly to data collection by the “government machine,” not just wartime; P’s numerous references to Foucault are relevant to this point.

In summary, AI represents the continuation of data analytics techniques first supported by state bureaus, secretly cultivated by intelligence agencies, and ultimately consolidated by internet companies into a planetary business of surveillance and forecasting. (20)

However, this version of the story (which P shares with many of the above critical histories) presumes the targets of control are passive; P reiterates Gramsci’s “everyone is an intellectual” argument, to note that we also have to see (in classic Operaist form) that [the agency of workers is in fact first, and the ruling class’s move to control or coopt it is reactionary, secondary.]

this book aims at rediscovering the centrality of the social intelligence that informs and empowers AI. It also contends – in a more radical thesis – that such social intelligence shapes the very design of AI algorithms from within.

This book is intended as an incursion into both the technical and social histories of AI, integrating these approaches into a sociotechnical history that may identify also the economic and political factors that influenced its inner logic. Rather than siding with a conventional social constructivism and going beyond the pioneering insights of social informatics, it tries to extend to the field of AI the method of historical epistemology …

Where social constructivism generically emphasises the influence of external factors on science and technology, historical epistemology is concerned with the dialectical unfolding of social praxis, instruments of labour, and scientific abstractions within a global economic dynamics.

This links to the political epistemology of feminist critics of science and technology. P states that the title of the book has a double meaning, because current AI tech arose out of the drive to automate pattern recognition; he provides the etymology:

‘Master’ and ‘pattern’ share a common political etymology. The English term ‘pattern’ comes from the French patron and the Latin patronus. Both have the same root of the English ‘paternal’ and ‘father’, that is the Latin pater. The Latin patronus means also protector, also in relation to servants. The French patron has the meaning of leader, boss, or head of a community, which, in patriarchal contexts, implies a model to follow. (223n39)

He turns to the distinction between the original, symbolic AI, and the newer, connectionist AI:

The two lineages pursue different kinds of logic and epistemology. The former professes that intelligence is a representation of the world (knowing- that) which can be formalised into propositions and, therefore, mechanised following deductive logic. The latter, in contrast, argues that intelligence is experience of the world (knowing-how) which can be implemented into approximate models constructed according to inductive logic. (22)

[But is connectionist AI really inductive? It seems like it creates vast categories, and locates tokens inside those, and would thus be deductive. Wouldn’t induction would be a small-data, not a big-data approach? Also, LLMs do not have “experience of the world.”]

P points out that “neither of these two paradigms has managed to fully imitate human intelligence,” but the latter is better at pattern recognition and automation of tasks, and so is the foundation of the current boom. P implies the irony that the hype and ideology surrounding the first kind of AI is now applied to its rival.

Against a tradition which repeats the overly celebrated saga of the Dartmouth workshop, this book highlights the origins of artificial neural networks, connectionism, and machine learning as a more compelling history of AI about which, especially regarding Rosenblatt’s work, critical and exhaustive literature is still missing.

The rest of the chapter provides an overview of rest of the book, then summarizes:

This book proposes the labour theory of automation, in the end, not only as an analytical principle to dismantle the ‘master algorithm’ of AI monopolies but also as a synthetic principle: as a practice of social autonomy for new forms of knowledge making and new cultures of invention. (28)