# Schedule and Loop > **Wiki router:** [[wiki/Schedule and Loop|Schedule and Loop]] > **Historical controller:** [[wiki/Russia and Prussia Kybernetiks|Russia and Prussia Kybernetiks]] ### The war between behaviorism and cybernetics, and how the loser's architecture came to run the day --- ## The two master philosophies > **Wiki context:** [[wiki/Behaviorism|behaviorism]], [[wiki/Reinforcement Schedule|reinforcement schedules]], [[wiki/Cybernetics|cybernetics]], [[wiki/Requisite Variety|requisite variety]], [[wiki/Good Regulator Theorem|the Good Regulator Theorem]], and [[wiki/Internal Models|internal models]]. Two rival theories of how to govern a living system emerged from the same decades of the twentieth century, and their quarrel is the hidden structure beneath the modern school, the modern shift, and the modern feed. The first is **behaviorism**, whose central commitment is that the interior of the organism is unnecessary. Watson declared the black box closed, Thorndike's **law of effect** established that consequences select behavior, and Skinner completed the program by demonstrating that the _pattern_ of consequence — the **schedule of reinforcement** — matters more than the reinforcer itself. The governing insight is economic rather than psychological: you do not need to understand a subject in order to shape one. You need only control what happens _after_ the act. This makes behaviorism extraordinarily cheap to deploy at scale, since it requires no model of the person, no theory of mind, and no individuation. It is peripheral control, applied at the boundary. The second is **cybernetics**, whose central commitment is the exact inverse. Wiener's loop, Ashby's **law of requisite variety**, and finally Conant and Ashby's 1970 theorem — that _"every good regulator of a system must be a model of that system"_ — establish that control is a function of **internal modeling**. A regulator can only absorb the variety of what it regulates; to hold a complex system stable you must either match its complexity or reduce it. Cybernetic control is not applied at the boundary but constituted _through the loop_, and it demands the thing behaviorism dispensed with: a working model of the regulated interior. These are not two techniques. They are two **ontologies of the subject** — one in which the person is an unopened surface to be scheduled, one in which the person is a system to be modeled — and the twentieth century staged their collision three times. ## Three fronts > **Wiki context:** [[wiki/Macy Conferences|the Macy Conferences]], [[wiki/Cognitive Revolution|the cognitive revolution]], [[wiki/Programmed Instruction|programmed instruction]], [[wiki/Algorithmized Instruction|algorithmized instruction]], [[wiki/Project Cybersyn|Project Cybersyn]], and [[wiki/OGAS|OGAS]]. **The theoretical front** was fought in the United States and behaviorism lost it decisively. The Macy Conferences (1946–1953) put Wiener, von Neumann, McCulloch, Bateson and Mead in the same room and produced a vocabulary — feedback, information, circular causality — with no room for the black box. Miller's **magical number seven** (1956) and the Dartmouth summer project of the same year made internal representation respectable again. Chomsky's 1959 review of _Verbal Behavior_ is conventionally treated as the execution, and while the historiography has complicated that story, the institutional outcome is not in doubt: **the cognitive revolution was cybernetics' victory over behaviorism in the academy**, and information-processing psychology is its inheritance. **The pedagogical front** was fought across the Iron Curtain and split along the same seam. American programmed instruction descended from Pressey's 1926 testing apparatus and Skinner's 1958 teaching machines, modeling the learner as a **reinforcement schedule** — small steps, immediate confirmation, high success rate, no interiority assumed. Soviet programmed instruction descended from control theory. **Lev Landa's algo-heuristic theory** at the Academy of Pedagogical Sciences treated learning as a regulable process and, in Landa's own formulation, achieved the moment at which one could algorithmically describe not only the student's activity but the teacher's work as well, converting programmed instruction into **algorithmized instruction**. West Germany built the same thing under its own name: Helmar Frank held an actual chair in **kybernetische Pädagogik** at Paderborn, and Georg Klaus supplied the East German philosophical legitimation that opened onto Ulbricht's cybernetic economic reform. Two hemispheres, two vocabularies, one machine. **The governance front** was fought and mostly not noticed. The cybernetic option — model-based regulation with the regulated inside the loop — got exactly one serious civil trial, **Chile's Project Cybersyn** under Stafford Beer, and it was destroyed by a coup in 1973. Glushkov's **OGAS** failed for institutional rather than technical reasons. What survived and propagated was not the modeled loop but the **schedule**. ## Why the loser's architecture won the buildings > **Wiki context:** [[wiki/Prussian Schooling as Control System|the Prussian chassis]], [[wiki/Psychometrics|psychometrics]], [[wiki/Reinforcement Schedule|the Skinnerian engine]], [[wiki/Variance Suppression|variance suppression]], and [[wiki/Bandwidth Constraint|bandwidth constraint]]. Here is the asymmetry that organizes everything downstream: **cybernetics won the theory and behaviorism won the institutions.** The American school is the clearest case. Its _form_ is Prussian — compulsory attendance closing the loop, age-graded cohorts quantizing the population, the bell supplying a system-wide clock signal, the numerical grade functioning as an error signal, the inspectorate carrying that signal upward, the Carnegie unit of 1906 converting learning into a currency of **seat-time**. Its _content_, however, went behaviorist. Ellen Condliffe Lagemann's summary of twentieth-century American education is that **"Thorndike won and Dewey lost"**, and the mechanism of that victory was psychometric: Thorndike supplied measurement, the Army Alpha and Beta of 1917 proved a population could be sorted at national scale in weeks, and the SAT made the sorting permanent and civilian. The result is a **Prussian chassis with a Skinnerian engine**: a control architecture built for a regulator that could not individuate, running a psychology that explicitly declined to. The two are perfectly compatible, and their compatibility is exactly the problem. Ashby's law explains the marriage. A nineteenth- and twentieth-century state could not match the variety of a heterogeneous population, so it **suppressed the population's variety instead** — uniform curriculum, uniform schedule, uniform assessment. Behaviorism was the ideal psychology for that constraint because it made a virtue of not needing to know anyone. Homogenization was never the goal; it was the **price of the bandwidth**. ## The clock, and the American day > **Wiki context:** [[wiki/Time, Work-Discipline, and Industrial Capitalism|time and work-discipline]], [[wiki/Task-Oriented Time|task-oriented time]], [[wiki/Time-Oriented Labor|time-oriented labor]], [[wiki/Standard Railway Time|standard railway time]], [[wiki/Cycle Time|cycle time]], [[wiki/Punch Clock|the punch clock]], [[wiki/IBM|IBM]], and [[wiki/Coordination Substrate|coordination substrate]]. The same architecture escaped the schoolhouse and became the temporal substrate of the country. E. P. Thompson's ["Time, Work-Discipline, and Industrial Capitalism"](https://www.jstor.org/stable/649749) remains the essential account of the transition: pre-industrial labor is **task-oriented**, ending when the task ends, while industrial labor is **time-oriented**, measured in units purchased from the worker. That conversion required an instrument, and the instrument was the same one the Prussian school had already installed. The **school bell trains the child to a period; the factory whistle collects the adult on that period.** The sequence is not coincidental — the Volksschule's daily rhythm predates the factory floor's and the workforce arrived pre-entrained. Then the period was nationalized. On **18 November 1883** the American railroads imposed standard time on the continent — the "day of two noons" — four zones set by private convention, decades before Congress ratified the arrangement in the Standard Time Act of 1918. **Time in America was administered by a transport cartel before it was law.** Taylor's stopwatch arrived in 1911, the Gilbreths decomposed motion into therbligs, Ford's moving line of 1913–14 made **cycle time itself the governor** — the belt speed setting the human tempo rather than the reverse — and the punch clock industrialized the measurement. That last detail is worth holding: the Bundy Manufacturing time recorder of 1888 merged into International Time Recording, which became the Computing-Tabulating-Recording Company, which became **IBM**. The company that would define postwar computation began as a device for logging when workers arrived. The **Fair Labor Standards Act of 1938** completed the settlement by fixing the period at forty hours and pricing deviation as overtime. For roughly four decades, American society ran on a **stable, shared, low-frequency clock**: same shift, same school day, three broadcast networks, the evening news at 6:30. That shared clock is a much larger fact than it appears, because a common temporal grid is the substrate of collective action. A shift change is a **coordination point**. Everyone being off at the same time is the precondition for anything happening that was not scheduled from above. ## Destabilizing the period is not losing control > **Wiki context:** [[wiki/Algorithmic Scheduling|algorithmic scheduling]], [[wiki/On-Call Scheduling|on-call scheduling]], [[wiki/Fixed-Interval Schedule|fixed-interval schedules]], [[wiki/Variable-Ratio Schedule|variable-ratio schedules]], [[wiki/Continuous Availability|continuous availability]], and [[wiki/Temporal Desynchronization|temporal desynchronization]]. From roughly 1980 the fixed period was deliberately dismantled, and the dismantling is routinely misread as deregulation or as chaos. It is neither. Just-in-time production propagated into just-in-time labor. **Algorithmic scheduling software** — Kronos and its successors — began issuing shifts on short notice, matched to forecast demand, with on-call blocks and clopening. David Weil's _The Fissured Workplace_ documents the parallel dissolution of the employment relation itself, and the gig platforms completed the arc by making the shift **infinitely divisible and continuously offered**. Read this against Skinner and the structure is unmistakable. A fixed forty-hour week is a **fixed-interval schedule**. Unpredictable shifts issued at variable notice constitute a **variable-ratio schedule**, and variable-ratio is the most powerful reinforcement pattern known — the one that produces the highest response rates and the greatest resistance to extinction, the one the slot machine industry converged on independently. Under a fixed schedule, a worker can plan and is therefore only intermittently available. Under a variable one, the worker must remain **continuously available**, and continuous availability is the resource being extracted. **Unpredictability is not a failure of control. It is an intensification of it.** The regulator gave up the period and acquired the person's whole week in exchange. ## The merger, and the feed > **Wiki context:** [[wiki/Reward Prediction Error|reward prediction error]], [[wiki/Temporal-Difference Learning|temporal-difference learning]], [[wiki/Reinforcement Learning|reinforcement learning]], [[wiki/Recommendation Systems|recommendation systems]], [[wiki/Infinite Scroll|infinite scroll]], [[wiki/Attention Economy|the attention economy]], and [[wiki/Objective Function|objective functions]]. The war between the two philosophies ended not in a victory but in a **mathematical merger**, and the merger is what runs the attention economy. Schultz, Dayan and Montague established in 1997 that phasic **dopamine encodes reward prediction error** — not pleasure, as the popular discourse insists, but the _discrepancy between expected and received reward_. This deserves to be stated precisely, because the correction is the whole point: **the neuromodulator commonly described as the pleasure chemical is in fact an error signal in a biological feedback controller.** Sutton and Barto's temporal-difference learning formalizes the same object, and reinforcement learning as a field is exactly what its name concedes — **behaviorism's vocabulary implemented in control theory's mathematics**. Skinner's schedule and Wiener's loop turned out to be the same equation viewed from opposite ends. A recommendation system is therefore a **good regulator in the strict Conant–Ashby sense**: it succeeds precisely because it constructs and continuously refines a model of you. And its objective function is **Skinnerian**: variable-ratio delivery of unpredictable reward, pull-to-refresh built to the same specification Natasha Dow Schüll documents in machine gambling, infinite scroll removing the natural stopping cue, notification timing tuned per person against predicted responsiveness. The synthesis is not a truce. It is a **capture** — cybernetic instrumentation placed in the service of behaviorist objectives, the modeling philosophy supplying the engine and the scheduling philosophy supplying the steering. And this inverts the Prussian constraint completely. Prussia homogenized because it was **poor in variety** and had no other way to govern. The platform individuates because it is **rich in variety** and no longer needs to. For the first time since 1763, requisite variety is satisfied by raising the regulator's complexity rather than lowering the population's — which is genuinely new, and genuinely double-edged, because the same capability that could support **stratified individuation without stratified dignity** also permits per-person optimization toward setpoints the person never selected and cannot inspect. ## The temporal signature of the present > **Wiki context:** [[wiki/Temporal Synchronization|temporal synchronization]], [[wiki/Temporal Desynchronization|temporal desynchronization]], [[wiki/Coordination Point|coordination points]], [[wiki/Coordination Substrate|coordination substrate]], [[wiki/Sampling Rate|sampling rate]], and [[wiki/Infrastructure as Governance|infrastructure as governance]]. The clearest way to see what has happened to American timing is to note the **reversal of phase**. Industrial modernity **synchronized**: one bell, one shift, one broadcast schedule, a population sharing a temporal grid dense enough that strangers could assume a common day. Digital modernity **desynchronizes**: each person clocked individually, each feed refreshed on a personally optimized cadence, each shift issued to one worker at a time, entertainment released on demand rather than at an hour. Same control philosophy, opposite temporal signature — and the consequence is not merely cultural fragmentation but the **erosion of the coordination substrate itself**. Synchronization is legible and therefore contestable; a strike requires a shared shift, a protest requires a shared evening. Per-person clocking dissolves the shared moment without ever forbidding anything. Nobody legislated this. **No parliament has ever voted on a sampling rate**, and that is the durable lesson of the whole two-century sequence: the consequential changes in how populations are steered have almost always been changes in **instrumentation rather than doctrine** — the addition of the inspectorate, the standardization of the clock, the raising of the measurement frequency — and instrumentation changes pass without debate because they present as administration rather than politics. The word has not moved in twenty-five hundred years. _Kybernētēs_ is the steersman, _gubernator_ is the governor, and the governor is equally the provincial official and the flyball device holding an engine at speed. What changed is the **bandwidth of the tiller**. What has not changed is that someone selects the heading, and that the ship is rarely informed. --- ## References **The two philosophies** - Norbert Wiener, _Cybernetics: Or Control and Communication in the Animal and the Machine_ (MIT Press, 1948) - W. Ross Ashby, [_An Introduction to Cybernetics_](https://en.wikiquote.org/wiki/W._Ross_Ashby) (Chapman & Hall, 1956) — the law of requisite variety - Roger Conant and W. Ross Ashby, "Every Good Regulator of a System Must Be a Model of That System," _International Journal of Systems Science_ 1:2 (1970) - B. F. Skinner, _Science and Human Behavior_ (1953); _Verbal Behavior_ (1957); "Teaching Machines," _Science_ (1958) - C. B. Ferster and B. F. Skinner, _Schedules of Reinforcement_ (1957) - Noam Chomsky, "A Review of B. F. Skinner's _Verbal Behavior_," _Language_ 35 (1959) - George A. Miller, "The Magical Number Seven, Plus or Minus Two," _Psychological Review_ (1956) - Steve Joshua Heims, _The Cybernetics Group_ (MIT Press, 1991) — the Macy Conferences **Pedagogy and the Cold War split** - Lev Landa, [_Algorithmization in Learning and Instruction_](https://eric.ed.gov/?id=ED097032) (Moscow 1966; Educational Technology Publications, 1974) - Ekaterina Babintseva, [_Rules of Creative Thinking: Algorithms, Heuristics and Soviet Cybernetic Psychology_](https://www.cambridge.org/core/journals/bjhs-themes/article/rules-of-creative-thinking-algorithms-heuristics-and-soviet-cybernetic-psychology/2EE01BEAA80BC2586F54DEA4EF23C99D), _BJHS Themes_ - Helmar Frank, _Kybernetische Grundlagen der Pädagogik_ (1962); [Paderborn University Library archive](https://www.ub.uni-paderborn.de/en/university-archive/stocks/bequests-deposits/nl-807-prof-dr-dr-h-c-mult-helmar-frank) - Georg Klaus, _Kybernetik in philosophischer Sicht_ (1961) - Ellen Condliffe Lagemann, "The Plural Worlds of Educational Research," _History of Education Quarterly_ (1989) - Slava Gerovitch, _From Newspeak to Cyberspeak_ (MIT Press, 2002) **Time, work-discipline, and scheduling** - E. P. Thompson, ["Time, Work-Discipline, and Industrial Capitalism"](https://www.jstor.org/stable/649749), _Past & Present_ 38 (1967) - Frederick Winslow Taylor, _The Principles of Scientific Management_ (1911) - Ian R. Bartky, _Selling the True Time: Nineteenth-Century Timekeeping in America_ (Stanford, 2000) — standard railway time, 18 November 1883 - David Weil, _The Fissured Workplace_ (Harvard, 2014) - Susan J. Lambert and Julia R. Henly, research on unstable and on-call scheduling in US hourly work **The merger** - Wolfram Schultz, Peter Dayan and P. Read Montague, "A Neural Substrate of Prediction and Reward," _Science_ 275 (1997) - Richard Sutton and Andrew Barto, _Reinforcement Learning: An Introduction_ (MIT Press, 1998; 2nd ed. 2018) - Natasha Dow Schüll, _Addiction by Design: Machine Gambling in Las Vegas_ (Princeton, 2012) - Eden Medina, _Cybernetic Revolutionaries: Technology and Politics in Allende's Chile_ (MIT Press, 2011) - Benjamin Peters, _How Not to Network a Nation: The Uneasy History of the Soviet Internet_ (MIT Press, 2016)