Introduction. The chapter reads System 4, against Beer’s own text, as “considerably larger than a sensor”. On this reading System 4 is the modeling and anticipation capacity of the viable system. It holds an account of the environment and of the system’s own relation to it. A model in this sense must:
- form expectations;
- compare incoming signals against them;
- revise its account quickly enough for the revision to be usable in action.
Sensing is “a component of that work rather than a description of it.”
The homeostat. The chapter reads the loop between System 3 and System 4 as carrying predictions rather than reports alone. In the usual reading the loop is a hand-over, with System 4 reporting what it has seen outside and System 3 turning the report into instruction for the operations. Beer describes something less one-directional: a “homeostatic loop between sensation and action”, and a feed-forward loop whose predictions “are compared with real-life outcomes”. If those passages carry that weight, what System 4 sends into the homeostat is a prediction about what the operations will show next, and an instruction issued on its strength is “a wager”, framed so that the next round of operational reporting can settle it. The chapter marks that reading as its own, supported by Beer’s passages rather than his phrase: “prediction channel” occurs in none of the eight Beer books searched, and it is System Five, not the command channel, that Beer routes System Four’s indications into.
A warning about vocabulary. The chapter notes that contemporary predictive-processing vocabulary fits this description closely. Any use of it is the authors’ comparison, not a claim about what Beer wrote or intended.
The missing forward model. The chapter reports a negative search result. In the defined Beer corpus, no function distinct from System 4 appears to do forward modeling (predicting the sensory consequences of the system’s own action). The chapter treats this as provisional: corpus boundary, editions, method, terms and date were not recorded. “A negative textual result with no documented method is a note to ourselves, not evidence.” Taken at face value, it gives two options, both costly:
- forward modeling belongs inside System 4, so teaching System 4 as scanning removes one of its central operations; or
- it does not, so the five-system architecture lacks a separately specified function needed to anticipate the consequences of one’s own action.
The chapter declines to choose between them by analogy with neural architecture.
The outer limit. System 4 models change in the environment while the system stays the same kind of system. It does not by itself account for a system becoming a different kind of system. That argument (morphogenesis) is left to Chapter 7.
Important authors. Stafford Beer (1926–2002), British cybernetician, author of the Viable System Model. The chapter does not name which of his books contain the System 4 passages.
Importance for cybernetics / VSM. The model requirement for System 4 is usually grounded in the good regulator theorem, which Chapter 3 examines (with Francis and Wonham’s internal model principle). Whether Beer’s own text carries an anticipatory architecture, and where the predictions sit in it, decides how much of that architecture the tradition kept when it kept the diagram.
Importance for the chapter. Carries “What Beer gave System Four to do”, “The model without a forward model” and “Change in the world, not change of kind”. Feeds items 2, 3 and 4 of “The receipt: what falls with the substrate”.
Sources in the reading list.
- Beer’s System 4 discussions — for the modeling and anticipation passages and the sentence that closes the sensation–action loop (BOTF 1981: 182–83, 189; HOE 1979: 235, 240).
- Chapter 3 of this book, The Law We Quote Is Not the Law Ashby Wrote — for the good regulator theorem and the internal model principle (referenced in receipt item 2).