Open-System Behavior

Definition

A "closed" or "state-determined" system's next state depends only on its current state (S(t+1) = F(S(t))), which guarantees its behavior eventually settles into a repeating cycle — simple, and simple to predict. An "open" system's next state also depends on an input arriving from outside (S(t+1) = F(S(t), I)), so it has no single "the behavior," only a set of possible behaviors, one selected by whichever input actually shows up. Weinberg suggests writing this set with a capital "Behavior" to distinguish the whole repertoire from any one observed instance of it.

In the Book

Weinberg builds the open/closed distinction from a music box that cycles predictably until you kick it — the kick isn't part of the state, it's an input that selects among possible next cycles. He formalizes the idea with a vending machine whose behavior is governed by an explicit program: total coins inserted determines whether a candy bar releases, change is returned, or the machine keeps waiting. He then applies the same lens to people: the legal system treats a "murderer" as if his behavior were state-determined and will repeat, while the modern psychiatric view often treats a single act of murder as environment-triggered — one specific input ("the one last nag") arriving into an otherwise unremarkable Behavior set, comparable to Johnny, the "model pupil virtually all year" who sets fire to the library one Friday. Because open-system behavior resists reduction to a single line, Weinberg notes we compulsively force one anyway: by naming a typical instance, an exceptional-but-alarming instance (murderer, liar, cheater — note there is no equivalent single-act word for a habitual truth-teller), or a statistical average — or, more directly, by acting as the system's environment ourselves and constraining its input, as when a suspected time bomb is dunked in a pail of water rather than merely observed.

Why It Matters

This explains why "explaining behavior" by looking only inside a person, organization, or system routinely fails, and why the same actor can look wildly inconsistent across contexts without actually being unpredictable — the inconsistency is often just different points in a legitimate repertoire, keyed to different inputs. It also names a lever: since open-system behavior is selected by environment, outcomes can often be changed more effectively by changing what's fed in than by trying to redesign the system's insides.