p.561 I believe that observed regularities of behavior can be fruitfully understood as "behavioral rules" that arise because of uncertainty in distinguishing preferred from less-preferred behavior. Such uncertainty requires behavior to be governed by mechanisms that restrict the flexibility to choose potential actions, or which produce a selective alertness to information that might prompt particular actions to be chosen.

p.561 predictable behavior will evolve only to the extent that uncertainty prevents agents from successfully maximizing.

p.562 Standard choice theory... assumes for the purpose of theoretical explanation that there is no gap between an agent's competence and the difficulty of the decision problem to be solved (hereafter called a "C-D gap")*... the presence of a C-D gap will introduce uncertainty in selecting most preferred alternatives, which will tend to produce errors and surprises. Such mistakes are by their nature unpredictable and erratic... I believe it is mistaken and that essentially the opposite view is true.

*Posing the problem in terms of a gap in an agent's decision competence relative to the difficulty of a decision problem was suggested to me [Heiner] by Axel Leijonhufvud.

p.563 it is not the absence of a C-D gap, but rather its presence which conditions regularity in behavior. [JLJ - think of solving a Rubik's cube. The cube presents to us a deeply scrambled configuration. We need to proceed by following a step-by-step procedure. Good/advanced "cubers" also follow a step-by-step procedure, but one that allows sequences to be skipped if "beneficial" situations/orientations are discovered by chance. God would simply unscramble the cube in a minimal number of steps, but in what would appear to be a haphazard method. In summary, a typical cuber would need a regular procedure, and would proceed in a predictable way. Heiner also uses a Rubik's cube example on p.563.]

p.565 In general, there is greater uncertainty as either an agent's perceptual abilities become less reliable or the environment becomes more complex.

p.565 the more general issue is whether some process - conscious or nor - will cause (or prevent) an "alertness" or "sensitivity" to information that might prompt selection of an action.

p.567 to satisfy the Reliability Condition, an agent must ignore actions which are appropriate for only "rare" or "unusual" situations. Conversely, and agent's repertoire must be limited to actions which are adapted only to relatively likely or "recurrent" situations.

p.568 An agent need only be capable of determining when to select particular actions from a limited range of allowable alternatives. To do so does not require an ability to understand why the resulting behavior patterns evolved. [JLJ - sounds like something useful for machine implementation. The machine can gather data, ask itself questions, construct diagnostic tests, but it will ultimately never, I feel, truly "understand".]

p.569 Predictable behavior... will evolve to the extent that agent's are unable to maximize because of uncertainty.

p.570 uncertainty becomes the basic source of predictable behavior.

p.571-572 Genuine uncertainty results from a gap in an agent's decision competence relative to the difficulty in selecting more preferred alternatives, so that error and surprise cannot be avoided... uncertainty may involve... the ability to identify potential actions which might be selected, or contingencies that might affect the consequences of future behavior.

p.572 it is the alertness or sensitivity to information that determines the patterns and complexity of rules manifested in behavior.

p.574 My departure from standard choice theory was suggested by the general pattern of animals having a larger C-D gap than humans, yet regularity in their behavior is much more noticeable than for humans... greater uncertainty will in general require a more inflexible structure of rules

p.585 I have argued that uncertainty is the basic source of predictable behavior, and also the main conditioning factor of evolutionary processes through which such behavior evolves. Uncertainty exists because agents cannot decipher all of the complexity of the decision problems they face, which literally prevents them from selecting most preferred alternatives... In contrast, standard economics analyzes the special case of no uncertainty in selecting most preferred options.