Chapter 3 Consequentialism and adaptation
1.2 Consequentialism and the standard picture of rationality 89
1.3.1 Maximizing or satisficing?
Working on decision making, Simon showed that, while expected utility theory requires very complex calculations and complete information in order to single out the decision that
maximises one’s expected utility, people have too limited cognitive resources and usually lack the basic information needed to solve the decision problem they are faced with. As a result, it would be impossible for them to perfectly work out the outcomes of the alternative options among which they have to choose. Being things so, Simon held that optimization cannot be our ultimate goal when making decisions or performing other type of human behaviour. Specifically, according to him, people reason and make decisions under environmental and cognitive constraints, such as limited memory and the need to reduce information costs.1 Simon proposed an analogy to explain this point: “human rational behaviour […] is shaped by a scissors whose two blades are the structure of the task environments and the computational capabilities of the actor” (Simon 1990: 7). According to him, if you want to understand how the human mind works, you have to look at both blades of the scissor (see also Gigerenzer & Selten 2001b). Because of these constraints, in order to count as rational a subject ought to make satisficing choices; when faced with a decision problem, she ought not to seek optimal or maximizing solutions, but only acceptable solutions. As Simon puts it, a rational decision maker ought not to try to maximize her expected utility because that is often an impossible task: in ordinary life, all of us usually behave as “satisficers”. For example, if decision makers recognize and choose the type of response appropriate to the circumstance in which the choice is being made, satisficing avoids unsatisfactory outcomes and reduces cognitive effort. Within this normative framework, rationality requires of decision makers that they should search for information and process it in ways that are less expensive than classical models of rationality require (Simon 1976: 136).
1 In subsequent research, the first aspect has received an overwhelming amount of attention. But recently, as we will see, the interest for the second aspect is widely increased.
1.3.2 Is bounded rationality really normative?
In the fifties, Simon wondered whether the “[…] model of economic man provides a suitable foundation on which to erect a theory – whether it be a theory of how firms do behave, or of how they ‘should’ rationally behave” (Simon 1955: 99). Likewise, Gigerenzer and Selten (2001b: 6) have pointed out with respect to Simon’s view of rationality that
“bounded rationality means rethinking the norms as well as studying the actual behaviour of minds […]”.Understood in these terms, Simon’s proposal has to be considered as an alternative to the classical normative models of rationality, not an account that accepts them and analyzes whether and to what extent reasoning performances deviate from their normative principles. The real question is then whether Simon’s proposal succeeds in providing an appropriate normative framework for assessing human reasoning. As is noted by Nickerson (2007: 21), “a decision need be neither optimal nor even as close to optimal as one is capable of achieving in order to meet the dictates of rationality. How close to optimal a decision must be to be considered rational appears to be an open question”. On the one hand, any deviations from the classical normative standards of rationality may always be reconsidered as being reasonable enough, given some cognitive or environmental constraints that were previously disregarded. On the other hand, however, “there are so many ways in which rationality can be bounded that we can never be sure we have the right one” (Watts 2003: 66; as reported by Nickerson 2007: 358). In order to determine whether satisficing choices are good enough for a certain individual, an evaluator must know something about her, such as what her goals are and what her level of aspiration is: while having this information is sometimes possible, it is often too difficult a task for anyone.
Moreover, as is pointed out by Edmund Henden (2007), it is wrong to think that Simon and his followers hold that it is always rational to make satisficing choices, i.e., prefer good enough choices to optimal ones. This seems to be very counter-intuitive. A more plausible
way to understand their claim is to assume that, according to the supporters of bounded rationality, it is not always rational to aim at satisficing choices rather than optimal ones.
As Henden argues (2007: 341), “this weaker claim is consistent with the possibility that there may be circumstances in which genuine satisficing in fact would be irrational”.
Consequently, he goes on, pointing out that showing few examples in which being a satisficer is not rational is not sufficient to reject satisficing. Again, however, we should conclude that proponents of bounded rationality need an argument about what conditions need to be present in order for a case of satisficing to be rational (Henden 2007: 341-342).
1.4 From maximization to satisficing: a final balance
The comparison between the standard picture and the concept of bounded rationality provides useful suggestions for the development of any satisfactory consequentialist account of rationality. In particular, I identify two fundamental conditions that any consequentialist approach to rationality assessment should satisfy.
First, any consequentialist account of rationality that should be used as a benchmark against which to evaluate human reasoning must take human cognitive limitations into account. On the one hand, this condition focuses on the issue of the limitations of human rationality. On the other hand, however, it also assumes that by taking these limitations into account we can reach an adequate notion of rationality. As seen above, behaviour that is optimal in terms of the classical normative standards is seldom required for achieving people’s realistic, ordinary goals. According to Simon, maximization or optimization is often an impossible task because the calculations required in maximizing one’s utility or in achieving optimal solutions are usually computationally intractable. If so, the standard picture of rationality seems to require of humans to make inferences that are far beyond their computational capacities.
There is then a second condition which has important implications for how we should think about normative standards of rationality according to a consequentialist view:
consequentialism should reject “the assumption that the same standards of good reasoning apply in all environments – that they are context invariant” (Samuels et al. 2004: 170).
Clearly, different contexts can affect the efficiency of a reasoning strategy in different ways. Accordingly, the second condition states that an appropriate consequentialist account of rationality should index standards of evaluation to the environment in which reasoning takes place.
According to Simon (1981; 1990), these two requirements could be satisfied by developing a science of adaptive systems. By the expression “adaptive system”, Simon means “a system resulting from his reactions to modelling forces of the environment to which every system must be adapted in order to survive” (Simon 1990: 3). Following similar considerations, starting with the late eighties the notions of adaptation and environment have received increasing attention in the works of philosophers of mind, cognitive psychologists and anthropologists.