Ecological Rationality

In working on cognitive adaptations and their relation with rationality, a further step should be to identify the relevant features of the domain for which a particular cognitive

mechanism was selected and then proceed to decide in which other actual domains, displaying similar characteristics, the cognitive mechanism may still operate. Such an analysis has been developed by Gerd Gigerenzer and the ABC group. They have amended to a certain extent the hypothesis proposed by evolutionary psychologists such as Leda Cosmides and John Tooby and have gone one step further (see mainly Gigerenzer 2000;

Gigerenzer 2008; Gigerenzer & Selten 2001a; Gigerenzer et al. 1999). They propose a different view on the rationality question, according to which one needs ecological standards for assessing human reasoning. While evolutionary psychologists hold that evolution has endowed humans with an indefinite number of domain-specific cognitive mechanisms, Gigerenzer and his collaborators maintain that humans are equipped with “an adaptive toolbox” of fast and frugal heuristics. These heuristics make them, not only adaptively, but also ecologically rational.

3.1 Ecological validity

Gerd Gigerenzer and his colleagues introduce their proposal by appealing to three different forms of rationality, that is, bounded, ecological, and social rationality, and arguing that only an analysis of the relationship among these three aspects of rationality can lead to a comprehensive understanding of what it amounts to. By relying on these three forms of rationality, they hold that we can understand human behaviour and cognition in their adaptation to specific environments and discover the heuristics that guide them (Gigerenzer

& Todd 1999: 25). In particular, a leading role is attributed to the notion of ecological rationality, which is characterized as rationality adapted to the environment. Indeed, behind this proposal, there is the idea that “the success of boundedly rational heuristics depends on their ability to exploit the information structures in the ecological and social environment”

(Todd & Gigerenzer 1999: 360). More specifically, Gigerenzer holds that a heuristic is

“ecologically valid” when it is successfully applied in an environment that has a relatively fixed and homogeneous structure: the rationality of heuristics does not amount to logicality, but rather is ecological (Gigerenzer & Todd 1999: 18-19). In other words, a heuristic process, as well as any adaptation, can be evaluated as rational or irrational only in relation to an environment. In particular, Gigerenzer and Goldstein (1996: 665-666) have argued that ecological standards of rationality should replace classical ones as the appropriate benchmark against which to assess reasoning performances, and these standards should be based on: (a) the relevance of the environment; (b) the recognition of cognitive limitations;

and (c) the identification of specific algorithms existing in the human mind.

3.2 The adaptive toolbox

At a cognitive level, the ecological approach to rationality assumes that humans have at their disposal an adaptive toolbox (see, e.g., Gigerenzer 2001; Gigerenzer & Todd 1999).⁵ It consists of a set of domain-specific cognitive mechanisms, which can be inherited, learned or designed: just as screwdriver and hammers are designed for specific classes of activity, so are the cognitive mechanisms belonging to the adaptive toolbox. In characterizing those cognitive mechanisms, Gerd Gigerenzer (2001: 6) specifies that they include both Simon’s satisficing and what he calls fast and frugal heuristics. While the former involves “search across alternatives […] assuming that a criterion is given”, fast and frugal heuristics work in the opposite way, that is, they “search for criteria or cues, assuming that the alternatives are given” (Gigerenzer 2001: 121). According to Gigerenzer and his collaborators, heuristics are “fast” because they require simple computations and

“frugal” because they need little information to come to a conclusion; so, they are considered to be very effective in environments in which people have limited time and

5 It is “adaptive” because it is a product of natural selection, and is a “toolbox” because it is composed by a number of domain-specific cognitive tools that can be used in a flexible way (Gigerenzer & Todd 1999).

information for making decisions and drawing inferences (Gigerenzer & Todd 1999: 24).

Their frugality, however, does not compromise their effectiveness. Indeed, if fast and frugal heuristics are applied to the right kind of environment, they are very effective and lead reliably to good solutions. Gigerenzer and his collaborators hold that, in order to explain how a given heuristics works, one needs to identify (1) its building blocks (such as rules for search, stopping search, and drawing conclusions), and (2) the relevant characteristics of the environments in which that heuristic works well (e.g., Gigerenzer 2001: 501-502;

Gigerenzer & Todd 1999: 24-25).

By way of illustration, consider, for example, the “recognition heuristic”. Suppose that you are asked to decide which of two objects has a higher value on some criterion.

Gigerenzer and Goldstein (1999) have empirically demonstrated that, in order to make such a decision, people are naturally inclined to follow the following rule: “if one of two objects is recognized and the other is not, then infer that the recognized object has the higher value”

(Gigerenzer & Goldstein 1999: 41). This is what they call the “recognition heuristic”. In particular, how the recognition heuristic works could be illustrated by referring to an experimental case in which lack of information increases the success of a prediction (Goldstein & Gigerenzer 1999: 43). In this experiment, two groups of subjects, respectively composed by German and American college students, were given the name of two American cities and then asked which has a larger population:

Which U.S. city has more inhabitants: San Diego or San Antonio?

While 100% of German students answered correctly (San Diego), only 62% of American students did so. According to Goldstein and Gigerenzer, this surprisingly large difference is due to the fact that German students, who do not know so much about medium-sized American cities than American students, are unlikely to have ever heard of San Antonio, and thereby chose the city which they immediately recognize. This immediate choice

without any calculation seems to be appropriate in this and other similar cases because it is more likely that a foreign city one recognizes has more inhabitants than one that he does not. So, reliance on the heuristic that guides subjects to choose the city they immediately recognize has been shown to be useful. However, it would be an error to think that the recognition heuristic is completely reliable because its success depends on whether or not some particular contingent conditions occur. In particular, when relying on the recognition heuristic, one must know neither too much nor too little about the target domain. In the case mentioned above, the American students cannot apply the recognition heuristic because they know too much about American cities. Because of their familiarity with American cities, the recognition heuristic cannot be properly used as a filter in deciding which of them has more inhabitants. Similarly, if the German students had known nothing about American geography, they would have not been in a position to apply the heuristic.

3.3 Ecological rationality and normativity

Gigerenzer and his collaborators hold that classical principles of rationality, which are justified a priori, when brought to bear on real-world activities yield inappropriate standards of rationality and prove unecological. In their view, reasoning should be always evaluated in terms of its success in dealing with relevant problems which are situated in a natural or social environment (e.g., Gigerenzer & Todd 1999: 21-22). For example, although the recognition heuristic cannot be rationally justified, it usually leads to good outcomes when applied in an appropriate environment. So, in order to judge people’s reasoning performances, we should rely on an ecological standard of rationality according to which good reasoning will be always environmentally situated. Gigerenzer and his collaborators hold that these standards correspond to “multiple correspondence criteria relating to real-world”: reasoning performances must be compared with the actual

requirements of their environment, which include “making accurate decisions, in a minimal amount of time, and using a minimal amount of information” (Gigerenzer & Todd 1999:

22). In their view, those criteria should receive different weights in relation to the environment in which reasoning performances occur; in some cases, for example, making a quick decision can be more important than focusing on its accuracy and this must influence how the final decision is evaluated. Indeed, “[t]he function of heuristics is not to be coherent. Rather, their function is to make reasonable, adaptive inferences about the real social and physical world given limited time and knowledge” (Gigerenzer & Todd 1999:

22). So, every assessment should be constrained by the environment and people should not be asked to solve problems which are contextualized in environments that are not so relevant for them. On the contrary, according to Gigerenzer (1991; 1996), in the reasoning tasks created by Kahneman and Tversky subjects are located into artificial experimental contexts and receive information to which they are not ecologically adapted.⁶ Both the researchers of the “heuristics and biases” approach and Gigerenzer assume that people are naturally inclined to rely on simple heuristics but, from that, they draw different conclusions: according to Kahneman and Tversky, these heuristics are less reliable than classical principles derived from deductive logic, standard probability theory and expected utility theory, while, according to Gigerenzer, when applied in the appropriate context heuristics are about as reliable as the classical normative principles (for an interesting discussion, see Samuels et al. 2002).

6 See also Gigerenzer 1991; Gigerenzer & Hoffrage 1995; Gigerenzer & Hug 1992; Hertwig & Gigerenzer

4. The implications of adaptive and ecological rationality for the evaluative question