Abstract
In chapters two and three I have argued that the cooperatively virtuous agent is rational, yet in certain situations — most notably, in the one-shot prisoners’ dilemma — she excludes considerations of rationality from deliberation and chooses cooperatively. Therefore, it is questionable whether being cooperatively virtuous is itself rational. This question will be the focus of attention in this and the next two chapters. We are looking for a justification of the cooperative virtues in terms of considerations of instrumental rationality.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
For example, Nozick (1974), Rawls (1971), and Williams (1971).
Prichard (1968). As Schmidtz (1995) points out, Prichard was not the first to claim this. Recently there have been a number of philosophers who take Prichard’s side on the legitimacy of the issue. For example, Brock (1977) and McDowell (1978).
Here I will discuss the latter possibility only. Below, I make some remaiks on the position that the choices of the cooperatively virtuous agent are rational once one has succeeded in becoming a cooperatively virtuous person. I
This is the same conclusion we reached in section12, chapter one where we saw that cooperative virtues cannot be replaced by a system of sanctions that wopld have the same effect as the cooperative virtues.
Gauthier (1986) defends an indirect justification of cooperative virtues. Though this interpretation is implausible in view of later work (in particular Gau er (1994b) and Gauthier (1994a)), he certainly gives reasons for it. The rational a ent, according to Gauthier, “makes a choice about how to make further choices; he chooses on utility- maximizing grounds, not to make further choices on those grounds.” Ga thier (1986, 158). Gauthier goes on to defend the rationality of these further choices.
Danielson (1992). The main difference between constrained maximization and reciprocal cooperation, as Danielson sees it, is that the latter disposition, unlike the former, requires the agent to exploit unconditional cooperators.
Kavka (1978). Kavka, unlike Gauthier and Danielson, thinks that an indirect approach implies making oneself less than completely rational on grounds of rationality.
I follow Elster’s discussion of the (im)possibility of achieving essential by-products. See Elster (1983).
As Elster observes correctly, the client should not just forget about her hypnosis, she should also forget about certain background memories and beliefs which make it possible for her to discern inconsistencies in her memories and beliefs.
The supposition is incorrect because it cannot be claimed that such agents face prisoners’ dilemmas in the first place. So any proposal of this kind should try to combine the following two propositions. (1) The cooperatively virtuous agent can find herself in situations of threatening sub-optimality, such as the one-shot prisoners’ dilemma, where these situations are defined in terms of the relevant preferences of all agents concerned. (2) The choices of the cooperatively virtuous agent are rational in that they maximize the satisfaction of the whole of her preferences. This is a hard task and it leads authors to come up with complicated constructions. For example, Sen (1974) argues that the mdtivations of the cooperatively virtuous agent should be regarded as preferences over orderings of preferences. These preferences are characterized as as if preferences. So the cooperatively virtuous agent has the standard prisoners’ dilemma preferences, but also prefers to have the preferences that come with, for example, the impure coordination game. She then proceeds to act as if she has the later preferences. The characterization of such preferences in terms of an as if mental state no doubt is inspired by Sen’s resistance agains behaviorist interpretations of preferences. However, I do not see how the concept of an as if prefe ence can satisfy proposition (2) above. If one does not actually have a certain preference, ac ing as if one has that preference does not make the action rational.
et another construction is proposed by Den Hartogh (1985, 334–337). Den Hartogh argues that cooperative virtues should be understood as synoptic preferences. The idea of a synoptic preference is that it is a preference the agent develops in response to the whole of the choice problem. For example, the cooperatively virtuous agent may find herself in a prisoners’ dilemma because of her first-order preferences. She realizes her predicament and, as a result of this realization, develops the preference for conditional cooperation.
is proposal has two fundamental weaknesses. First, synoptic prefe ences, which are process-oriented preferences, are assumed to be commensurable in some way with first-order preferences for outcomes. Such commensurability needs to be established without it being the case that the process-oriented preference is reduced to an outcome-oriiented preference. Secondly, the cooperative virtues refer to the existence of social norms. This is explicitly argued for by Den Hartogh. Therefore, they are essentially forms of respe t for these norms which can only be explained in terms of constraint. Constraint, however it not a preference, whether of the as if, or of the synoptic kind.
I abstract here from the question of how one could compare the utilities of the agent ex ante and ex post. Under standard utility theory such comparisons are impossible.
Proponents of this view can be found in the field of sociobiology, ethology, as well as in the social sciences. Among these are biologists such as Hamilton (1964), Lumsden and Wilson (1981), Trivers (1971) and Smith (1982). In addition there are many social scientists such as Axelrod (1984) and the economist Frank (1988).
It is important to remember that this is a metaphor. The individuals in the population do not need to have fiendish intentions, or any intentions whatsoever, in this “snuggle”.
This term was first introduced by Spencer (1867), one of the first defenders of Darwin’s theory. Darwin then adopted the term and used it in Darwin (1874). SCe also Richards (1987), Bradie (1994), and Dawkins (1978). Spencer never gave a careful definition of what he meant by fitness. This has caused some critics to argue that survival of the fittest is an empty concept.
Gould and Lewontin (1979) have argued that evolutionary explanations should not be interpreted as adaptationist explanations. The former are scientifically acceptable whereas the latter are not, according to these authors. The reasons for this are, first, that adaptationist explanations contain unverified steps; they are “just so stories”. The accusation Gould and Lewontin make is that adaptationist explanations are satisfied with giving elegant explanations without making sure that all steps in the explanation can be verified or are verified. Secondly, some typical characteristics of species are not adaptations at all, but are by-products of adaptations. For example, the reason human beings have chins has nothing to do with some function of the chin. Rather, it is the by-product of human beings having two rows of teeth. To complicate matters, such by-products can acquire a function later on in evolution. Though the trait in question then obviously has a function, it would be wrong to claim that it evolved for that function, as an adaptation to circumstances. Dennett (1995), who defends an adaptationist interpretation of evolution, argues against this view.
Dawkins (1978).
Elster (1979, ch. 1).
The occurrence of random mutations that cannot be explained by hereditary principles was first described by De Vries (1901).
Ruse (1986) gives a good analogy of the difference between an evolutionary process using only mutants as input and the process which uses inputs from the entire population stock of different traits. It is like the difference between writing an academic study in nineteenth-century Tahiti where the mail boat brought some literature, maybe even relevant literature, every month, and writing it in the Smithsonian Institute. If evolution depended on mutants only, it would be like writing the study in Tahiti and the resulting diversity and complexity would be far less than what we witness today.
Dawkins (1978).
Elster (1979, ch. 1) describes evolution in these terms.
In the late seventies a rival theory to this view was brought forward, the theory of punctuated equilibria.This theory states that evolution is a process of relatively slow or no change, interrupted by periods of relatively quick change. An outline of the main ideas of the theory of punctuated equilibria can be found in Eldredge (1983), Eldredge and Gould (1972) and Gould (1982).
Up to this date the philosophical debates about the unit of selection continue. For an excellent overview, as well as for a well-argued position in this debate, see Sober (1984).
It is often claimed by critics of evolutionary theory that the notion of survival of the fittest amounts to a tautological concept. According to them, fitness is defined as the number of fertile offspring. That is, the offspring of the survivors of natural selection. But then it seems that survival of the fittest means survival of the survivors, which clearly is a tautologous category. However this is a rather malicious reading of the concept. In evolutionary theory, fitness is best interpreted as the degree of adaptation to the environment. Survival is a strong indication of fitness. Survival is not identical with fitness, just as body size is not identical with weight.
This means that the explanandum of evolutionary theory is the emergence of traits in groups, or rather, populations, though the explanation itself is individualist. Therefore, one cannot use evolutionary theory to explain why a particular individual has certain characteristics. Evolutionary explanations are suitable for explaining the relative frequency of these traits in a population. See also Sober (1984) for a discussion.
Wallace (1864). This is remarkable because Wallace’s views at the time were mostly shared by Darwin, to the extent that an outline of Wallace’s theory sent to Darwin for his comments prompted the latter to publish Origin of the Species much earlier than he initially planned. See Richards (1987).
The first systematic argument for this view can be found in Williams (1966).
Dawkins (1978; 1986).
See the overview in Sober (1984).
For a critique of Dawkins’ use of the concept, see Smith (1978) and Stent (1977).
Lumsden and Wilson (1981, ch. 2 and 3) list an impressive array of studies that suggest universal similarities in perception, in cognitive capacities, and in “basic social norms” and “patterns of culture”. These studies range from psychology, anthropology, sociology to art history and economics.
Lumsden and Wilson (1981, 370).
Lumsden and Wilson (1981, p. 43–48).
Shepher (1971).
See Cosmides (1989).
Wilson (1975, 3).
Lorenz (1996).
Williams (1966).
See also the discussion in section 6.2, especially note 19.
I abstract here from the observation that these factors interact in rather complex ways. For example, heredity and rapid cycling time reduce variability among individuals. See Alexander (1980, 36–43) for a discussion.
Wilson (1975, 30).
This is true only a few species of vertebrates. Some species of rodent apparently fulfill these conditions. One example that has amazed scientists lately is the naked mole rat, a mammal that has more or less the same behavior patterns as the social insects, with similar phenotypic stratification.
Wilson (1975). Probably, Wilson thinks this is the case because of its work on ants. These conditions are true for most species of ants. Wilson (1971)
Alexander (1980, 47–53); Smith (1982, ch. 13).
For example, Mackie (1978). Admittedly, Mackie moves away from evolutionary theory saying he is concerned with social evolution instead of biological, but this makes it worse. For this means that he is using some sort of functionalism to explain the emergence of norms. These types of explanations are suspicious because they often, if not always, lack a feedback mechanism to explain why the effect of the explanandum causes the explanandum to emerge. See Elster (1989b).
his discussion of Dawkins (1978) on the concept of group selection, Mackie admits that this is the weak point even in his own theory. But he insists that in biology, as opposed to the social sciences, talk of group selection is allowed. He claims that ESS-es that are detrimental for the population cause these populations to go extinct, while on the other hand, populations with optimal ESS-es flourish and will take over the territory of the extinct population. This, of course, assumes that Wilson’s conditions mentioned above are satisfied. Whether this is the case depends very much on the specific circumstances in which species live. Apart from this, it does not strengthen Mackie’s theory at all. Mackie claims to have identified a feedback mechanism for biological evolution, but in social evolution, which is at issue here, such a mechanism has yet to be established.
Hamilton (1963; 1964).
Wilson (1971).
Although there are indications that through some freak arranging of the chromosomes in the zygotes a similar effect can be seen in the termite family. See Colman (1983).
Colman (1983).
Hamilton (1963). The pioneering work on imprinting was done by Konrad Lorenz and is reported in several publications, for example, Lorenz (1981). A systematic treatment can be found in Hess (1973). Incidentally, the mechanism of imprinting is one of the central examples of secondary epigenetic rules which Lumsden and Wilson (1981) give.
Trivers (1971).
This is the case if the chance of future encounters (a) is larger than the ratio of the decrease in fitness as a result of helping (c) and the increase in fitness as a result of being helped (b). In formula:,v> db. This is the sanie inequality that is discussed in section 2.7.
Trivers (1971) reporting the findings of Eibl-Eibesfeldt (1955).
See section 2.7.
Represented in this way we abstract from the asynchronous character of the original situation. Obviously the two individuals do not rescue each other simultaneously. Note that the natural interpretation of these percentages as the associated risk of a general policy of rescuing others when in need, already implies that we are dealing with a repeated two-person game.
See for example, Alexander’s discussion of reciprocal altruism, Alexander (1980, 48–51). Alexander, like many biologists, argues that to understand human cooperation, one needs to combine kin selection with reciprocal altruism. However, he does not give a detailed model in which the two are combined. Moreover, Axelrod has acknowledged Trivers’ theory on several occasions as a major inspiration for his work on tit-for-tat. See Axelrod and Hamilton (1981), Axelrod (1984; 1986).
Trivers (1971, 54).
Eibl-Eibesfeldt (1959).
Moreover, the range of cooperative behaviors that Trivers seeks to explain is extremely diverse. For example, it includes n-person cooperation among humans For genuine n-person cooperation, it is hard to see how either interpretation of reciprocal altruism is compatible with it. Yet Trivers cites it as evidence of the existence of reciprocal altruism.
It has been argued that reciprocal altruism presupposes too much for this reason. The abundance of psychological predispositions needed as a defense against cheaters are considered too complicated to have evolved in an evolutionary processi Colman (1983) presents an overview of this particular line of criticism.
us stated this objection is not very convincing. The notion that certain traits are too complex to have evolved has been discredited on numerous occasions Iby evolutionary scientists where it involved physiological characteristics (e.g., the structure of the eye). To gain plausibility, the objection needs to bring out why fine-tuned cognitivle and behavioral dispositions are fundamentally different from these characteristics.
Frank (1987) and Frank (1988).
Hence his name for the model, the commitment model. Frank (1988).
This is a common misunderstanding of the phenomenology of guilt. Guilt presupposes that one knows one has violated some rule. This, in turn, presupposes the ekistence of such a rule. To explain the stability of such a rule as a result of guilt-avoidance alone is circular. Guilt is not a cooperative virtue. It refers to the presence of a rule which is deemed authoritative. It is a sign of being cooperatively virtuous to some extent, but not a cooperative virtue itself.
Frank is unclear just how these distributions have to be interpreted. Are individuals assigned a different S-value at each encounter? Or are the S-values given and do these distributions describe the population? Moreover, what exactly is the nature of the variance? Is it due to slightly different signals in each individual or should they be attributed to the perception of the potential partners? Franssen (1997, 171) shows that it matters for the interpretation and plausibility of the model as a whole how we interpret these distributions.
If the intervals do not overlap at all, S, is a guarantee for the honesty or dishonesty of i and if the intervals overlap completely, S, is totally unreliable as a signal of i’s disposition. The overlapping of the two intervals can be interpreted, as Frank does, as incomplete mimicry of the honest trait by dishonest individuals. Harrington (1989) shows how important Frank’s assumption Uo < U„ is. Once this does not hold, it no longer follows that a population of only dishonest agents can always be invaded by honest agents.
Franssen (1997, 169) notes that in fact Frank lets each agent in this population play the following game: This way Frank’s model avoids the possibility of unintentional exploitation between two honest agents. An honest agent who thinks her opponent is dishonest will choose to work alone.
Frank is unclear just how these distributions have to be interpreted. Are individuals assigned a different S-value at each encounter? Or are the S-values given and do these distributions describe the population? Moreover, what exactly is the nature of the variance? Is it due to slightly different signals in each individual or should they be attributed to the perception of the potential partners? Franssen (1997, 171) shows that it matters for the interpretation and plausibility of the model as a whole how we interpret these distributions.
If the intervals do not overlap at all, Si is a guarantee for the honesty or dishonesty of i and if the intervals overlap completely, Si is totally unreliable as a signal of i’s disposition. The overlapping of the two intervals can be interpreted, as Frank does, as incomplete mimicry of the honest trait by dishonest individuals. Harrington (1989) shows how important Frank’s assumption UD< U„ is. Once this does not hold, it no longer follows that a population of only dishonest agents can always be invaded by honest agents.
The distributions Frank uses are derived from two identical normal distributions around µD=2 and r’,=3. These are “cut off’ at two standard deviations from the mean and then renormalized. The result is two partly overlapping bell curves. Frank (1988, 263).
Franssen (1997) shows convincingly that this expression for the expected utility implies that agents know their own signal value, i.e., they know how honest they will appear to others.
S* depends also on the shape of the distribution of SH and SD. See also Harrington (1989).
This assumption is necessary for otherwise it is by no means certain that the honest partner, with S > S*, will enter into cooperation with the honest agent with S > S*. I will discuss the plausibility of this assumption in the next section.
Harrington (1989) points out that this can only be true in a sufficiently small population. In an infinite population where h-40, the chance that the other is an honest agent is infinitesimally small regardless of his value of S.
Frank’s critics show that this is the case only in virtue of his assumptions about fo and f„ and the fraction x2 - xl, which turn out to be quite restrictive. See Harrington (1989).
- x2
ank introduces a further refinement to the model, by assuming that the observation of people’s S-value is not costless, which results in a more robust equilibriUm. That does not take away the worries that Harrington formulates. Harrington’s observations hold both for the basic and the refined model. I will not go into the refined model here either.
The experiment is already described in Frank (1988), where some preliminary results are given. The results of the entire experiment are published in Frank, Giiovich, and Regan (1993).
Surveys of this research are in DePaulo and Rosenthal (1979); DePaulp, Zuckerman, and Rosenthal (1980); and Zuckerman, Paulo, and Rosenthal (1981).
Frank gives some reasons why this result may be less important than I suggest here. The most important one being that laboratory situations do not resemble real-life situations. Another consideration which he does not mention is that there may be interference with the content of the discussion during the thirty minutes that the subjects were given before reaching their predictions. Orbell, Kragt, and Dawes (1988) shows that the content of the discussion, especially whether or not individuals are allowed to exchange promises with regards to how they are going to play the game, raises the level of cooperative choices. Dawes, Kragt, and Orbell (1990) explain this not by referring to the possibility of making better predictions, but by reference to a concept of group-identity.
Of course I am not claiming that individuals really have to know all this in the full blown sense epistemologists talk about. It is sufficient that they behave “as if’ they know.
Incidentally, this would explain the nature of the distributions of f (S). These distributions then are the representations of the “noise” in the perception of the signals. That is, an individual knows very well its own signal and its own type (D or H), but has difficulties in determining the exact nature of the potential partner. The distributions then are not averages of the two subpopulations (i.e., the honest and dishonest sections of the population), but describe the chance of perceiving a particular signal in each encounter. This fits rather poorly with the idea that evolutionary models, like that of Frank, are supposed to describe averages of the population instead of individual psychological and perceptual characteristics.
This means that the strict biological explanation is to be suplplemented with a developmental explanation.
See Franssen (1997, 170).
Tinbergen (1952).
Gould (1977).
The example is that of Frank (1988, 96–97).
Davies and Halliday (1977).
Note that the full disclosure principle derives from the fact that each competitor knows its own signal value, but is unsure of that of the other. This is consistent with the observations regarding the commitment model in section 2.6. The full disclosure principle has a number of interesting applications in economics. In his classic paper, Akerlof (1970) introduces the full disclosure principle to explain the extreme drop in price between new and second-hand cars.
Frank, Gilovich, and Regan (1993).
Franssen (1997, 189–201).
Especially Herrnstein (1970), Ainslie (1975), Ainslie and Herrnstein (1981), Ainslie (1982), and Ainslie (1992).
This is one of the more robust results of experimental psychology. The effect can be found in many settings. Also, there is evidence that this effect is “hard-wired”, i.e., it is the result of evolutionary processes that have influenced the older brain structures.
The term was coined by Herrnstein (1970). Of evolutionary theory and individual psychological descriptions. Secondly, the
Frank (1988, 84).
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2002 Springer Science+Business Media Dordrecht
About this chapter
Cite this chapter
Verbeek, B. (2002). Evolution of the Cooperative Virtues: An Indirect Justification. In: Instrumental Rationality and Moral Philosophy. Theory and Decision Library, vol 33. Springer, Dordrecht. https://doi.org/10.1007/978-94-015-9982-5_6
Download citation
DOI: https://doi.org/10.1007/978-94-015-9982-5_6
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-6026-6
Online ISBN: 978-94-015-9982-5
eBook Packages: Springer Book Archive