Cultural Transmission, Disproportionate Prior Exposure, and the Evolution of Cooperation

by Noah Mark
Citation
Title:
Cultural Transmission, Disproportionate Prior Exposure, and the Evolution of Cooperation
Author:
Noah Mark
Year: 
2004
Publication: 
American Sociological Review
Volume: 
69
Issue: 
1
Start Page: 
144
End Page: 
149
Publisher: 
Language: 
English
URL: 
Select license: 
Select License
DOI: 
PMID: 
ISSN: 
Abstract:

Reply to Bienenstock andMcBride

Cultural Transmission, Disproportionate Prior Exposure, and the Evolution of Cooperation

Noah Mark

Stanford University

thank Bienenstock and McBride (2004 [henceforward B&M]) for their considera- tion of my paper and for the opportunity to revisit issues my paper raised. B&M offer three major criticisms of my paper. (1) My model's tendency toward cooperation depends on a decoupling of behavior and fitness that is the- oretically untenable. (2) My model is not a viable explanation for cooperation in small groups and thus has ambiguous relevance to the understanding of real-world cooperation. (3) My model's tendency toward cooperation is not robust to minor changes in its specification. B&M conclude that my "model cannot account for the evolution of cooperation except when there is a large initial population of cooperators" (p. 142).

Before addressing each of these criticisms individually, I note an important characteristic of B&M's comment: They do not identify the relevance of these criticisms to my paper's con- tribution. My paper called attention to a new the- oretical argument about cultural transmission that may be important to understanding the prevalence of cooperation. I used a formal model to illustrate this argument. However, the logic and substantive significance of this argu- ment are independent of any particular model and can be understood without a formal model.

The theoretical argument is this: Suppose people learn cooperative and exploitative behav- iors by interacting with people who practice these behaviors-i.e., they acquire cooperative and exploitative behaviors through cultural

Direct correspondence to Noah P. Mark, Department of Sociology, Stanford University, Stanford CA 94305-2047 (nmark@stanford.edu). I thank Jennifer Curnming, James Kitts, Michael Macy, Miller McPherson, Cecilia Ridgeway, Joseph Ruff, and ASR Editor Jerry Jacobs for helpful comments and discussion.

transmission. Also, suppose the amount of expo- sure a person has to a given behavior positive- ly affects the probability that she or he will acquire that behavior. Then people who have become cooperative will on average have been exposed to more cooperative people and/or more cooperative behavior than have people who have not become cooperative, and people who have become exploitative will on average have been exposed to more exploitative people andlor more exploitative behavior than have people who have not become exploitative. In other words, cooperators have had dispropor- tionate prior exposure to cooperation, and exploiters have had disproportionate prior expo- sure to exploitation. Accordingly, cooperators have disproportionately benefited from the cooperative acts of others, and exploiters have disproportionately been hurt by the exploitative acts of others. If the benefits of being the tar- get of cooperative acts tends to make a person more influential as a behavioral model (and if the costs of being exploited make a person less influential as a behavioral model), then the dis- proportionate prior exposure generated by cul- tural transmission will create a cultural evolutionary force toward cooperation.

B&M do not challenge the logical coherence or the plausibility of this argument. Also, they do not even address this argument as if it were something that we could understand independ- ent of a particular formal model. A critique of my analysis is important only if we can use it to change, refine, or reject the above argument. Thus, B&M9s critiques of my model are, at best, undeveloped. However, individual considera- tion of these critiques reveals that they contain incorrect claims, rest on untenable assumptions, ignore what is new about my model, and redis- cover properties of my model that my paper already described.

COMMENT AND REPLY 145

Nevertheless, each of these criticisms has an important motivation. Considering them clari- fies misunderstandings and suggests one direc- tion for improving my model so that it more effectively illustrates my argument.

THE DECOUPLING OF BEHAVIOR AND FITNESS

B&M claim that in my model, behavior and fitness are decoupled that this decoupling is what creates my model's tendency toward coop- eration, and that this decoupling is at odds with the theoretical motivation for my model's assumption that individuals tend to imitate those who have received large payoffs in the past. The characteristic of my model to which B&M are referring when they speak of the decou- pling of behavior and fitness is this: In my model, the fitness level of maximally fit coop- erators cannot be acquired by cooperating, and the fitness of minimally fit defectors cannot be acquired by defecting. This is B&M's most important criticism. It does not reveal any flaw in my theoretical argument about cultural trans- mission, and, as I discuss below, no character- istic of my model is incompatible with its theoretical motivation. However, considering this criticism can help us to distinguish an unnecessary and potentially distracting char- acteristic of my model from the characteristic of my model that draws attention to and illus- trates my argument.

As I discuss in my paper, the high fitness level of new cooperators (C3s) and the low fit- ness level of new defectors (DOs) is what cre- ates my model's tendency toward cooperation. The fitness level of a new cooperator is the fit- ness level one acquires by having defected against a cooperator in the previous round. What I did not clarify was that a portion (213) of a new cooperator's fitness results from her prior expo- sure to a cooperator, and the remaining portion

(113) results from her prior defection. My the- oretical argument is about the former and not the latter. In other words, it is about how coopera- tors' disproportionate prior exposure to coop- eration elevates the mean fitness of cooperators relative to the mean fitness of defectors and creates an evolutionary force toward coopera- tion. It is not about how the prior defections of new cooperators could create an evolutionary force toward cooperation. Thus, an ideal model for illustrating my argument would be one in which the fitness of new cooperators was ele- vated only by prior exposure to cooperation, not the combination of prior exposure to coop- eration and prior defection. Work in progress develops such a model and shows how dispro- portionate prior exposure alone creates an evo- lutionary force toward cooperation. Although my present model was sufficient to reveal the new argument that is my paper's focus, the fact that, in my model, the switching ofbehavior by actors also elevates the mean fitness of coop- erators relative to the mean fitness of defectors potentially distracts attention from this argu- ment.

I have noted how B&M's criticism helps to suggest an improvement to my model, but I still need to directly address their criticism. The term "decoupling" mischaracterizes the feature of my model that B&M call "decoupling." Their concern is that, in my model, an actor's fitness does not always reflect the effect of her own behavior on her own fitness, and they call this less-than-perfect reflection "decoupling." The fact that actors' fitness levels do not perfectly reflect the effects their own behaviors have on their own fitness levels does not mean that behavior and fitness are decoupled. The fitness levels of actors can reflect more than one effect of their behavior. In my model, the fitness lev- els of actors partly reflect the effects their own behaviors have on their own fitness levels and partly reflect the effects their behaviors have on the targets of their behaviors. The latter is illus- trated by high fitness cooperators and low fit- ness defectors. The fitness of high fitness cooperators is not decoupled from cooperation; it is the consequence of having been the target of cooperation. Likewise, the fitness of low fit- ness defectors is not decoupled from defection; it is the consequence of having been the target of defection. In my model, behavior and fit- ness are partly coupled by the effects of actors' own behaviors on their own fitness levels and are partly coupled by actors' disproportionate prior exposure to the behaviors they display. B&M's claim that behavior and fitness are decoupled in my model rests on the implicit assumption that there is only one mechanism- the effect of an actor's own behavior on her own fitness-that can legitimately couple behavior and fitness in an explanation for coop- eration. This atomistic view denies the socio- logical understanding that differences in the benefits individuals receive often reflect factors

46 AMERICAN SOCIOLOGICAL REVIEW

other than differences in abilities to acquire benefits. My model reflects the sociological understanding that the benefits an individual has received may partially reflect the structural posi- tion she has occupied-i.e., the individual(s) with whom she has interacted-as well as her own ability to acquire benefits.

Showing that a mechanism other than the effect that a behavior has on the fitness of the individual who performs it can couple behav- ior and fitness and that such a mechanism can elevate (rather than lower) the relative mean fitness of cooperators is not new (Hamilton 1964). Nor is the employment of an under- standing of social structure in an explanation for the evolution of cooperation (Axelrod 1984; Sober and Wilson 1998; Takahashi 2000; Trivers 197 1). What is new about the mechanism I propose-disproportionate prior exposure-is that it is generated by cultural transmission itself, even in the absence of a clustered social struc- ture, assortative pairing, or selective coopera- tion.

As noted above, B&M claim that the char- acteristic of my model that they refer to as decoupling is at odds with my model's assump- tion that individuals tend to imitate those who have received large payoffs in the past. This claim is incorrect. What B&M see as problem- atic in "decoupling" is that the mean fitness of actors of a given behavior is not fully deter- mined by-and does not perfectly reflect-the effect that behavior has on the fitness of actors who perform it. B&M claim that this charac- teristic of my model is at odds with an empiri- cally substantiated theoretical idea about status: Perceptions of a person's competence positive- ly affect that person's status and the influence that person exerts (Driskell and Mullen 1990; Wagner, Ford, and Ford 1986)'. Contrary to

' Note that B&M's claim of incompatibility rests on an implicit assumption that defection in a pris- oner's dilemma is an example of competence. From the perspective of theory on status (Berger and Zelditch 1998; Wagner, Ford, and Ford 1986), this assumption is invalid. For researchers of status, one's competence is one's ability to contribute to the achievement of a collective goal, but defection reduces the collective payoff as well as the payoff to the individual defected against. Therefore, the idea that perceptions of a person's competence positive- ly affect that person's status and influence is unrelated

B&M's claim of incompatibility, this idea is compatible with the following: (1) The status of a person can fail to accurately reflect her com- petence; (2) characteristics of an individual other than her competence can affect her status. The general form of B&M's argument is that for an explanation to be compatible with sociolog- ical knowledge about status, the explanation must hold that the status of individuals (as well as how influential they are) accurately reflects their competence. However, sociologists have long recognized that status differences between individuals do not always reflect differences between them in competence, especially when the status difference derives from a difference in social position (e.g., a difference in gender, race, age, or class).

POPULATION SIZES

B&M claim that my model is not a viable expla- nation for cooperation in small groups; there- fore, given that the small societies of human prehistory is "where cooperation is thought to have evolved," my model's "relevance for explaining the evolution of cooperation is ambiguous" (p. 142). B&M's argument for this claim rests on two untenable assumptions and ignores what is new about the behavior of my model.

B&M conduct their own simulation analysis of my model to support their claim. They sim- ulated systems initially composed of 495 max- imally fit defectors and 5 minimally fit cooperators, and cooperation evolved in 8% of these systems. Although direct extrapolation of this simulation result to the real world yields the unlikely prediction that people cooperated in only 8% of human societies, such extrapolation requires the assumption that intersocietal con- tact did not occur and that cooperation that evolved in some societies by disproportionate prior exposure was not selected for at the society

to my model, not incompatible with it. Nevertheless, perceptions of a person's ability to acquire high pay- offs may positively affect that person's status and influence, as perceptions of a person's competence do (and as the rewards a person has received do [Cook 1975; Harrod 19801). If we adopt the termi- nology of B&M and call the ability to acquire high payoffs "competence," it is important to note the compatibility of my model with theory on status, influence, and B&M's conception of competence.

level. Simulation analysis of my model shows that if one individual from a "society" composed of cooperators (a C2) migrates to a "soci- ety" composed of defectors (D 1 s), every member of the defector society eventually becomes a cooperator in more than 20% of cases (analysis available from author on request).= The fact that early humans lived in small societies may have even strengthened the evolutionary force toward cooperation created by cultural transmission and disproportionate prior exposure. This is because a large popula- tion organized into multiple small societies is a case of a clustered social structure, which past research has shown to promote the evolution of cooperation. Contributing further to the preva- lence of cooperation would have been any selec- tion for cooperation occurring at the society level (Lenski, Nolan, and Lenski [I9701 1995;

Soltis, Boyd, and Richerson 1995). As I noted in my paper, disproportionate prior exposure complements group-selectionist explanations by offering one account for how cooperation could have evolved within a group.

B&M claim that if disproportionate prior exposure cannot account for the evolution of cooperation in the small societies of human pre-history, "its relevance for explaining the evolution of cooperation is ambiguous" (p. 142). This claim requires the untenable assumption that the only setting in which human coopera- tion evolved was those pre-historic human soci- eties. A more plausible view is that cooperation has evolved and continues to evolve in multiple contexts and at multiple times. Cooperation evolves whenever an evolutionary force toward cooperation becomes sufficiently strong andlor sufficiently unopposed by counter-veiling forces. There are many different cooperative behaviors and many different situations in which individuals might cooperate. Different types of cooperation evolve in different places, at dif- ferent times, under different circumstances.

B&M's simulation analysis shows the prob- ability that a system will absorb into the all-D repellor increases as system size decreases. Some context for considering this result is use-

On the other hand, if an individual from a "soci- ety" composed of defectors migrates to a "society" composed of cooperators, the defector will switch to cooperation after one round of interaction.

COMMENT ANDREPLY 47

ful. The behavior of a probabilistic model can be decomposed into systematic and stochastic components. I chose to analyze my model as a system of difference equations. This approach isolates the systematic behavior of a proba- bilistic model (Coleman 1964), and the sys- tematic behavior of my model is the aspect of my model's behavior that is of interest. Analysis of my model showed that under conditions of unbiased random mixing, in which individuals meet strangers in one-shot prisoner's dilemmas with no option to exit, all-C was an attractor (the only attractor) and all-D was a repellor (the only repellorti.e., there was an evolutionary force toward cooperation. All prior research on the evolution of cooperation indicated (or accepted) that, under these conditions, all-C was necessarily the only repellor and all-D was necessarily the only attractor-i.e., there would necessarily be an evolutionary force away from cooperation and toward defection. In short, my paper is about a property of cultural transmis- sion that can create an evolutionary force toward cooperation under conditions previously believed to necessarily yield an evolutionary force away from cooperation. It was important to establish this difference in systematic behav- ior between my model and previously offered models before exploring how stochastic depar- tures from my model's systematic behavior would attenuate its tendency toward cooperation in small systems.

Noting that prior research indicated (or accepted) that, under these conditions, all-C was necessarily the only repellor and all-D was necessarily the only attractor also provides con- text for B&M's main conclusion, that my "model cannot account for the evolution of cooperation except when there is a large initial population of cooperators" (p. 142). Whether we define "a large initial population of cooperators" as 5%, 50%, or 95% of the population, my model's prediction that the percent of the pop- ulation who are cooperators will increase from the initial percent under these conditions is new and was previously thought impossible.

COOPERATION NOT ROBUST TO MINOR CHANGES IN MODEL

B&M claim that my model's tendency toward cooperation is not robust to minor changes in my model and, therefore, find my model's tenden- cy toward cooperation uncompelling. It is use-

48 AMERICAN SOCIOLOGICAL REVIEW

ful to distinguish the aspects of this claim that are new and unresolved from those that are nei- ther. First, I will note what is neither new nor unresolved. Sociologists already know that for any model that predicts the evolution of coop- eration, it is possible to change that model in such a way that it does not predict the evolution of cooperation. In my paper, I showed how one change in my model (a change in the sequenc- ing of when partners play the prisoner's dilem- ma and when they compare fitness levels) eliminated my model's tendency toward coop- eration. It is unfortunate that B&M do not address my analysis of this alternative "play- compare" model, because this model has the quality that B&M found most problematically missing from the model to which I devoted most attention. That is, in my play-compare model, behavior and fitness are coupled only by the effect that a behavior has on the fitness of its performer. Thus, my paper already showed that my model's tendency toward cooperation is sensitive to the combination of mechanisms by which behavior and fitness are coupled (i.e., it is sensitive to what B&M label the "decou- pling" of behavior and fitness).

What is new and unresolved is whether the changes to which my model's tendency toward cooperation is sensitive are "minor" and whether my model's tendency toward cooperation is "uncompelling." B&M do not say what they mean by "minor changes" or how they would distinguish between changes that are minor and those that are not. Nor do they say explicitly what is important about whether a change is minor. One potential view is this: If a result is sensitive to minor changes, the result is not compelling, but if none of the changes a result is sensitive to are minor, then the result may be compelling.

In my paper, I advocated a different approach to the effects that changes in a model have on the model's implications. For any change to which a model's tendency toward cooperation is sensitive, we should consider whether that change is substantively meaningful-i.e., whether the difference between the two models corresponds to (captures an aspect of) a differ- ence between different real-world situations. If we identify substantive significance in such a change, we can use it to develop a theoretical idea and possibly a testable hypothesis about the conditions under which cooperation is likely to evolve via the mechanism captured by the model. Thus, there is nothing wrong with the alternative specifications that B&M explore. These analyses are potentially important and useful. This is the case if we use them to devel- op our understanding of conditions under which the disproportionate prior exposure inherent to cultural transmission is likely to promote coop- eration and the conditions under which it is not.

CONCLUSION

My paper called attention to a previously unrec- ognized characteristic of cultural transmission- disproportionate prior exposure-and showed how this characteristic could create a cultural evolutionary force toward cooperation. I used a formal model to develop, clarify, and illustrate my argument about cultural transmission and disproportionate prior exposure. Although B&M critique my model, they challenge neither the logical coherence nor the plausibility of my theoretical argument, which is the core contri- bution of my paper. As I argue above, B&M7s three criticisms of my model are not compelling. Nevertheless, each of these criticisms is moti- vated by a thought experiment that could poten- tially lead to a new hypothesis about conditions under which the disproportionate prior exposure inherent to cultural transmission will or will not result in the evolution of cooperation. As I emphasized in my paper, this is a task to which future research should attend.

Noah E Mark is Assistant Professor of Sociology at Stanford University. He conducts theoretical research on social and cultural evolution. Work in progress seeks to identzb conditions under which cultural transmission willpromote the evolution of coopera- tion via disproportionate prior exposure and condi- tions under which it will not. Other works in progress seek to explain the emergence ofstatus inequality and the evolution of status behavio~

REFERENCES

Axelrod, Robert. 1984. The Evolution ofCooperation.

New York: Basic Books. Berger, Joseph and Morris Zelditch, Jr. 1998.Status,

Power, and Legitimacy: Strategies & Theories.

New Brunswick: Transaction Publishers. Bienenstock, Elisa Jayne and Michael McBride.

2004. "Explication of the Cultural Transmission

Model." Comment on Mark, ASR, June 2002.

American Sociological Review 69: 138-143. Coleman, James S. 1964. Introduction to

COMMENT AND REPLY qg

Mathematical Sociology. New York: Free Press of Glencoe.

Cook, Karen S. 1975. "Expectations, Evaluations, and Equity." American Sociological Review 40:372-388.

Hamilton, William D. 1964. "The Genetical Evolution of Social Behaviour. I, 11." Journal of Theoretical Biology 7:l-52.

Harrod, Wendy Jean. 1980. "Expectations from Unequal Rewards." Social Psychology Quarterly 43:126-130.

Lenski, Gerhard, Patrick Nolan, and Jean Lenski. [I9701 1995. Human Societies: An Introduction to Macro Sociology. Reprint, New York: McGraw- Hill. Mark, Noah P. 2002. "Cultural Transmission, Disproportionate Prior Exposure, and the Evolution of Cooperation." American Sociological Review 67:32344.

Sober, Elliott and David Sloan Wilson. 1998. Unto Others: The Evolution and Psychology of Unselfish Behavior. Cambridge, Massachusetts: Harvard University Press.

Soltis, Joseph, Robert Boyd, and Peter J. Richerson. 1995. "Can Group-Functional Behaviors Evolve by Cultural Group Selection?'CurrentAnthropology 36:473-94.

Takahashi, Nobuyuki. 2000. "The Emergence of Generalized Exchange." American Journal of Sociology 105: 1105-1 134.

Trivers, Robert L. 1971. "The Evolution of Reciprocal Altruism." The Quarterly Review of Biology 46: 3 5-5 7.

Wagner, David G., Rebecca S. Ford, and Thomas W. Ford. 1986. "Can Gender Inequalities Be Reduced?" American Sociological Review 51:47-61.

Comments
  • Recommend Us