Federico Cecconi and Domenico Parisi (1998)

Journal of Artificial Societies and Social Simulation vol. 1, no. 2, <https://www.jasss.org/1/2/1.html>

Abstract

The paper introduces the concepts of individual survival strategies (ISS) and social survival strategies (SSS) and presents three sets of simulations of a particular type of SSS: the Central Store (CS) strategy, according to which the individuals in a group contribute part of their resources to a central mechanism that can redistribute these resources or make other uses of them. CS and ISS both allow a group of individuals to survive in a favourable environment although group size is slower to reach a steady state in the CS group because of the lower selective pressure on individuals' resource production. However, only CS groups survive in a less favourable environment apparently because the CS functions as a safety net for the individuals in the group. Although CS strategies can have this and other advantages over ISS, if individuals are left free to decide whether or not to give their resources to the CS, they tend not to do so. In other words, they abandon the CS strategy and revert to ISS. Because CS strategies characterize an increasing number of human societies since Neolithic times an important research problem is to identify and reproduce in the simulations, how groups of individuals that tend to act egoistically and not to give their resources to the CS, can be induced to do so.

Keywords:

Artificial Life, Social Survival Strategies, Centralized Resources

Individual and Social Survival Strategies

1.1

To survive and reproduce an individual must possess, that is, have available for use, various types of resource. A resource can be defined as anything which, when it is used by an individual, directly or indirectly causes an increase in the individual's reproductive chances. Resources include food, artefacts, mating partners, the capacity to work and fight, knowledge, money, etc. An individual has available for use its personal resources, e.g., its capacity to work and fight, and the resources it directly produces such as food or artefacts. However, if some resource of individual A is used by individual B, it is B's reproductive chances that are increased, not A's. Many resources disappear when they are used, i.e., they are consumed. Therefore, usually, if B uses some resource of A, A cannot use the resource.

1.2

A group of individuals in which each individual uses all and only its own resources is said to adopt an individual survival strategy (ISS). In a group adopting an ISS (an ISS group) the reproductive chances of each individual depend exclusively (aside from chance factors) on the resource-producing characteristics of the individual, that is, on its capacity to work and fight, on the food and artefacts the individual is able to produce, etc.

1.3

However, in some groups resources are transferred from one individual to another. A group in which there is transfer of resources among individuals is said to adopt a social survival strategy (SSS). In an SSS group the reproductive chances of an individual do not depend exclusively on the resources possessed by the individual because of its personal characteristics or because the individual has directly produced the resources. Since the resources of one individual can be used by another individual, the reproductive chances of each particular individual also depend on the other individuals in the group, i.e., on which resources the other individuals possess or are able to generate, the resources of the individual are transferred to other individuals, on which resources of other individuals are transferred to the individual, etc.

1.4

There are various sub-types of SSSs (Parisi, 1997). One kind of resource, e.g. food, may be transferred from one individual to another to allow them to survive periods of lack of food (resource sharing). Or one type of resource can be transferred from individual A to individual B and at the same time another type of resource can be transferred from individual B to individual A (resource exchange). Or some particular resource can be used only for exchange and any other resource can be exchanged for this one (money-based exchange).

1.5

In this paper we will concentrate on a kind of SSS which appears to have played a crucial role in the more recent evolution of human societies: the "Central Store" strategy. The Central Store is a mechanism according to which all the individuals in a group transfer (part of) their resources to some collective entity, the Central Store (CS). Therefore, the individuals do not use some of their own resources and the CS comes to possess the resources contributed by the group which it can use for its own goals. The resources collected by the CS can be used in three different ways: (1) they can be redistributed to the members of the group; (2) they can be transformed into collective resources, that is, resources that cannot be produced by single individuals, which are then made available for use by the group; (3) they can be used by a 'chief' that manages the Central Store to increase its own reproductive chances. (Steward, 1997, calls this chief , the 'manager'.)

1.6

In the present paper we describe a number of simulations comparing ISS groups with SSS groups adopting a CS strategy (CS groups). In an ISS group each individual produces a varying quantity of food according to its individual level of food-producing ability and it uses (eats) all and only this food to survive and reproduce. In a CS group each individual gives a portion of the food it has been able to produce to the CS. The CS periodically redistributes the collected food to the individuals in the group.

1.7

We use neural networks to model individuals and a genetic algorithm to model social evolution. Networks survive and reproduce (i.e., generate copies of themselves with some random mutations) differentially based on the quantity of food they are able to eat. In ISS groups, a network eats all the food it is able to produce. In CS groups, networks give some of the food they produce to the CS and they eat the remaining food and, possibly, the food redistributed by the CS. At the beginning of evolution networks are assigned random connection weights and therefore they are not very good at producing food. However, the selective reproduction of the ablest individuals and the addition of random mutations (which can occasionally result in offspring that are better than their parents) tend to cause an evolutionary increase in food producing ability. We study not only how this ability evolves but also the evolutionary changes in group size; more particularly, we are interested in what conditions determine group survival or extinction.

1.8

We present three sets of simulations. In the first set, we show that if ISS and CS groups live (separately) in a favourable environment, i.e., an environment where plenty of food can be produced, both types of groups succeed in surviving and both eventually reach a comparable stable state in terms of group size and average ability to produce food. However, due to the lower selective pressure on CS groups, CS groups take more time to reach the final stable state than ISS groups.

1.9

In a second set of simulations we show some advantages of the CS strategy over the ISS. In these simulations the environment worsens after a group has reached a stable state in the more favourable environment. In the new environmental conditions less food can be produced. CS groups tend to survive although their group size decreases due to the lower carrying capacity of the environment. However, the ISS groups become extinct. Furthermore, if both a CS group and an ISS group live together in the same unfavourable environment and compete for the same resources, after a while the ISS group becomes extinct, leaving just the CS group.

1.10

In both the first and the second set of simulations the individuals in the CS group are required to give a prescribed portion of their food to the CS. Furthermore, there is no advantage in not looking for food because energy is decreased by some amount in each cycle independently of what the individual does. In a third set of simulations we ask what would happen if our individuals are left free to decide (in an evolutionary sense) to give or not to give their food to the CS or, even more radically, to produce or not to produce food, given the fact that an individual could try to live on the food redistributed by the CS without producing any food of its own. In these new scenarios an increasing number of individuals tend to choose, respectively, the behavior of not giving their food to the CS and the behavior of not producing food, relying entirely on the food redistributed by the CS. In both cases the CS soon becomes empty, the CS strategy is abandoned and the group reverts to an ISS. If the groups happen to live in a favourable environment, they survive as ISS groups. However, if they live in an unfavourable environment they become extinct. Thus, although adopting a CS strategy allows a group to survive in an unfavourable environment, the egotistical tendencies of the individuals cause the abandonment of the CS strategy and, consequently, the extinction of the group.

1.11

The paper is organized in the following way. In Section 2 we present our model of ISS and CS groups. In Section 3 we describe the three sets of simulations. In Section 4 we discuss the results of the simulations in a broader context and we point out to further research currently in progress.

A Model of ISS and CS Groups

2.1

Imagine a group of individuals living in an environment that contains randomly distributed food elements. The nervous system of each individual is simulated by a neural network. The input units of the network encode sensory information specifying the location of the single food element currently nearest to the individual. The output units encode motor actions with which the individual responds to the sensory input. The motor actions allow the individual to turn or to displace itself in the environment. Hence, at each time cycle an individual perceives the position of a single food element and it responds with some motor action. The other individuals are not perceived.

2.2

Each individual is given some 'energy' at birth but the energy is decreased by some small quantity in each cycle. When the energy reaches the zero level, the individual dies. There is only one way for an individual to survive: to eat. When an individual happens to step on a food element, the food element disappears and it is considered as captured by the individual. (This looks more like food gathering rather than food production in the sense of agriculture but for our purposes they are equivalent.) The individual eats the food element it has captured and, as a consequence, the energy of the individual is increased by some quantity. To survive, therefore, an individual must be able to respond to sensory information about the position of the nearest food element by generating motor actions that allow it to reach the food. If it is able to eat and survive, an individual can have offspring. Offspring are generated periodically at regular intervals. When an individual generates an offspring, it gives some of its current energy to the offspring.

2.3

Which output (motor action) an individual will generate at any given time in response to a particular input depends on the connection weights of its neural network. The networks that constitute the initial group are assigned random connection weights. Hence, each individual will tend to react to sensory input differently from all the other individuals but the behavior of most individuals will be haphazard or stereotypical because of the random connection weights. The individuals will not eat much and many will die without leaving offspring. As a consequence, the size of the group will tend to shrink and, if the mortality rate is greater than the birth rate, the entire group will become extinct. However, for purely chance reasons some individuals will have better connection weights than other individuals. These individuals will eat more and will have some offspring. Since the offspring inherit the connection weights of their parents, selective reproduction will tend to increase the average eating ability of the group and, consequently, group size. (We are assuming that the quantity of food present in the environment is not a limiting factor at this stage. Food is periodically (seasonally) reintroduced in the environment.) Hence, if the group succeeds in overcoming the initial period in which there is risk of extinction, the group can reach a stable group size which depends on the carrying capacity of the environment.

2.4

Reproduction is nonsexual: a single parent can generate one or more offspring. Offspring do not inherit exactly the connection weights of their parents since some of the weights are randomly changed at reproduction (genetic mutations). While in most cases these changes will result in offspring with a less good food capturing ability than their parents, in some cases an offspring will be better at capturing food than its parent - and these individuals will be more likely to reproduce than their less lucky siblings. Hence, genetic mutations are a second mechanism that will lead to an evolutionary increase in group's size and average ability. Selective reproduction and the constant addition of variability through genetic mutations will cause the evolutionary emergence of a good average level of food capturing ability in the group although different individuals will always tend to have different levels of ability and will reproduce differentially.

2.5

Notice that in this model the reproductive chances of an individual depend exclusively on the food capturing ability of that particular individual, i.e., on the goodness of the connection weights the individual has inherited from its parent (there is no learning during life) and, of course, on its being lucky enough not to end up in a local environment with many other individuals and, therefore, little food. There is no transfer of resources in the sense that an individual does not eat the food captured by another individual. Therefore, this is an ISS group (Figure 1).

Figure 1. A group of individuals adopting an individual survival strategy.

2.6

Imagine now a second group identical to the first except for a single feature. In the first group when an individual happens to step on a food element, the individual automatically eats the food and its energy level is increased. In the new group, when an individual captures a food element, the individual eats only some part of the food. The remaining part is contributed to a CS.

2.7

What happens to the food contained in the CS? The food contained in the CS is redistributed to the entire population. The redistribution scheme can take different forms. The food can be periodically and automatically distributed to all individuals in the population. Or when an individual reaches the zero level of energy and is about to die, the individual can eat the food contained in the CS and survive. Or some other scheme can be adopted. In all cases the CS functions as a safety net for all the members of the group (Figure 2). (Consider, however, that the CS might have other functions. The CS food could be redistributed to the 'richest' individuals rather than to the 'poorest' ones, thereby increasing the selective pressure on the food capturing ability.)

Figure 2. A group adopting a Central Store strategy. The individuals contribute (some of) the food they are able to capture in the environment to a Central Store and the food contained in the Central Store is redistributed to the entire group.

2.8

While in an ISS group the reproductive chances of an individual depend exclusively on the characteristics of that individual (in our case, on the individual's ability to capture food), in an CS group they also depend on the CS and, therefore, on the behavior of the other individuals contributing to the CS. If the CS is not empty, an individual can eat the food contained in the CS to survive and, possibly, reproduce. If the CS is empty, this is not possible. But the state of the CS, its being full or empty, is a function of the entire group, i.e., of the ability and, as we will see, also the readiness of all individuals in the population to look for food in the environment and to contribute the food they capture to the CS. Therefore, in a CS group the reproductive chances of an individual depend not only on the characteristics of the individual but also on the characteristics of the entire group. In the extreme case in which all individuals give all the food they are able to capture to the CS and they take the food from the CS when they need it to survive, the reproductive chances of each individual depend (almost) completely on the entire group and the characteristics of the particular individual (i.e., the individual's ability to capture food) become unimportant, especially if the group's size is large and the individual's contribution to the CS is marginal.

2.9

As the survival and reproductive chances of any particular individual become less and less dependent on its ability, because it can obtain food from the CS independently of its ability to capture food, the evolutionary pressure on individuals is weakened. In an ISS group the individuals that reproduce are the individuals that are most able to capture food (aside from fortuitous factors). Hence, evolution will tend to select for and to maintain in the population a high level of food capturing ability. In a CS group the correlation between the food capturing ability of an individual and the individual's reproductive chances is reduced. In the extreme case, when all the food captured is contributed to the CS and each individual takes all the food it needs from the CS, the correlation becomes zero and individuals are selected for reproduction on a random basis (neutral selection).

Simulations

A Comparison between ISS and CS Groups

3.1

We have done some simulations comparing ISS and CS groups. An ISS group and a CS group live in separate but identical environments. The initial group size is 150 individuals in both cases. The environment is a grid of 100x100 cells. In the initial environment, 50 per cent of the cells, randomly selected, contain a food element. Food is periodically reintroduced every 40 cycles (one "year") by filling 80 per cent of the cells with food. Each individual is modelled by a feedforward neural network with two input units, two output units and five hidden units. The two input units encode the angle and distance of the nearest food element. The two output units encode four possible motor actions: go one cell forward in the facing direction, turn 90 degrees right or left or do nothing.

3.2

Each individual in the initial group is born with 30 energy units. The energy is decreased by half a unit in every cycle. When an individual of the ISS group captures (eats) a food element, its energy is increased by 50 units. In the CS group, however, the individual gives almost three quarters of the food element's energy (36 energy units) to the CS and eats the remaining 14 units. The energy contained in the CS is redistributed to the entire group by giving 2 energy units to 10 per cent of the population, randomly selected, in each cycle.

3.3

When an individual's energy becomes zero, the individual dies. Provided that it is able to survive, an individual generates one offspring every 50 cycles. The offspring is placed in the environment in a randomly selected position near its parent and the parent gives half its current energy to the offspring. All the individuals die when they have reached a maximum age of 350 cycles. The simulation lasts 20,000 cycles.

3.4

The results¹ of the simulations are shown in Figures 3 and 4.

Figure 3. Evolutionary change in group size in ISS and CS groups across 20,000 cycles.

3.5

Figure 3 shows how group size changes in the two groups across 20,000 cycles. After the inevitable initial drop due to the low initial food capturing ability, group size starts to increase until it reaches a steady state of about 500 individuals at around cycle 14,000 for the ISS group and at around cycle 17,000 for the CS group. The greater selective pressure acting on the ISS groups is reflected in the more rapid increase in group size for this group compared with the CS group.

3.6

We have also run another set of simulations in which all individuals give all the food they are able to capture to the CS and all individuals live entirely on the food redistributed by the CS. In these simulations there is no selective pressure to evolve food capturing ability because there is no correspondence between an individual's ability to capture food and the individual's reproductive chances. Therefore, food capturing ability simply does not evolve and the group quickly becomes extinct.

3.7

However, in the present simulations an individual eats at least some of the food it is able to capture and, therefore, there is some correspondence between an individual's ability and its reproductive chances. This correspondence creates a sufficient selective pressure on the evolution of food capturing ability in all the 25 replications to save the CS group from extinction.

3.8

The effect of the difference in selective pressure acting on the behavior which is evolving in the two groups can be directly observed by measuring the food capturing ability of the individuals of the two groups in the artificial but identical conditions of a simulated 'laboratory'. We extracted a sample of individuals in each cycle from each of the two groups and tested them by putting each individual alone in a standard environment. Figure 4 shows the number of food elements captured in a fixed time period by the individuals of the ISS groups and by the individuals of the CS group. The food capturing ability of the ISS individuals increases more rapidly than the ability of the CS individuals. This explains the greater early increase in group size of the ISS group compared with the CS group (Figure 3). However, the difference in food capturing ability between the two groups disappears at around cycle 17,000, in correspondence to the reaching of an identical group size by the two groups.

Figure 4. Evolutionary change in food capturing ability in ISS and CS groups. The food capturing ability of individuals belonging to the two groups is tested in identical 'laboratory' conditions

3.9

We conclude from this initial set of simulations that, if the food contributed to the CS is redistributed equally to all individuals, a CS strategy implies a less strong selective pressure on the individuals of a group adopting this strategy than with an ISS. This causes a slower evolutionary increase in group size (after the initial drop) and in the food capturing ability of the CS individuals. However, after the CS strategy has had a chance to evolve, it results in the same group size and the same average ability as the ISS.

Advantages of the CS Strategy

3.10

What are the consequences of adopting a CS strategy? Adopting a CS strategy appears to have a number of advantages with respect to an ISS.

3.11

Imagine that a CS group and an ISS group live in two identical but separate environments and have reached a stable state, as did the two groups described in the preceding Section. Then, however, the environment changes for some reason. Less food is periodically reintroduced in the environment at the beginning of each new "year". While in the past a group was able to survive the lean periods which occur just prior to food reintroduction, now the group can find itself with very little food in these periods and risks becoming extinct. In the changed environment there may be periods in which group size becomes exceedingly small. If group size becomes too small, chance factors may cause the complete extinction of the group in both real (cf. Raup, 1991) and simulated (Cecconi and Parisi, 1994) groups.

3.12

An ISS group cannot do much to avoid extinction. If the environmental changes are sufficiently large and nasty, an ISS group is very likely to become extinct. In contrast, a CS group can protect itself from the negative effects of environmental change. During the longer and harsher 'winters' individuals can survive by eating the food stored in the CS.²

3.13

We have run a new set of simulations in which ISS groups and CS groups reach a stable population size of around 500 individuals in separate environments in which food is periodically reintroduced in 80 per cent of the cells, exactly as in our preceding simulations. Then, suddenly the environment changes and food is reintroduced into only 30 per cent of the cells. The results in terms of population size are shown in Figure 5.

Figure 5. Change in group size in ISS and CS groups after a worsening of environmental conditions. The ISS groups all become extinct after 6,000 cycles (average of 25 runs of the simulation) while all the CS groups succeed in avoiding extinction and in reaching a new stable state corresponding to the carrying capacity of the new environment

3.14

The ISS groups appear to be helpless in the new unfavourable environment and they all become extinct after an average of 6,000 cycles. In contrast, the CS strategy allows a group to avoid extinction. Group size drops sharply from the initial value of 500 individuals to around 150 individuals, corresponding to the carrying capacity of the new environment.

3.15

We conclude that CS groups may be more able to survive adverse environmental changes than ISS groups. Evidently, the CS functions as a safety net both for individuals that in the new more adverse environment are at risk of dying because of their limited skill in capturing food and for individuals that by chance find themselves in a local environment containing many other individuals and, as a consequence, little food. This safety net does not exist for the ISS groups which are therefore unable to tide over the worsening of environmental conditions.

3.16

Another advantage of the CS strategy over the ISS emerges if the two strategies happen to compete directly with one another. In the simulations described so far ISS and CS groups live separately in distinct environments. What would be the outcome if an ISS group and an CS group share the same environment and compete for the same resources?

3.17

We answered this question by first evolving an ISS group and a CS group in two separate environments with favourable conditions (food reintroduced in 80 per cent of the cells) and then putting the two groups together in the same environment.

3.18

The results vary as a function of the environmental conditions. If the environmental conditions remain favourable all sorts of results can be obtained in the different runs of the simulation. In some runs both groups survive or both groups become extinct. In other runs either the ISS group or the CS group remains as the only surviving group while the other group becomes extinct. However, if the two groups compete in a less favourable environment (food reintroduced in 30 per cent of the cells) the invariable result is that the CS group survives while the ISS group disappears.

3.19

We can summarize this Section by saying that a group adopting the CS strategy is able to survive in less favourable environments in which ISS groups become extinct. In Section 4 we will see that there are other possible advantages that accrue to CS groups and are not available to ISS groups. However, these advantages notwithstanding, CS strategies are delicate adaptations that are constantly exposed to the risk of being replaced by individual strategies or, even more seriously, of leading to the extinction of the group. In the next Section we will see what these risks are.

Cheating

3.20

In the simulations described so far an individual belonging to a CS group has no choice whether to eat the food it is able to capture in the environment or to renounce eating the food and contribute it to the CS. The quantity of food that can be directly eaten and the quantity of food that must be contributed to the collective store have been decided once for all by the researcher. Similarly, the individual has no choice about whether to look actively for food - thereby spending its energy - or to omit to look for food and simply eat the food contained in the CS. Energy is automatically decreased in each cycle whatever the individual does and, therefore, there is no advantage for the individual in omitting to look for food.

3.21

But imagine, more realistically, that an individual is free to decide how much of the food which it is able to capture it will eat and how much it will contribute to the CS. Or that the individual has the option of avoiding spending its energy looking for food. In these circumstances, social cheaters or free riders are likely to appear. Social cheating can take two forms. The first type is an individual that tends to eat most or all of the food it is able to capture and, therefore, tends to contribute little or no food to the CS. The second type is an individual that omits to look for food and relies on the CS for eating. If the first type of social cheating becomes widespread in the group, the group is likely to abandon the CS strategy and to revert to an ISS. If the second type of social cheating becomes widespread, the group can simply become extinct.

Receiving from and not Contributing to the CS

3.22

In the simulations described in the preceding Sections the percentage of food each individual contributes to the CS was fixed and identical for all individuals. Each individual gave about three quarters of the food to the CS and ate the remaining quarter. But imagine the situation is different. Individuals vary in their tendency to eat the food they are able to capture and in their contribution to the CS. Some individuals tend to eat most or even all the food whereas other individuals contribute most of their food to the CS. The tendency is genetically inherited. An offspring inherits from its parent, as well as the connection weights that control the individual's behavior of looking for food, a 'social gene' that specifies in a quantitative way its tendency to eat its food (low value of the social gene) or to contribute the food to the CS (high value of the social gene). Of course, mutations apply to this social gene. Because of mutations an offspring can inherit a quantitative value for its social gene that can be slightly greater or smaller than the quantitative value of its parent's gene. This implies that evolution can select for individuals that have particular values for their social genes.

3.23

Individuals that inherit low values for their social gene, i.e., that tend to eat most of their food, are more likely to reproduce than individuals that inherit high values for the gene. The former individuals will increase their energy with both the food they capture and with the food they receive from the CS. Hence, they will have more offspring than the more socially conscious individuals that deprive themselves of the food they are able to capture and give it to the CS. The results of our simulations generally confirm this prediction. However, the actual outcomes of the simulations depend on various factors.

3.24

The value of the social gene can vary from 0 to 1. A value of zero means that the individual eats all the food it is able to capture. A value of 1 means that the individual contributes all the expected food to the CS, i.e., three quarters of the food it succeeds in capturing. Intermediate values between 0 and 1 mean that the individual contributes the corresponding quantity to the CS.

3.25

In a first set of simulations the group evolves its food capturing ability in a favourable environment (food reintroduced in 80 per cent of the cells). The group starts either as an ISS group (all individuals have a zero value for their social gene) or as a CS group (all individuals have a value of 1 for their social gene).

3.26

The results are the following. If the group starts as an ISS group, the ISS is never abandoned. The average value of the gene never increases above 0.3 (not zero because of random mutations). The ISS group is able to evolve an appropriate level of food capturing ability and to survive. But a CS strategy never evolves.

3.27

If the group starts as a CS group, the outcome is different. As we saw in Section 2, a CS strategy implies a reduced selective pressure for evolving the ability to capture food and, as a consequence, this ability increases more slowly than in ISS groups. This causes the CS not to contain much food initially. CS groups with a hardwired tendency to contribute to the CS can tide over this initial difficult phase because all individuals are contributing three quarters of their food to the CS. Hence, an evolving CS group without the social gene can survive in a favourable environment (cf. Section 2).

3.28

The situation is different if the CS group has a social gene, i.e., the individuals in the group are (evolutionarily) free to decide whether to give or not to give their food to the CS. Individuals with lower values of their social gene tend to colonize the group because, as already remarked, these individuals eat more of their own food while receiving the same amount of food from the CS as the individuals with a higher gene value. But these more egotistical individuals contribute less to the CS. As a consequence, the CS contains still less food in the difficult initial phase and the individuals die before the group becomes a completely ISS group.

3.29

In a second set of simulations we first evolved either an ISS group or a CS group without the social gene in a favourable environment until the group stabilized. Then the environment worsens and the individuals in both groups are left free to decide (in an evolutionary sense) whether to act egoistically or socially by adding the social gene to their genotype. In other words, unlike the preceding simulations, the ability to capture food has already evolved when the individuals are given the freedom to decide to act egoistically or socially.

3.30

In these circumstances both ISS and CS groups become extinct. As we know from the simulations reported in Section 3.2, in an unfavourable environment only groups that adopt a CS strategy are able to survive. In the present simulations when a group becomes free to decide what strategy to adopt, the groups that are initially ISS groups (zero value of the social gene) tend to preserve their ISS and become extinct because of the unfavourable environment. The groups that are initially CS groups (gene value of 1) tend to abandon their CS strategy and become ISS groups. As a consequence, they also become extinct.

Omitting to look for Food

3.31

Let us turn to the second type of social cheating, which consists in omitting to look for food and becoming entirely dependent on the CS. To allow for the possible emergence of this type of social cheating we proceed in the following way. In the preceding simulations all four possible actions (go one cell forward, turn right or left, do nothing) involved a cost in energy. In the new simulations the option 'do nothing' has no energy cost. To find food an individual has to move forward and to turn (first three options). These actions decrease its energy. If the individual decides (because of its inherited connection weights) to respond to the sensory input by doing nothing (fourth option) its energy is not decreased but of course the individual will never find any food and therefore it will never contribute any food to the CS. The individual will survive by eating the food contributed to the CS by the non-cheating members of the group.

3.32

To measure the evolutionary history of individuals which could omit looking for food, we found the percentage of 'do nothing' actions in each cycle in the entire group across 20,000 cycles. We did this separately for fixed ISS and CS groups (i.e., groups lacking the social gene) that evolve in either a favourable or an unfavourable environment. The results are shown in Figure 6.

Figure 6. Percentage of 'do nothing' actions in each cycle across 20,000 cycles of evolution for ISS and CS groups living in either favourable or unfavourable environments. All groups become extinct except the ISS groups living in the favourable environment, which succeed in keeping the percentage of 'do nothing' actions at very low levels.

3.33

To understand the results of Figure 6 one must consider that in ISS groups the option 'do nothing' is not advantageous for the individual because in these groups, which lack a CS, the individual must necessarily rely on its own food to survive. Therefore, individuals that tend to do nothing capture and eat little food and, therefore, leave few offspring and their lineage is eliminated from the population. This explains why ISS groups keep the percentage of 'do nothing' actions at very low levels and succeed in surviving in the favourable environment. However, as we already know (cf. Section 3.2), ISS groups are unable to survive in the unfavourable environment.

3.34

In contrast, the behavior of doing nothing has the possibility of becoming widespread in CS groups because individuals that do nothing can survive by eating the collective food. Individuals that adopt this behavioral strategy will be favoured in the reproductive race since they do not incur the costs of looking for food that affect negatively the reproductive chances of the other individuals. Therefore, in the long run the group will tend to be colonised by these individuals.

3.35

Figure 6 shows that the selective pressure on avoiding spending energy by actively looking for food causes 70 per cent of all actions to become 'do nothing' actions after a certain number of evolutionary cycles in both the CS groups living in the favourable environment and the CS groups living in the unfavourable environment. However, this turns out to be suicidal behavior in the long run at the group level. Since fewer and fewer individuals are contributing to the CS because they simply do nothing and therefore have no food to contribute, the CS tends to be empty most of the time and it ceases to function as a safety net for the group. Individuals cannot eat their own food because they do nothing and therefore do not capture any food, but they also cannot eat the non-existent collective food. Hence, the group becomes extinct. Predictably, extinction is quicker for the CS groups living in the less favourable environment than for the CS groups living in the more favourable environment (cf. Figure 6).

Discussion

4.1

Social strategies that consist in collecting the resources produced by the individuals in a group in a CS and redistributing the collected resources to the group appear to have a number of advantages over individual strategies in which each individual uses all and only its own resources. In our simulations we have shown that CS groups can survive in less favourable environments in which ISS groups are unable to survive and they win the competition with ISS groups in these environments. There may be many other possibilities that are open to CS groups but not to ISS groups. For example, CS groups can colonize new environments that are more adverse than the environment currently inhabited by the group. This may be impossible for ISS groups.

4.2

Another possible advantage of the CS strategy is related to the division of labour. In the simulations described in this paper all individuals exhibit the same type of behavior: capturing the food which is present in the environment. There is no specialization or division of labour on the basis of which some individuals do A and other individuals do B. But imagine a group in which individuals can exhibit two different types of behavior. One behavior gives access to some resource that can be directly used to increase the individual's reproductive chances, e.g. food. The other type of behavior creates some new resource that can be indirectly but not directly used to increase an individual's reproductive chances, e.g., an artefact (such as a pot). If both types of behaviors are present in a group, the total amount of resources available to the group is greater than if only the first type of behavior is present since a pot can increase the usefulness or reproductive value of food by making it possible to preserve or cook it. However, no individual can specialise in producing artefacts unless there is a CS that provides the individual with food. (We are ignoring resource exchange as another possible social survival strategy that might support division of labour. On the relationship between resource exchange and the CS, see below.) In simulations described in Pedone and Parisi (1997) it was shown that behaviors that are not immediately remunerative for the individuals exhibiting them do not emerge and are not maintained in a group even if they would increase the total amount of resources available to the group, unless the group is a small group of genetically related individuals and the results can be explained in terms of kin selection theory (Hamilton, 1964; Trivers, 1985). Hence, another advantage of the CS strategy might be to facilitate the emergence of a division of labour in large groups of genetically unrelated individuals. If a division of labour implies activities that are not immediately remunerative for the individual displaying these activities, it may never evolve in an ISS group. However, it can evolve and stabilize in CS groups because CS resources can be used to reward the individuals exhibiting the not immediately remunerative activities.

4.3

Other advantages of the CS strategy that may help explain its crucial role in the recent evolution of human societies derive from the second type of use that can be made of the resources contributed to the CS by the individuals forming a group (cf. Section 1). The resources contributed to the CS can be redistributed to the group in their "raw" form or they can be transformed by the CS into a new kind of resource, that we have called collective resources. A collective resource is a resource (that is, something that can increase the reproductive chances of an individual using the resource) which cannot be produced by a single individual but only through the organized effort of many individuals. Examples of collective resources are a defensive wall, an irrigation system, the capacity to wage aggressive and defensive wars. These are resources that are typically produced by a CS using the individual resources contributed to the CS by the members of the group such as an individual's capacity to work or fight. Collective resources may greatly increase the reproductive chances of the individuals that use the resources.

4.4

The role of the CS in generating otherwise non-existent collective resources can be crucial in connection with another social survival strategy: resource exchange. Resource exchange consists of allowing another individual to use one's resources while expecting that the other individual will reciprocate by giving some other resource to the first individual. Resource exchange by itself is independent of a CS and therefore it can in principle exist in groups without a CS. However, resource exchange is often accompanied by behaviors that threaten to disrupt it: individuals can fail to reciprocate appropriately in an exchange. Resource exchange tends to be analyzed with theoretical models such as game theory that concentrate on the two parties involved in the exchange (Axelrod, 1984). Any role of a third party that would guarantee that the appropriate reciprocal behaviors are exhibited by the two partners tends to be ignored. In opposition to this, we believe that if we wish to understand the role played by resource exchange in the evolution of human societies, the study of resource exchange and its theoretical interpretation must include the role of this third party. This third party is the CS mechanism. One of the most important collective resources generated by the CS is the organization of resource exchange and, especially, the guarantee that individuals will behave appropriately in resource exchange.

4.5

We have seen that the CS strategy has many important advantages over an ISS. These advantages can perhaps explain why so many human societies have moved from ISS to CS since Neolithic times. In fact, CS can be a cause and at the same time an effect of the economic intensification and the production of resource surplus that increasingly characterize humans societies in the last 10-12 millennia. However, our simulations have also shown that when the individuals in a group are free to decide whether or not to give their resources to the CS, they tend not to or even not to produce resources at all. This means that either a CS strategy will never evolve in an ISS group or that the strategy, after it has emerged, is quickly abandoned and the group reverts to an ISS. This creates an apparent paradox. CS strategies appear to be very useful but they tend to be discarded by egoistically oriented individuals. On the other, the recent history of human societies clearly shows that these strategies have emerged and have been maintained in human groups. Therefore, the next important research question becomes: How have human groups been able to adopt CS strategies notwithstanding the egotistical tendencies of their members that are in opposition to these strategies?

4.6

All the advantages of CS strategies appear to be advantages for the group, not directly advantages for the individual. Group advantages are problematic for the individual because they only translate into advantages for the individual if the other individuals behave in the prescribed way - which is not always the case. Perhaps we could observe the emergence and preservation of CS strategies in our simulations if we could create the conditions for competition among groups and for group selection (Wilson and Sober, 1994). But biological group selection remains problematic, first, because the conditions that would allow group selection to operate are not frequently realized and, second, because the tension within each group between the interests of the group and the interests of the individuals remains.

4.7

The problem of explaining the fact of the emergence and maintenance of CS strategies in human societies therefore remains unsolved, although many interesting hypotheses can be and have been advanced. Given our simulation approach, to explain the emergence and maintenance of CS strategies means to run a simulation with a group of individuals that are free to decide whether to give or not to give their resources to the CS and to be able to observe the emergence of a CS strategy from an ISS and the maintenance of the CS strategy over time.

4.8

We would like to conclude this paper by indicating a particular direction that our research on the CS strategy is taking in some simulations currently in progress. The new simulations are based on the idea that the resources contributed to the CS by the members of a group can not only be used for redistribution to the group or for transformation into collective resources for use by the individuals in the group, but they can also be used by a 'chief' (an individual or a structure composed of a few individuals) to further its own goals and reproductive chances (cf. Section 1). This 'chief' is rewarded with (part of) the resources contributed to the CS if the 'chief' is able to guarantee, using all possible means, that the individuals in the group do actually contribute their resources to the CS. In other words, the 'chief' is an individual or set of individuals whose reproductive chances tend to co-vary directly with the continuing existence and prosperity of the CS (Steward, 1997). 'Chiefs' tend to be responsible not only for guaranteeing the appropriate behaviors with respect to the CS on the part of the group's individuals but also for deciding the rules of the CS that specify who must give to the CS, what and how much, who must receive from the CS, what and how much, and for organizing the transformation of individual resources into collective resources. Although 'chiefs' may not be the only explanation for the emergence and maintenance of the CS, the emergence of 'chiefs' (Earle, 1997) appears to be critically linked to the emergence of CS strategies.

Notes

¹

The results of all the simulations described in this paper are the average of 25 different replications (runs) of the same simulation each with a different 'seed' for generating the initial sets of random connection weights, the spatial distribution of food, the genetic mutations, etc.)

²

One might think that even individual food storage could solve this problem, with no need for a central store. Individuals might store some of the food they are able to produce in individual stores and consume this food in case of necessity. Parisi (1997) describes some simulations comparing groups with individual and central stores. For the role of food storage in hunting societies, see Ingold, 1983, Testart, 1982. For the role of food strategies in primitive economies, see Colson, 1980.)

Acknowledgements

The paper has benefited from discussion with the participants to the cultural modelling month in May 1995 at the Santa Fe Institute, Santa Fe, New Mexico.

References

AXELROD, R. The Evolution of Cooperation. New York, Basic Books, 1984.

CECCONI, F. and PARISI, D. Learning during reproductive immaturity in evolving populations of neural networks. Institute of Psychology, CNR, 1994.

COLSON, E. In good years and bad: food strategies of self-reliant societies. Journal of Anthropological Research, 1980, 35, 1-29.

EARLE, T.K. How Chiefs Come to Power. The Political Economy in Prehistory. Stanford, Stanford University Press, 1997.

HAMILTON, W.D. The genetical evolution of social behavior. Journal of Theoretical Biology, 1964, 7, 1-16.

INGOLD, T. The significance of storage in hunting societies. Man, 1983, 18, 553-571.

PARISI, D. What to do with a surplus. In R. Conte, P. Terna, & R. Hegselmann (eds.) Simulating Social Phenomena. Berlin, Springer, 1997.

PEDONE, R. and PARISI, D. In what kinds of social groups can "altruistic" behaviors evolve? In R. Conte, P. Terna, & R. Hegselmann (eds.) Simulating Social Phenomena. Berlin, Springer, 1997.

RAUP, D.M. Extinction. New York, Norton, 1991.

STEWARD, J. Evolutionary Transitions and Artificial Life. Artificial Life, 1997, 3, 101-120

TESTART, A. The significance of food storage among hunter-gatherers: residence patterns, population densities, and social inequalities. Current Anthropology, 1982, 25, 523-537.

TRIVERS, R. Social Evolution. Menlo Park, CA, Benjamin/Cummings, 1985.

WILSON, D.S. and SOBER, E. Reintroducing group selection to the human behavioral sciences. Behavioral and Brain Science, 1994, 17, 585-608.

Return to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, 1998