printed by Tolerance and Sexual Attraction in Despotic Societies: A Replication and Analysis of Hemelrijk (2002) Hagen Lehmann, JingJing Wang & Joanna J. Bryson University of Bath, Artificial models of natural Intelligence Most primate societies are characterised by hierarchical structures with more or less despotic value. Males are usually dominant over females, but in periods of sexual attraction male “tolerance” towards females rises. This has been explained as a probable cognitive strategy (Yerkes 1939) — an exchange for copulation, which is adaptive in that it also therefore produces offspring (Goodall 1986, deWaal 1989, Stanford 1996). Tolerance is considered one of the most basic forms of conflict resolution. It might be difficult to see tolerance as an action to be selected, since it seems like a form of inaction. However, if an individual is very inclined to preserve resources (including its own social rank), then expressing tolerance can require considerable inhibition of strong inclinations. Hemelrijk showed in a model called “DomWorld” that the “tolerance” of males towards females is created as a side effect due to the rise of female dominance during periods of sexual attraction and that this is the consequence of more frequent approaches of males towards females during these periods (Hemelrijk 2002). In her model males gain no benefit from “tolerating” females and they only do so at high aggression levels as what she calls “respectful timidity”, because some of the females have become dominant over them. Our study is an attempt to replicate Hemelrijk's results concerning the effects of sexual attraction and intensity of aggression between sexes in both despotic and egalitarian primate societies. We used the description of her individual-based model, DomWorld, to create our own version which simulates an artificial primate group. We did so in order to gain a deeper understanding of the structure of her model. We hope, this will enable us to create our own simulation of group dynamics in primate societies in the future. BACKGROUNDPURPOSE METHODS RESULTS CONCLUSIONS REFERENCES Our replication of Hemelrijk was successful. Our results show as hers did that in groups with a high level of aggression females are able to gain higher positions in the social hierarchy and eventually take over the group. Sexual attraction amplifies this effect. Agent-based models of social interactions can be copious sources of hypotheses and theories, since they allow for constructive modeling followed by relatively simple testing. They show how simple behavior at an individual level produces complex behavioral patterns on a social level. However, they are only valid if correlates can be found in real animals for the mechanisms that produce the model’s results. Our replication leads to the following questions: 1. Can two factors, the intensity of aggression and sexual attraction, really account for the change in female dominance at the time of tumescence? If the model is accurate, we should see increased conflicts involving the females at this time, and if the conflicts are violent, we should see the females immediately take a high dominance position if they happen to win. 2. Is the confounding of aggression with increase in the dominance value admissible or not? If the model holds, then changes of dominance rank should require fewer conflicts in despotic species than in egalitarian ones Self-confidence after winning a fight may increase, if one wins against a much stronger opponent. But in a real fight the body size and strength is at least as important as the psychological status of the individual. If the empirical data does not support Hemelrijk’s model in the above two questions, then we will have sufficient cause to look at more complex models, possibly including enough cognitive state for reciprocation. The number of virtual female interactions increases significantly in conditions with sexual attraction at both intensities of aggression (Mann-Whitney, N = 10, U = 0, p < tailed, Mann-Whitney, N = 10, U = 0, p < tailed). The aggression level amplifies the result, even though the effect for the aggression is rather weak (Mann- Whitney U-Test, N = 10, U = 24 p <.05 2-tailed). Figure 1: Number of female interactions during different conditions Figure 2 shows the dominance of virtual females as the summed number of males ranking below each female at different times in different conditions. We can see that the values in conditions with high aggression level increase over the time, but that they stay constant in conditions with a low aggression level. Figure 3 shows the distribution of the coefficient of variation of dominance values for both sexes. If the aggression is high, there will be a steeper hierarchy --- the difference between rank values will be larger. This is true for both within and between sexes. Attraction amplifies this result, despite the fact that some females may outrank some males in this condition. Figure 6: Change of dominance values at a low level of aggression (blue = male, red = female). Figure 5: Change of dominance values at a high level of aggression (blue = male, red = female). Figure 3: Distribution of coefficients of variation of dominance value for both sexes. The last two figures show the change of dominance values for both sexes in conditions with high and with low levels of aggression. With high aggression a constant change in the dominance structure is noticeable and therefore a more and more differentiated hierarchy. With low aggression there is only very little change in the dominance values. This creates a very stable hierarchy with no chances for females to gain a higher position in the group then any male. Figure 2: Female dominance as the summed number of males ranking below each female. Our simulation was based on the model described by Hemelrijk (2002). We used NetLogo 2.1, a purpose-built modelling tool that provides a relatively easy, high-level language for quickly constructing models and visualising results. As described by Hemelrijk the world is a continuous space of 200 x 200 units. Agents have real-valued locations and can move in any of 360 directions. Each agent has certain individual attributes, and a shared set of interaction rules that determine social behavior. The dominance value of each agent determines its hierarchical status and is calculated with the with a formula based on individual rank (DomValue): StepDom (the intensity of aggression) and produces w, the victory or loss value. It is calculated as follows: The relative dominance value is compared with a randomly drawn number between 0 and 1. If it is greater then the drawn number, the agent wins. That means, as higher rank of the agent as more likely it is to win. For the experiment we used the parameter settings Hemelrijk used in different studies. We set up 8 agents, four of each sex (N = 8). Each agent had an personal space of 2, a vision angle of 120 degrees, a maximum perception range of 50 units and a near- perception range of 24 units. The search angle was 90 degrees, the fleeing distance was 2 units, the fleeing angle was 45 degrees at random direction away from the opponent and the chasing distance was 1 unit in the direction of the opponent. To resemble the difference in physical strength between males and females both sexes started out with different winning or loosing tendencies and females had only 80% of the aggression intensity the males had. The experiment was conducted with 4 different conditions. We used two levels of aggression. In the high level the intensity of aggression of males was 1 and of females 0.8, in the low aggression level it was 0.1 for males and 0.08 for females. In both levels attraction was whether turned on or off, resulting in 4 different condition. Each condition was run for 10 times, resulting in a total number of 40 runs. by courtesy of Mr. Minoru Kinoshita, Kyoto University - Goodall, J. (1986). The chimpanzees of Gombe: patterns of behavior. — Belknapp Press of Harvard University Press, Cambridge MA and London. - Hemelrijk C.K. (2002). Despotic societies, sexual attraction and the emergence of male ”tolerance”: an agent-based model — Behaviour 139, p Hemelrijk C.K. (2002). Self-organization and natural selection in the evolution of complex despotic societies. — Biol. Bull. 202, p Yerkes, R.M. (1939). Social dominance and sexual status in chimpanzees. — Quart. Rev. Biol. 14, p de Waal, F.B.M. (1989). Foodsharing and reciprocal obligations among chimpanzees. — J. Human Evolution 18, p Stanford, C.B. (1996). The hunting ecology of wild chimpanzees: Implications for the evolutionary ecology of Pliocene hominids. — Am. Anthropol. 98, p FUTURE WORK We are planning to build our own model based on empirical research data on different macaque species, done in the lab of Bernard Thierry. We plan to implement not only instantaneous action selection, but also simple cognitive state including emotions, drives such as foraging & grooming, and internal models of social relations, which may form the foundation of coalitions.