Agent-based simulations of the evolution of psychopathy
– by Dražen Domijan
Agent-based simulations offer valuable insights into processes that are difficult or impossible to observe directly. They enable systematic variation of relevant factors and examination of their influence on the behavior of simulated agents. For example, one can examine and compare how the same trait evolves under favorable conditions in an environment with abundant resources, or in a harsh environment where resources to sustain life are scarce. In this work, we were inspired by a recently developed agent-based model by Dr. Martina Testori and her colleagues. We are grateful to Dr. Testori for openly sharing their simulation code, which increases the transparency and reproducibility of computational studies and also significantly reduces the time required to develop our extended model.
Testori et al. adopted a public goods game in which generous and selfish agents interact to produce resources that are shared equally among all members of the society. Agents then reproduce and are replaced by their offspring in the next generation. The number of offspring each parent produces is proportional to their fitness. If an agent’s fitness is below the survival threshold, the agent dies without leaving offspring. Reproduction is governed by a probabilistic mechanism that incorporates genetic propensity for selfishness and behavioral contribution to the development of selfish and generous personality traits. This mechanism enables offspring to inherit their parents’ phenotype or to switch to the opposite phenotype with a certain probability of mutation.
To simulate psychopathic traits in more detail, Testori et al. introduced two additional behavioral components of the psychopathic personality, beyond mere selfishness. Firstly, psychopathic individuals tend to engage in risky behaviors that may lead to their premature deaths and removal from society. In the model, this is represented by a parameter called the mortality rate, meaning that a fraction of psychopaths were eliminated from the game before they had the opportunity to reproduce. Secondly, psychopathic individuals tend to engage in destructive acts that harm public goods, thereby imposing a community cost on available resources. The mortality rate and community cost are free parameters that are allowed to vary independently, enabling their effects on the evolution of psychopathy to be assessed in simulations.
When we began work on this project, our attention was immediately drawn to the finding that, in most reported simulations in the Testori et al. study, selfish agents dominated society. In some combinations of mortality rate and community cost, they constitute more than 80% of the population. However, this is not a plausible evolutionary outcome, as a recent meta-analysis suggests that the frequency of psychopaths in the human population is quite low. Estimates range from 1.2% to 4.5%, depending on the instrument used to assess psychopathy. Some authors even suggest a more conservative estimate of around 1% in the general population. This raises an important question: what societal forces keep the fraction of psychopathic individuals at such low levels?
Inspired by the work of Hiroaki Chiba-Okabe and Joshua Plotkin, we hypothesized that institutions could serve as a stabilizing factor that alleviate the adverse consequences of the actions of selfish agents. Here, institution should be understood in the broadest sense, involving any set of rules that regulate interactions among members of a society. In our model, the role of the institution is to act as an external agent that partially reallocates resources from psychopathic agents to cooperators, thus helping to reduce the resource inequality generated in the public goods game. In this way, the institution increases the fitness of generous agents who are disadvantaged in exchanges with selfish agents. We devised several scenarios in which the institution imposes punishment on selfish agents, or combines punishment of selfish agents with rewards for generous agents. We also considered the effect of conditional cooperation, that is, a scenario in which cooperators adjust their contributions to the public goods game by taking into account information about the level of cooperation in the previous generation.
Our simulations showed that both punishing selfish agents and rewarding generous agents were necessary to reduce the proportion of selfish agents in society to levels closer to those observed empirically. Furthermore, these measures must be accompanied by a high mortality rate among selfish agents. However, although we approached our target, we were unable to achieve exactly 1% selfish agents in the final population. Our results suggest that selfish agents are highly resistant and manage to survive institutional interventions. This supports the conclusion that psychopathy may have adaptive value because it strongly contributes to fitness. This is consistent with several recent studies showing that certain features of psychopathy, especially manipulation and emotional coldness, are positively associated with fertility as a crucial component of evolutionary fitness. Interestingly, across simulations we also observed that population size increases as institutional interventions become stronger. Thus, institutional redistribution of resources contributes to population growth.
To keep our model as simple as possible, we made several simplifications and left many topics unaddressed. For example, we treated personality traits as if they were consistent with genetic makeup. However, it is possible that psychopaths alter their behavior in situations where psychopathic behavior is punished or prosocial behavior is rewarded. We modeled psychopathy as a single phenotype, although it is better described as a behavioral syndrome that can be subdivided into several distinct phenotypes. We also kept the environment constant across generations, although it may fluctuate over time. We hope our work will inspire further computational explorations of the societal and ecological conditions that shape the evolution of psychopathy.
