ABSTRACT. Cooperative behaviour has been extensively studied, in both evolutionary biology and the social sciences, as a choice between cooperation and defection. However, in many cases, the possibility to not participate or to exit a situation is also available. This type of problem can be studied through the optional public goods game. The introduction of the "Loner" strategy, allows players to withdraw from the public goods game, radically changing the dynamics of cooperation in social groups and leading to a never-ending cooperator-defector-loner cycle. While pro-social punishment has been found to help increase cooperation, anti-social punishment -- where defectors punish cooperators -- causes the downfall of cooperation in both experimental and theoretical studies. In this paper, we extend the theory of the optional public goods game, introducing reputational dynamics in the form of social norms that allow agents to condition both their participation and contribution decisions to the reputation of their peers. We benchmark this setup both with respect to the standard optional public goods game and to the variant where all types of punishment are allowed. We find that a social norm imposing a more moderate reputational penalty for opting out than for defecting, increases cooperation. When, besides reputation, punishment is also possible, the two mechanisms work synergically under all norms that do not punish loners too harshly. Under this latter setup, the high levels of cooperation are sustained by conditional strategies, which largely reduce the use of pro-social punishment and almost completely eliminate anti-social punishment.

ABSTRACT. My current research concerns experimental Cournot games. While theory predicts that players will collude in repeated games, we typically see a lot of volatility. I think that this is due to incentives being too low. As a result, the goals of players are not aligned: some try to maximize absolute profits while others try to earn more, or more often than their opponents. This suggests that these experiments should be interpreted as exercises in coordination. For instance, it is non-trivial how two cooperative players should play against a competitive player that wants to earn more than the others. My hypothesis is that players communicate with each other through patterns in quantities (see slides).

Before building an agent-based model that uses these patterns, I want to learn which patterns are important and how. For this, I will use machine learning. I am currently building agents that are modeled after a Learning Classifier System (LCS). An LCS can be optimized by letting it play against itself. Although optimized, my LCS agents qualify as Minimal Intelligence agents in a social sense: the extent to which they perceive signals from other players depends on the quality of their perceptors. By evolving the perceptors (and by assuming that cooperative players are somewhat impatient), it must be possible to demonstrate the value of being aware of patterns in achieving collusion (if any). An LCS will produce rules that are intelligible, and therefore should be helpful in developing an ABM at a later stage.

The submission consists of this abstract, and a presentation + speaker notes (adapted from a presentation for the Barcelona Summer Forum 2021).

ABSTRACT. The outcomes from strategic decision-making (such as market entry or technology adoption) depend on structural features of situations and types of players involved. Even in identical situations, players differ in their perceptions of situations, goals, and strategic sophistication. Informed by behavioral and experimental economics, we present two heuristic player types under strategic uncertainty in new situations, exemplified within the simplest class of games, two players–two actions (2x2) games, e.g., Prisoner's dilemma or Entry games. One type anticipates others' behavior and (iteratedly) best-replies to beliefs (called iterated reasoning heuristic), while the other is guided by goals, e.g., equality or social optimum, ignoring procedural details, such as others' reasoning (called efficiency heuristic). To understand the implications of structure and player types, we develop a behavioral system of 2x2 games. Due to the fundamental differences of the two types, the large set of 2x2 games collapses to four distinct classes for efficiency types and five, albeit different ones, for iterated reasoning, and to 14 in a joint system based on (behavioral) game-theoretic features. Thus, advanced knowledge of players' strategic capabilities or goals considerably simplifies the strategic analysis by inducing a categorization of games. Furthermore, we predict differences in the heterogeneity of behavior and outcomes, depending on player types and structure. We also discuss extensions of these situations and how heuristics may help agents gauge and respond to others' thinking better

ABSTRACT. Humans are capable of solving cooperation problems following social norms. Social norms dictate appropriate behaviour and judgement on others in response to their previous actions and reputation. Recently, the so-called \emph{leading eight} norms have been identified from many potential social norms that can sustain cooperation through a reputation-based indirect reciprocity mechanism. Despite indirect reciprocity being claimed to extend direct reciprocity in larger populations where direct experiences cannot be accumulated, the success of social norms have been analysed in models with global information and evolution. This study is the first to analyse the leading eight norms with local information and evolution. We find that the leading eight are robust against selfish players within most scenarios and can maintain a high level of cooperation also with local information and evolution. In fact, local evolution sustains cooperation under a wider set of conditions than global evolution, while local reputation does not hinder cooperation compared to global reputation. Four of the leading eight norms that do not reward justified defection offer better chances for cooperation with quick evolution, reputation with noise, larger networks, and when unconditional defectors enter the population.

ABSTRACT. This paper introduces a model of a stylized organization that is comprised of several departments that autonomously allocate tasks. To do so, the departments either take short-sighted decisions that immediately maximize their utility or take long-sighted decisions that aim at minimizing the interdependencies between tasks. The organization guides the departments' behavior by either an individualistic, a balanced, or an altruistic linear incentive scheme. Even if tasks are perfectly decomposable, altruistic incentive schemes are preferred over individualistic incentive schemes since they substantially increase the organization's performance. Interestingly, if altruistic incentive schemes are effective, short-sighted decisions appear favorable since they do not only increase performance in the short run but also result in significantly higher performances in the long run.

ABSTRACT. The paper studies how the heterogeneity of minimal intelligence agents shapes the effects of coordination mechanisms in organizational decision-making. For this, the study employs an agent-based simulation based on the framework of NK fitness landscapes. The results suggest that tighter coordination does not universally level out heterogeneity in decision-making across decision-makers but could result in even higher differences between homogenous and heterogenous decision-making styles.