Markus Brede

Playing against the fittest: A simple strategy that promotes the emergence of cooperation

Understanding the emergence and sustainability of cooperation is a fundamental problem in evolutionary biology and is frequently studied in the framework of evolutionary game theory. A very powerful mechanism to promote cooperation is network reciprocity, where the interaction patterns and opportunities for strategy spread of agents are constrained to limited sets of permanent interactions partners. Cooperation survives because it is possible for closeknit communities of cooperation to be shielded from invasion by defectors. Here we show that parameter ranges in which cooperation can survive are strongly expanded if game play on networks is skewed towards more frequent interactions with more successful neighbours. In particular, if agents exclusively select neighbors for game play that are more successful than themselves, cooperation can even dominate in situations in which it would die out if interaction neighbours were chosen without a bias or with a preference for less successful opponents. We demonstrate that the “selecting fitter neighbours” strategy is evolutionarily stable. Moreover, it will emerge as the dominant strategy out of an initially random population of agents.

Short Versus Long Term Benefits and the Evolution of Cooperation in the Prisoner's Dilemma Game

In this paper I investigate the evolution of cooperation in the prisoners dilemma when individuals change their strategies subject to performance evaluation of their neighbours over variable time horizons. In the monochrome setting, in which all agents per default share the same performance evaluation rule, weighing past events strongly dramatically enhances the prevalence of cooperators. For co-evolutionary models, in which evaluation time horizons and strategies can co-evolve, I demonstrate that cooperation naturally associates with long-term evaluation of others while defection is typically paired with very short time horizons. Moreover, considering the continuous spectrum in between enhanced and discounted weights of past performance, cooperation is optimally supported when cooperators neither give enhanced weight to past nor more recent events, but simply average payoffs. Payoff averaging is also found to emerge as the dominant strategy for cooperators in co-evolutionary models, thus proposing a natural route to the evolution of cooperation in viscous populations.

Coordinated and uncoordinated optimization of networks.

In this paper, we consider spatial networks that realize a balance between an infrastructure cost (the cost of wire needed to connect the network in space) and communication efficiency, measured by average shortest path length. A global optimization procedure yields network topologies in which this balance is optimized. These are compared with network topologies generated by a competitive process in which each node strives to optimize its own cost-communication balance. Three phases are observed in globally optimal configurations for different cost-communication trade offs: (i) regular small worlds, (ii) starlike networks, and (iii) trees with a center of interconnected hubs. In the latter regime, i.e., for very expensive wire, power laws in the link length distributions P(w)∝w(-α) are found, which can be explained by a hierarchical organization of the networks. In contrast, in the local optimization process the presence of sharp transitions between different network regimes depends on the dimension of the underlying space. Whereas for d=∞ sharp transitions between fully connected networks, regular small worlds, and highly cliquish periphery-core networks are found, for d=1 sharp transitions are absent and the power law behavior in the link length distribution persists over a much wider range of link cost parameters. The measured power law exponents are in agreement with the hypothesis that the locally optimized networks consist of multiple overlapping suboptimal hierarchical trees.

Patterns in randomly evolving networks: Idiotypic networks

We present a model for the evolution of networks of occupied sites on undirected regular graphs. At every iteration step in a parallel update, I randomly chosen empty sites are occupied and occupied sites having occupied neighbor degree outside of a given interval (t(l),t(u)) are set empty. Depending on the influx I and the values of both lower threshold and upper threshold of the occupied neighbor degree, different kinds of behavior can be observed. In certain regimes stable long-living patterns appear. We distinguish two types of patterns: static patterns arising on graphs with low connectivity and dynamic patterns found on high connectivity graphs. Increasing I patterns become unstable and transitions between almost stable patterns, interrupted by disordered phases, occur. For still larger I the lifetime of occupied sites becomes very small and network structures are dominated by randomness. We develop methods to analyze the nature and dynamics of these network patterns, give a statistical description of defects and fluctuations around them, and elucidate the transitions between different patterns. Results and methods presented can be applied to a variety of problems in different fields and a broad class of graphs. Aiming chiefly at the modeling of functional networks of interacting antibodies and B cells of the immune system (idiotypic networks), we focus on a class of graphs constructed by bit chains. The biological relevance of the patterns and possible operational modes of idiotypic networks are discussed.

Architecture of idiotypic networks: Percolation and scaling Behavior

We investigate a model where idiotypes (characterizing B lymphocytes and antibodies of an immune system) and anti-idiotypes are represented by complementary bit strings of a given length d allowing for a number of mismatches (matching rules). In this model, the vertices of the hypercube in dimension d represent the potential repertoire of idiotypes. A random set of (with probability p) occupied vertices corresponds to the expressed repertoire of idiotypes at a given moment. Vertices of this set linked by the above matching rules build random clusters. We give a structural and statistical characterization of these clusters, or in other words of the architecture of the idiotypic network. Increasing the probability p one finds at a critical p a percolation transition where for the first time a large connected graph occurs with probability 1. Increasing p further, there is a second transition above which the repertoire is complete in the sense that any newly introduced idiotype finds a complementary anti-idiotype. We introduce structural characteristics such as the mass distribution and the fragmentation rate for random clusters, and determine the scaling behavior of the cluster size distribution near the percolation transition, including finite size corrections. We find that slightly above the percolation transition the large connected cluster (the central part of the idiotypic network) consists typically of one highly connected part and a number of weakly connected constituents and coexists with a number of small, isolated clusters. This is in accordance with the picture of a central and a peripheral part of the idiotypic network and gives some support to idealized architectures of the central part used in recent dynamical mean field models.

Multiplex lexical networks reveal patterns in early word acquisition in children

Network models of language have provided a way of linking cognitive processes to language structure. However, current approaches focus only on one linguistic relationship at a time, missing the complex multi-relational nature of language. In this work, we overcome this limitation by modelling the mental lexicon of English-speaking toddlers as a multiplex lexical network, i.e. a multi-layered network where N = 529 words/nodes are connected according to four relationship: (i) free association, (ii) feature sharing, (iii) co-occurrence, and (iv) phonological similarity. We investigate the topology of the resulting multiplex and then proceed to evaluate single layers and the full multiplex structure on their ability to predict empirically observed age of acquisition data of English speaking toddlers. We find that the multiplex topology is an important proxy of the cognitive processes of acquisition, capable of capturing emergent lexicon structure. In fact, we show that the multiplex structure is fundamentally more powerful than individual layers in predicting the ordering with which words are acquired. Furthermore, multiplex analysis allows for a quantification of distinct phases of lexical acquisition in early learners: while initially all the multiplex layers contribute to word learning, after about month 23 free associations take the lead in driving word acquisition.

The evolution of cooperation on correlated payoff landscapes

We study the evolution of cooperation in the prisoners dilemma game on time-invariant heterogeneous payoff landscapes on regular and heterogeneous networks. Correlations in the landscape structure and their implications for the evolution of cooperation are investigated. On regular networks we find that negatively and neutrally correlated payoff landscapes strongly enhance cooperation, while positively correlated landscapes may suppress the evolution of cooperation. On heterogeneous networks, cooperation is facilitated if payoff stochasticity is positively correlated with network heterogeneity and may be suppressed otherwise.

How evolution learns to generalise: using the principles of learning theory to understand the evolution of developmental organisation

One of the most intriguing questions in evolution is how organisms exhibit suitable phenotypic variation to rapidly adapt in novel selective environments. Such variability is crucial for evolvability, but poorly understood. In particular, how can natural selection favour developmental organisations that facilitate adaptive evolution in previously unseen environments? Such a capacity suggests foresight that is incompatible with the short-sighted concept of natural selection. A potential resolution is provided by the idea that evolution may discover and exploit information not only about the particular phenotypes selected in the past, but their underlying structural regularities: new phenotypes, with the same underlying regularities, but novel particulars, may then be useful in new environments. If true, we still need to understand the conditions in which natural selection will discover such deep regularities rather than exploiting ‘quick fixes’ (i.e., fixes that provide adaptive phenotypes in the short term, but limit future evolvability). Here we argue that the ability of evolution to discover such regularities is formally analogous to learning principles, familiar in humans and machines, that enable generalisation from past experience. Conversely, natural selection that fails to enhance evolvability is directly analogous to the learning problem of over-fitting and the subsequent failure to generalise. We support the conclusion that evolving systems and learning systems are different instantiations of the same algorithmic principles by showing that existing results from the learning domain can be transferred to the evolution domain. Specifically, we show that conditions that alleviate over-fitting in learning systems successfully predict which biological conditions (e.g., environmental variation, regularity, noise or a pressure for developmental simplicity) enhance evolvability. This equivalence provides access to a well-developed theoretical framework from learning theory that enables a characterisation of the general conditions for the evolution of evolvability.

Detecting and Characterizing Eating-Disorder Communities on Social Media

Eating disorders are complex mental disorders and responsible for the highest mortality rate among mental illnesses. Recent studies reveal that user-generated content on social media provides useful information in understanding these disorders. Most previous studies focus on studying communities of people who discuss eating disorders on social media, while few studies have explored community structures and interactions among individuals who suffer from this disease over social media. In this paper, we first develop a snowball sampling method to automatically gather individuals who self-identify as eating disordered in their profile descriptions, as well as their social network connections with one another on Twitter. Then, we verify the effectiveness of our sampling method by: 1. quantifying differences between the sampled eating disordered users and two sets of reference data collected for non-disordered users in social status, behavioral patterns and psychometric properties; 2. building predictive models to classify eating disordered and non-disordered users. Finally, leveraging the data of social connections between eating disordered individuals on Twitter, we present the first homophily study among eating-disorder communities on social media. Our findings shed new light on how an eating-disorder community develops on social media.

Harvesting heterogeneous renewable resources: Uncoordinated, selfish, team-, and community-oriented strategies

Using the example of a fishing fleet harvesting in different fishing zones with different carrying capacities and growth rates, we investigate strategies for the exploitation of distributed renewable resources by a crowd of agents without centralized coordination. In agent-based simulations we compare the performance of uncoordinated random harvesting, team playing, selfish individualistic and community-oriented (Collective Intelligence or COIN) behaviours operating with long and short time-horizon planning. Demonstrating the usefulness of COIN-based harvesting, more cooperative long-term planning strategies are found to relieve the pressure on the resource, reduce fluctuations and diminish the risk of overharvesting. Further, the outcome of an evolutionary dynamics where strategies in the agent population spread proportional to relative economic performance are strongly influenced by the harvesting pressure. In order of decreasing resource abundance we find that first an uncoordinated random, then a cooperative COIN-strategy and later the selfish strategy for an overharvested resource dominate the agent population. We also report that increasing harvesting pressure increasingly favours short-term and more individualistic strategies.