Advances in Complex Systems

Since social-economic systems increase interdependency, a crucial question arises: Is an interconnected world a safer or a more dangerous place to live? Over the last few years, we have witnessed the dark side of increasing interdependencies. As such, there is a growing need to focus on how to mitigate networked risk and to enhance the system resilience to the impact of a large-scale shock. The traditional engineering approach has been to design systems that are less vulnerable to damage from hazard events. On the other hand, system resilience is the ability to recover from failure and provide the continuity of system function. The goal of the present paper is to investigate the gain from risk sharing. We propose a mechanism of risk sharing that may enhance the resilience of the networked systems. The proposed risk sharing protocols are based on coordinated incentives of agents to survive collectively by absorbing external shocks. The key issue we would like to analyze is how the gain from risk sharing depends on the capacity of each agent to absorb shock and on the interconnections patterns among agents with risk sharing rules. We demonstrate that risk sharing is beneficial from a systems point of view when the agents' capacities to shocks is high and detrimental when it is low. In particular, we evaluate the effectiveness of risk sharing in two domains. In the first domain, in which networked agents have the possibility of cascading failure, risk sharing is useful in mitigating systemic failure, especially if the agents are running at high load. In the second domain, we evaluate the ratio of safe agents who invest in risky portfolios or projects collectively. In this case, risk sharing is only beneficial if the agents' risk absorbing capacity is high.

In this paper, we present a new agent-based model for the simulation of tax compliance and tax evasion behavior (SIMULFIS). The main novelties of the model are the introduction of a "behavioral filter approach" to model tax decisions, the combination of a set of different mechanisms to produce tax compliance (namely rational choice, normative commitments and social influence), and the use of the concept of "fraud opportunity use rate" (FOUR) as the main behavioral outcome. After describing the model in detail, we display the main behavioral and economic results of 1,920 simulations calibrated for the Spanish case and designed to test for the internal validity of SIMULFIS. The behavioral outcomes show that scenarios with strict rational agents strongly overestimate tax evasion, while the introduction of social influence and normative commitments allows to generate more plausible compliance levels under certain deterrence conditions. Interestingly, the relative effect of social influence is shown to be ambivalent: it optimizes compliance under low and middle deterrence conditions, but not when deterrence is made harder. Finally, SIMULFIS economic outcomes are broadly in line with theoretical expectations, thus supporting the reliability of the model.

In the Internet era, online social media emerged as the main tool for sharing opinions and information among individuals. In this work, we study an adaptive model of a social network where directed links connect users with similar tastes, and over which information propagates through social recommendation. Agent-based simulations of two different artificial settings for modeling user tastes are compared with patterns seen in real data, suggesting that users differing in their scope of interests is a more realistic assumption than users differing only in their particular interests. We further introduce an extensive set of similarity metrics based on users' past assessments, and evaluate their use in the given social recommendation model with both artificial simulations and real data. Superior recommendation performance is observed for similarity metrics that give preference to users with small scope — who thus act as selective filters in social recommendation.

Socio-technical systems involve agents who create and process knowledge, exchange information and create ties between ideas in a distributed and networked manner: webloggers, communities of scientists, software developers and wiki contributors are, among others, examples of such networks. The state-of-the-art in this regard focuses on two main issues which are generally addressed in an independent manner: the description of content dynamics and the study of social network characteristics and evolution. This paper relies on recent endeavors to merge both types of dynamics into co-evolutionary, multi-level modeling frameworks, where social and semantic aspects are being jointly appraised. Case studies featuring socio-semantic graphs, socio-semantic hypergraphs and socio-semantic lattices are notably discussed.

We study a system in which N agents have to decide between two strategies θ_i(i ∈ 1 … N), for defection or cooperation, when interacting with other n agents (either spatial neighbors or randomly chosen ones). After each round, they update their strategy responding nonlinearly to two different information sources: (i) the payoff a_i(θ_i, f_i) received from the strategic interaction with their n counterparts, (ii) the fraction f_i of cooperators in this interaction. For the latter response, we assume social herding, i.e., agents adopt their strategy based on the frequencies of the different strategies in their neighborhood, without taking into account the consequences of this decision. We note that f_i already determines the payoff, so there is no additional information assumed. A parameter ζ defines to what level agents take the two different information sources into account. For the strategic interaction, we assume a Prisoner's Dilemma game, i.e., one in which defection is the evolutionary stable strategy. However, if the additional dimension of social herding is taken into account, we find instead a stable outcome where cooperators are the majority. By means of agent-based computer simulations and analytical investigations, we evaluate the critical conditions for this transition toward cooperation. We find that, in addition to a high degree of social herding, there has to be a nonlinear response to the fraction of cooperators. We argue that the transition to cooperation in our model is based on less information, i.e., on agents which are not informed about the payoff matrix, and therefore rely on just observing the strategy of others, to adopt it. By designing the right mechanisms to respond to this information, the transition to cooperation can be remarkably enhanced. Our results are obtained in an evolutionary PD game with fixed payoffs and a fixed four-player neighborhood, where agents follow a stochastic better response dynamics.

Massively distributed data mining in large networks such as smart device platforms and peer-to-peer systems is a rapidly developing research area. One important problem here is concept drift, where global data patterns (movement, preferences, activities, etc.) change according to the actual set of participating users, the weather, the time of day, or as a result of events such as accidents or even natural catastrophes. In an important case — when the network is very large but only a few training samples can be obtained at each node locally — no efficient distributed solution is known that could follow concept drift efficiently. This case is characteristic of smart device platforms where each device stores only one local observation or data record related to a learning problem. Here we present two algorithms to handle concept drift. None of the algorithms collects data to a central location, instead models of the data perform random walks in the network, while being improved using an online learning algorithm. The first algorithm achieves adaptivity by maintaining young as well as old models in the network according to a fixed age distribution. The second one measures the performance of models locally, and discards them if they are judged outdated. We demonstrate through a thorough experimental analysis that our algorithms outperform the known competing methods if the number of independent local samples is limited relative to the speed of drift: a typical scenario in our targeted application domains. The two algorithms have different strengths: while the age distribution approach is very simple and efficient, explicit drift detection can be useful in monitoring applications to trigger control action.

Energy waste due to inhabitants' behavior in residential buildings has emerged as a potential research area due to the increasing worldwide population and growing energy needs. However, existing approaches for simulating energy consumption are mainly limited to office buildings and are based on static profiles. In this paper we propose a 4-step co-simulation methodology to assess how inhabitants' interactions with household appliances affect energy consumption. The approach is validated using a case study showing how human activities influence the energy consumption patterns of a refrigerator. The fridge was specifically chosen because it is a high energy-consuming appliance that is strongly affected by inhabitants' behaviors. In addition, modeling the fridge is nontrivial, and in choosing this appliance we show that it is possible to apply the approach to less complex appliances. A co-simulation approach is adopted with the fridge being physically modeled in Matlab and with human behavior being modeled in the Brahms language and simulation environment. The consumption distribution from the simulated scenario is compared with the actual distribution (using data from a consumption database), to find optimum values of tuning parameters with less than 10% variation. This methodology enables us to simulate how human behaviors affect energy appliance consumption.

In this paper, we report on the theoretical foundations, empirical context and technical implementation of an agent-based modeling (ABM) framework, that uses a high-performance computing (HPC) approach to investigate human population dynamics on a global scale, and on evolutionary time scales. The ABM-HPC framework provides an in silico testbed to explore how short-term/small-scale patterns of individual human behavior and long-term/large-scale patterns of environmental change act together to influence human dispersal, survival and extinction scenarios. These topics are currently at the center of the Neanderthal debate, i.e., the question why Neanderthals died out during the Late Pleistocene, while modern humans dispersed over the entire globe. To tackle this and similar questions, simulations typically adopt one of two opposing approaches, top-down (equation-based) and bottom-up (agent-based) models of population dynamics. We propose HPC technology as an essential computational tool to bridge the gap between these approaches. Using the numerical simulation of worldwide human dispersals as an example, we show that integrating different levels of model hierarchy into an ABM-HPC simulation framework provides new insights into emergent properties of the model, and into the potential and limitations of agent-based versus continuum models.

Online systems, where users purchase or collect items of some kind, can be effectively represented by temporal bipartite networks where both nodes and links are added with time. We use this representation to predict which items might become popular in the near future. Various prediction methods are evaluated on three distinct datasets originating from popular online services (Movielens, Netflix, and Digg). We show that the prediction performance can be further enhanced if the user social network is known and centrality of individual users in this network is used to weight their actions.

The speed of social and technological changes is constantly increasing. Change is a pre-requisite for economic development but this increasing speed carries additional costs that may largely affect the ability of the social systems to adopt them. While the financial costs constitute a limit to adoption, the psychological and social costs may also profoundly change the adoption potential. In an agent-based model of a heterogeneous population of adopters we explore the consequences of increasing the speed of novelty introduction on their satisfaction and the degree to which new trends may permeate the system. We show that introduction speed has a diametrically different impact on different adopter groups: opinion leaders are most satisfied when the mainstream individuals are least content and vice versa. Moreover, introduction speed profoundly affects the ability of trends to penetrate the system — the lower the introduction speed, the higher the level of penetration of the social system. With high speed of introduction, only a small fraction of the most attractive novelties is able to permeate the system. In sum, these results show that both the wellbeing of the social system as well as its capacity to adopt novelties are dependent on the speed of change.

We analyze emotionally annotated massive data from Internet relay chat (IRC) as well as from BBC forum website and model the dialogues between chat participants by assuming that the driving force for the discussion is the entropy growth of emotional probability distribution. This process is claimed to be responsible for a power-law distribution of the discussion lengths observed in the dialogues. We perform numerical simulations based on the noticed phenomenon obtaining a good agreement with the real data. Finally, we propose a method to artificially prolong the duration of the discussion that relies on the entropy of emotional probability distribution.

Simulation with agent-based models is increasingly used in the study of complex socio-technical systems and in social simulation in general. This paradigm offers a number of attractive features, namely the possibility of modeling emergent phenomena within large populations. As a consequence, often the quantity in need of calibration may be a distribution over the population whose relation with the parameters of the model is analytically intractable. Nevertheless, we can simulate. In this paper we present a simulation-based framework for the calibration of agent-based models with distributional output based on indirect inference. We illustrate our method step by step on a model of norm emergence in an online community of peer production, using data from three large Wikipedia communities. Model fit and diagnostics are discussed.