Naming game. Models, simulations and analysis (Q822688)

This computationally oriented book is mostly about some interesting variants of the naming game (NG), a semiotic model of consensus inspired by Luc Steels' experiments with autonomous robots and the language games proposed by Ludwig Wittgenstein in his book \textit{Philosophical Investigations} (1953). Most of these variants aim to simulate the dynamics of a multi-agent system according to the NG's rules in real-life scenarios. Chapter 1 gives a general overview of the research conducted in this field until the publication of the book. Therefore, it misses some recent advances where the word learning process has been implemented according to a Bayesian framework proposed by Joshua Tenenbaum, to take into account the real behavior of human learners, and hence the uncertainty that characterizes their learning tasks. Chapter 2 presents a detailed introduction to the NG model, commonly referred to as the minimal NG, by means of pedagogical cartoons and some useful flowcharts. It also contains a succinct presentation of some complex networks on which the NG time-evolution is routinely simulated to study the possible emergence of global consensus. Therefore, the first two chapters of the book could be useful to readers who wish to learn the basic facts about the naming game model and its computational implementation. In the rest of the book, the NG variants are extensively studied by carrying out computer simulations of the multi-agent system embedded in different complex networks to compute typical observables, such as the total number of words, the number of different words, the success rate, and other possible quantities of interest, e.g., the convergence time required to reach global consensus. However, the analysis of the scaling behavior with the system size is generally missing, and hence the reader is referred to the relevant references in the bibliography of the book. Chapter 3 considers an NG dynamics where the agents can forget the names due to the finite size of their memory while in Chapter 5, the authors study how the communications with errors between the agents can affect the dynamics. Essentially, these two models seek to provide agents with some human-like features that we observe in everyday life, e.g. the word learning is an error-prone process. In Chapter 4, the authors illustrate two models that involve multiple agents communicating simultaneously while placed on some complex networks. Such variants of the NG model aim to simulate the broadcasting and group discussion between the agents as commonly observed in real-world communications. In Chapter 6, the authors study the NG dynamics on multi-local-world networks for the latter is capable of capturing some peculiar features of the Internet and also compare their findings with those obtained from numerical simulations on other typical underlying topologies, e.g., free-scale networks. In the last two chapters (6 and 7), the authors present the multi-word naming game that attempts to explain the emergence of sentences in a natural language such as English, and a multilingual naming game scenario in which the English and Chinese speaking agents try to communicate via translators, respectively. Overall, researchers who wish to reproduce the results presented in the book by their own computer codes might find this monograph very useful, as it is accompanied by several tables containing detailed information about the simulation settings and the numerical results, as well as a large number of plots about the outputs obtained by the authors in silico experiments.