arXiv:1310.7769v1 [cs.SI] 29 Oct 2013 Evolution of interaction networks On the evolution of interaction networks: primitive typology of vertex and prominence of measures. Renato Fabbri, 1, a) Vilson Vieira da Silva Junior, b) Ricardo Fabbri, c) Deborah Christina Antunes, d) Marilia Mello Pisani, e) Luciano da Fontoura Costa, f) and Osvaldo N. Oliveira Jr. g) Instituto de F´ ısica de S˜ao Carlos, Universidade de S˜ao Paulo (IFSC/USP) (Dated: 16 December 2013) This article reports a minimal and general characterization of interaction networks evolution. Such a task involves a selection of aspects to investigate, which lead to: 1) activity distribution in time and among participants, 2) a sound and stable classification of vertex: peripheral, intermediary and hub sectors, 3) composition of basic measures into components with greater dispersion. While time patterns of activity are not obvious, participant activity follow concentrations expected by scale-free networks. Comparison with ideal Erd¨ os-R´ enyi network with the same number of edges and vertexes revealed as a sound criterion for distinguishing sectors on the networks. Principal components in basic measures spaces revealed interesting and regular patterns of independence and dispersion. This includes a ranking of measures that most contribute to dispersion: 1) degree and strength measures, 2) symmetry related quantization, and 3) clusterization. Results suggested typologies for these networks and participants. Further work include considerations of text production, psychoanalysis inspired typologies, participatory democracy exploitation of observed properties, and better visualization support for network evolution. PACS numbers: 89.75.Fb,05.65.+b,89.65.-s Keywords: complex networks, social network analysis, pattern recognition, statistics ‘The conception of personality structure is the best safeguard against the inclination to attribute persistent trends in the individual to something ”innate” or ”basic” or ”racial” within him. The Nazi allegation that natural, biological traits de- cide the total being of a person would not have been such a successful political device had it not been possible to point to numerous instances of relative fixity in human behavior and to challenge those who thought to explain them on any basis other than a biological one.’ - T. Adorno (The Authoritarian Personality, Studies in Prejudice Series, 1950) a) http://ifsc.usp.br/˜fabbri/; Electronic mail: fabbri@usp.br b) http://automata.cc/; Electronic mail: vilson@void.cc; Also at IFSC-USP c) http://www.lems.brown.edu/˜rfabbri/; Electronic mail: rfab- bri@iprj.uerj.br; Instituto Polit´ ecnico, Universidade Estadual do Rio de Janeiro (IPRJ) d) http://lattes.cnpq.br/1065956470701739; Electronic mail: deb- orahantunes@gmail.com; Departamento de Psicologia, Universi- dade Federal do Cer´ a (DP/UFC) e) http://lattes.cnpq.br/6738980149860322; Electronic mail: mar- ilia.m.pisani@gmail.com; Centro de Cincias Naturais e Humanas, Universidade Federal do ABC (CCNH/UFABC) f) http://cyvision.ifsc.usp.br/˜luciano/; Electronic mail: ldf- costa@gmail.com; Also at IFSC-USP g) www.polimeros.ifsc.usp.br/professors/professor.php?id=4; Elec- tronic mail: chu@ifsc.usp.br; Also at IFSC-USP I. INTRODUCTION Networks evolution has received dedicated attention from the research community for more than a decade, with punctual and rich investigations, such as 1,2 . This ar- ticle observes interaction network evolution, driven from public email lists, with focus on topological aspects. While significant measures will depend on the model and system characteristics 3,4 , this work considers only di- rected, weighted and human interaction networks. Undi- rected and unweighted representation of such networks is also seen in the literature 5 , which can be obtained by simplification. Although all networks considered origi- nated from email lists, coherence with literature suggests that results hold for a more general class of interaction networks, such as observed in platforms (LinkedIn, Face- book). The GMANE email archive 6 was used to obtain the networks studied. This archive consists of more than 20,000 email lists and more than 130,000,000 messages 7 . Lists span a variety of topics, mostly technology-related. It can be seen as a corpus with metadata of its mes- sages, like time, place, sender, etc. It has been used as an archive and as a news gateway. Its use in scientific re- search is reported in studies of isolated lists and of lexical innovations 5,8 . In this article, a minimal topological characterization is addressed. This purposes to support ongoing work in text production and typologies of online participants 9,10 .