SCube: A Tool for Segregation Discovery

Segregation discovery consists of finding contexts of segregation of social groups distributed across units or communities. The SCube system implements an approach for segregation discovery on top of frequent itemset mining, by offering to the analyst a multi-dimensional (segregation) data cube for exploratory data analysis. The demonstration first guides the audience through the social science concepts, then describes the architecture of the SCube system and how it is able to deal with both relational data and with attributed graph data modeling social networks. Finally, we go through two running case studies in the context of occupational segregation in the boards of company directors using a 2012 snaphost of the Italian companies, and a 20-year long dataset of Estonian companies. I. SOCIAL SEGREGATION Ethical issues in data and knowledge engineering are gaining momentum in the last few years. In addition to the traditional field of privacy, techniques for data analysis are being designed or enhanced to take into account moral values such as fairness, transparency, accountability, and diversity. See e.g., the resources listed in https://fatconference.org. We have recentely developed a data-driven technique for adddressing segregation of social groups through multi-dimensional data analysis [1]. The approach is implemented in the SCube system, which we propose to demonstrate. Social segregation refers to the “separation of socially defined groups” [2]. People are partitioned into two or more groups on the grounds of personal or cultural traits that can foster discrimination, such as gender, age, ethnicity, income, skin color, language, religion, political opinion, membership of a national minority, etc. [3]. Contact, communication, or interaction among groups are limited by their physical, working or socio-economic distance. This can be observed when dissecting society in organizational units (neighborhoods, schools, job types). Due to the ubiquitous presence and pervasiveness of ICT, segregation is shifting from ancient forms of spatial segregation to modern cultural segregation. For instance, it has been warned that the filter bubble generated by personalization of online social networks may foster idelogical segregation [4], opinion polarization [5], and informational segregation [6]. A technology that enables to assess the extent, nature, and trends of social segregation is of extreme interest for a wide audience: social scientistics, public policy makers, regulation and control authorities professional associations, civil rights societie, and investigative journalists. Business decision makers should also care of business practices, or even automated decisions and recommendation, that segregate customers and users into stereotypes. young middle age (SA) elder ∗ re gio n (C A) female sex (SA) male ∗ ∗