“It takes a village to raise a child.” This old pro­verb reminds us that community is a defining feature of human life, an organizing principle –whether it’s a literal village or any other social group to which we belong. Humans form communities based on their common interests, opinions, preferences or cultures. Like-minded individuals gather in forums, conferences and special events and share their experiences and knowledge, doing things together that they couldn’t do alone.

Today, the emergence of ubiquitous computing and networks has expanded social connectivity to an unprecedented scale. Billions of people come together on social networking sites, building social ties and forming virtual communities. More and more companies now provide platforms for people to interact with each other, exchanging and sharing ideas, goods and services. These complex networks have opened up a whole new world of thinking about communities. As a result, community detection has emerged as a new research direction in social network analysis.

Larisa Adamyan is a PhD student in "Machine Learning" at Humboldt University of Berlin. She is also the co-founder of Expanenta, an analytics service for evaluation of offline marketing performance. Expanenta helps customers track traditional advertisements (print, TV, radio) as simply as online campaigns, by using machine learning technology.

On the surface, defining a community is simple – it’s a group of individuals who interact with each other more frequently than with those outside the group. Except in a world now driven by widespread connectivity, people aren’t as simple anymore: they tend to belong to many communities, and their connections within those are constantly evolving. This is where community detection really comes into its own as a field of research. To truly understand and map what makes communities tick, we need to design algorithms to help us discover overlaps in these networks. Machine learning and graph theory lie at the heart of these algorithms. They provide us with tools and methods to analyze any social network in terms of nodes (individuals or users) and edges connecting these nodes (relationships). At this level, a community is a group of nodes that are more densely connected to each other than to the rest of the network. Algorithms detect communities by finding groups of nodes in a network with high edge density.

As our thinking about communities has expanded, we’re seeing how the concept and related methods for detecting them are applied across disciplines involving networked systems – not just those formed by humans. For example, in the human brain the equivalent of communities would be the modules of cells that are densely connected with each other and responsible for specific tasks. In molecular networks of protein-to-protein interactions, the communities are the groups of proteins that contribute to the same cellular function. Community detection can help improve marketing performance by targeting influential users of each community and reaching their community through them, or providing useful recommendations based on users’ common interests. Then in fields ranging from physiology to nutrition to providing services, detecting and mapping communities is critical in understanding the structure of complex networks.

The good news is that with bigger data sets, faster computers and better algorithms, we can now quite literally detect what makes humans tick – and better understand how exactly all those virtual villages can help us tackle the most pressing challenges of our time.

Illustration: Valentin Berger

Dieser Artikel ist in unserer Dezember-Ausgabe 2018 „Sharing Economy“ erschienen.