Most people will be familiar with popular social networking services such as Facebook and Twitter. Recently, organisations have begun to appreciate the benefits of adopting these types of platforms in order to facilitate interaction and collaboration between their employees. For example, there has been a growing trend towards the adoption of enterprise social networking services such as Yammer, Slack and IBM Connections. In addition, most organisations continue to rely on traditional forms of online communication such as email and IRC. Clearly, an organisation's real-world social network plays an important role in the functioning of the organisation and can tell us much about an employee's informal transversal competencies. By analysing an organisation's virtual social networks, we can obtain useful insights into characteristics of this real-world social network.
With that in mind, the DEVELOP project aims to apply social network analysis to understand an organisation's social network and an employee's position within this network. In particular, using the concept of social capital, we aim to study how information about the social characteristics of an employee can be used to influence and improve career planning. For example, we might infer evidence of informal workplace transversal competencies such as leadership, mentorship and collaboration. This evidence may then be used when identifying career goals and when suggesting learning interventions towards achieving these goals. By considering the issue of concept drift, we also aim to identify changes to the social network over time and to consider how this may affect career planning.
In data analytics, a social network is usually modelled as a graph. In its simplest form, a graph is just a set of vertices (also called nodes or points) and a set of edges (also called arcs or lines) between vertices. Thus, in social networks, vertices are used to represent social actors (e.g. employees) and edges are used to represent social relationships. While basic graphs allow us to model simple social networks involving one type of social relationship, more expressive graphs allow us to model very elaborate social networks. For example, multi-relational graphs allow us to model various different types of social relationships within a single graph. Likewise, attributed graphs allow us to model contextual information about social actors and the nature of their social relationships.
The above graph is a multi-relational and attributed graph involving three people (vertices): Alice, Bob and Carol. In this case, we have a number of social relationships (edges) describing an email communication network and a friendship network. The fact that we have arrows rather than lines between vertices indicates that the relationships are directed. For example, the arrow from Alice to Bob is an edge which means that Alice emails Bob, i.e. it does not mean that Alice and Bob email each other. Conversely, there is no arrow from Bob to Alice which means that Bob does not email Alice. Clearly, these people are all friends. The additional information associated with some vertices and edges are called attributes. For example, we have attributes stating that Bob is a manager and that Alice works in human resources. This information may be important when interpreting an individual’s unique social network. Similarly, we have attributes stating that Alice emails Bob significantly more than she emails Carol. This information is then useful for understanding the strength of the relationship.
There are numerous challenges for graph-based analysis of workplace social networks. Many of these are theoretical (e.g. in identifying social groups and cliques or in measuring social capital) while many are practical (e.g. where large organisations may have tens or hundreds of thousands of vertices and millions of edges). Of course, there are also many important legal and ethical concerns regarding the analysis of personal data in any form. For this reason, an important part of the DEVELOP project is to ensure that we address these concerns and that we maintain transparency with regards to how we acquire and analyse data.
Grandjean, M. (2016). Social Network Analysis Visualization. Retrieved 24 June 2016, from https://commons.wikimedia.org/wiki/File:Social_Network_Analysis_Visualization.png