Social Network Analysis Concepts Applications Analysis of Social

Social Network Analysis: Concepts, Applications, Analysis of Social Structures Using Secondary Data Liaquat Hossain Sciences of Learning Winter Institute Workshop The University of Hong Kong, 14 January 2014 Email: lhossain@hku. hk

Agenda • • Introduction to Social Networks Visualising & Interpreting Social Networks Social Network Data Collection & Analysis Social Networks Correlates to Coordination – Enron Email Communication Corpus – Visualisation of Enron Communication Network • Research Framework • Methods • Results 2

Introduction to Social Networks 3

Social Network Analysis • Social Network – A set of actors and relations that hold the actors together • SNA – “The disciplined inquiry into the patterning of relations among social actors, as well as the patterning of relationships among actors at different levels of analysis (such as persons and groups)” (Breiger, 2004) 4

Growth of Social Networks • Social Networks research has been growing exponentially over the past decades Source: Otte & Rousseau (2002) 5

Distribution of Social Network Research 6

At the Beginning of the Hype Curve Source: IBM (2005) 7

Concepts: Relations and Ties • Relations – connects actors (eg. Friendship, colleagues, kinship) • Tie - a relation that is established between two or more entities from the moment information or resources are exchanged • Multiplex Tie – Connections based on many relations • Ties are characterised by content, direction and strength 8

Visualising & Interpreting Social Networks 9

Collaboration Network Source: Otte & Rousseau (2002) 10

Academic Collaboration Ties (1990 -1994) 11

Academic Collaboration Ties (1995 -1999) 12

Academic Collaboration Ties (2000 -2004) 13

Sexual relations Colorado Spring data By Gender (Morris, Moody et al. ) Male Female 14

Social Network Data Collection & Analysis 15

Whole or Sociocentric Approach • Focus is on measuring structural patterns of interactions and how the patterns explain outcomes (eg. concentration of power or resources within groups) • Actors are usually known or determined ie. network boundaries are a priori defined • A roster of names is usually needed to formulate the adjancency matrix 16

An Adjacency Matrix Seeks advice from John Ken Smith Jacky Carl Winnie Will Daren Fiona Fetch John - 0 0 0 3 0 0 0 2 3 Ken 0 - 0 0 0 1 1 0 0 0 Smith 1 0 - 0 0 2 0 1 0 0 Jacky 0 0 0 - 3 0 0 Carl 0 1 0 0 - 0 0 0 2 0 Winnie 0 0 0 3 3 - 0 0 Will 0 3 2 0 3 0 - 0 2 0 Daren 0 0 0 3 3 0 0 - 0 0 Fiona 0 0 1 0 3 0 0 0 - 0 Fetch 0 0 0 3 3 0 0 - 17

Sociocentric Network Approach • What if the population of interest is over 100 or 1, 000? • For – 30 actors there would be 435 (undirected) or 870 (directed) interactions – 100 actors there would be 4950 (undirected) or 9900 (directed) interactions • Scrutinising through long lists of names and identifying multiplex ties with actors on the list causes fatigue and recall problems (Bernard et al, 1982) 18

Egocentric Network Approach (. . cont) • “Ego” is the actor we are interested in. “Alters” are the people referred to by the ego as having a tie with • Coleman, Katz et al’s (1957) medical innovation diffusion study adopted egocentric approach from doctors – To whom did you turn to for advice and information? – With whom did you most often discuss cases in an ordinary week? – Who were the friends between your colleagues whom you saw often socially? Ego 19

Egocentric Network Approach (. . cont) • Socio-demographic attribute data: Sex, Marital Status, Education • Occupational attribute data: Age, professional associations, social associations, journals read • Relational data: Density, Centrality and Centralisation 20

Social Networks Correlates to Coordination 21

Introduction to Enron Dataset • We used the Enron Email Dataset to study the correlation between centrality and coordinative ability. • The Enron Dataset is: – A collection of over 250, 000 emails… – of about 150 employees… – from a real organisation… – over a period of 3. 5 years 22

Enron Email Corpus versions • Various instances of the email corpus: – Federal Energy Regulatory Commission (619, 449 emails, 158 employees) – Stanford Research Institute then rectified data integrity problems – Cohen (CMU) made the dataset public (517, 431 emails, 151 addresses) – Corrada-Emmanuel (250, 484 emails, 149 users) – Shetty and Adibi (252, 759 emails, 151 users) 23

Visualisation of Enron Communication Network 1997 -2002 24

Preliminary Results of Enron Communication Network (1997 -98) 25

Preliminary Results of Enron Communication Network (1999) 26

Preliminary Results of Enron Communication Network (2000) 27

Preliminary Results of Enron Communication Network (2001) 28

Preliminary Results of Enron Communication Network (2002) 29

Research Framework 30

Methodology summary – The methodology of the study involves four research phases: • extraction and cataloguing of coordination key phrases • calculation of coordination score bounded by project scope • construction of social network matrices using the centrality measures • hypothesis testing of the association between network centrality and coordination 31

Text Mining for Coordination 32

Key Phrase Extraction 33

Calculating Coordination Weights 34

Text Mining and Coordination Score A text-mining application was created to deploy the model to calculate coordination score from the Enron dataset. 35

Social Network Matrices and Centrality • Measurement: – Degree – Closeness – Betweenness • Directional Analysis: – In-Centrality – Out-Centrality UCINET 6 Output of Closeness 36

Project Based Coordination Results • Coordination and centrality across three projects The coordination and centrality scores are ranked. 6 of the 8 cases ranked equally. 37

Organisational Position and Coordination Employee title and role, along with coordination and centrality scores 38

Mapping average coordination against position in organisational hierarchy 39

Results Summary 40