Social Network Analysis Workshop This talk will be a workshop featuring an overview of basic theory and methods for social network analysis and an introduction to igraph. The first half of the talk will be a discussion of the concepts and the second half will feature code examples and demonstrations. Igraph is a package in R, Python, and C++ that supports social network analysis and network data visualization. Ian McCulloh holds joint appointments as a Parson’s Fellow in the Bloomberg School of Public health, a Senior Lecturer in the Whiting School of Engineering and a senior scientist at the Applied Physics Lab, at Johns Hopkins University. His current research is focused on strategic influence in online networks. His most recent papers have been focused on the neuroscience of persuasion and measuring influence in online social media firestorms. He is the author of “Social Network Analysis with Applications” (Wiley: 2013), “Networks Over Time” (Oxford: forthcoming) and has published 48 peer-reviewed papers, primarily in the area of social network analysis. His current applied work is focused on educating soldiers and marines in advanced methods for open source research and data science leadership. More information about Dr. Ian McCulloh's work can be found at https://ep.jhu.edu/about-us/faculty-directory/1511-ian-mcculloh

1. ep.jhu.edu
11100 Johns Hopkins Road
Laurel, MD 20723-6099
Social Network Analysis (SNA)
15 May 2018
Ian McCulloh, Ph.D.
Parson Fellow, Bloomberg School of Public Health
Senior Lecturer, Whiting School of Engineering
Senior Scientist, Applied Physics Laboratory
imccull4@jhu.edu
1

2. ep.jhu.eduep.jhu.edu
1. PhD Computer Science/Social Networks, Carnegie Mellon University.
2. 50+ peer reviewed papers.
3. Author of Wiley’s textbook on Social Network Analysis.
4. Current research: social media and the neuroscience of persuasion.
5. 20 years in the US Army – Network targeting.
6. Organizing the North American Social Network (NASN) Conference
in DC 27-30 NOV 2018
Background

3. ep.jhu.eduep.jhu.edu
Getting Started
https://www.rstudio.com
https://ep.jhu.edu/programs-and-courses
•605.633—Social Media Analytics
•605.634—Crowdsourcing and
Human Computation
•605.632—Graph Analytics

4. ep.jhu.eduep.jhu.edu
• Study of sociology
• Organizational behavior (leadership, management)
• Influencing groups (public health, propaganda, marketing)
• Increasing engagement with social media (computer science)
• Cool algorithms/heuristics (math, computer science)
Why Social Networks?

5. ep.jhu.eduep.jhu.edu
• Study of sociology
• Organizational behavior (leadership, management)
• Influencing groups (public health, propaganda, marketing)
• Increasing engagement with social media (computer science)
• Cool algorithms/heuristics (math, computer science)
Why Social Networks?

6. ep.jhu.eduep.jhu.edu
Study of Sociology
Moreno (1934) Sociometry
• 2nd grade classroom in the US
• Social physics
• The birth of social networks
Network Science
• 1999 Albert-Barabasi Scale-free networks
• 2004 NRC Report, Army interest
• Often lacks empirically grounded social theory that has been
developed over 65years – communities are merging

7. ep.jhu.eduep.jhu.edu
• Study of sociology
• Organizational behavior (leadership, management)
• Influencing groups (public health, propaganda, marketing)
• Increasing engagement with social media (computer science)
• Cool algorithms/heuristics (math, computer science)
Why Social Networks?

8. ep.jhu.eduep.jhu.edu
Organizational Behavior
• Who is the most powerful person in this network?

9. ep.jhu.eduep.jhu.edu
Organizational Behavior

10. ep.jhu.edu
Organizational Spectrum
23 May 1
Agility Efficiency
• Well-defined task/purpose
• High repetition
• Standards/quality control
• Reduce waste/minimize costs
• Hierarchy/supervision
• Unity of leadership
• Social interaction = distraction
• Pioneering/no defined task
• High novelty/not done before
• Creative/diversity of ideas
• Innovation/maximize new ideas
• Organic/collaboration
• Flat structure/many bosses
• Social interaction = value

11. ep.jhu.edu
Measuring Organizational Efficiency
1
Four Properties:
1. Connected
2. Hierarchic (no reciprocity)
3. Efficient (no cross-talk)
4. Least Upper Bound
1 −
𝑉
𝑛 𝑛 − 1 /2
1 −
𝑉
max 𝑉
1 −
𝐿
max 𝐿
1 −
𝑈
max 𝑈
V = reciprocal link
L = #links above nk-1
U = #pairs without LUB

12. ep.jhu.edu
What do we seek in an agile network
1
AGILE
• Knowledge Exchange
• Resource Exchange
• Reduced Management Overhead
• Innovation
• Cognitive Diversity
• Inclusion
• Time (social opportunity)
EFFICIENT
• Connected
• Efficient
• Hierarchic
• Least Upper Bound
It is not clear that Agile is the opposite of Efficient.
We wish to maximize connectivity and minimize efficiency and hierarchy.

13. ep.jhu.edu
What are we missing?
1
Network “horizons” suggest the likelihood of knowledge/resource
exchange between actors approaches 0, as distance >3
Relationships take time and resources
How many meaningful conversations?
• Software developers 5-6/day
• Managers 15-20/day
500-2000 Facebook friends!!!

14. ep.jhu.edu
How to Create Truly Agile Networks
• Minimize the diameter of the network
• Minimize the average degree of actors (a.k.a. density)
• Maximize cognitive diversity
Diameter = 2
Density = 0.4
Diameter = 1
Density = 1.0
Diameter = 2
Density = 0.5
Need to explore tradeoffs in diameter and density

15. ep.jhu.eduep.jhu.edu
• Study of sociology
• Organizational behavior (leadership, management)
• Influencing groups (public health, propaganda, marketing)
• Increasing engagement with social media (computer science)
• Cool algorithms/heuristics (math, computer science)
Why Social Networks?

16. ep.jhu.eduep.jhu.edu
Reasoned Action Theory
𝐵 = 𝑤1 𝐴 + 𝑤2 𝐼𝑁 + 𝑤3 𝐷𝑁 ∗ 𝑤4 𝑃𝐵𝐶
• B = Behavior
• A = Salient Attitudes
• IN = Injunctive Norms
• DN = Descriptive Norms
• PBC = Perceived Behavioral Control
• wi = Weight applied to factor

17. ep.jhu.eduep.jhu.edu
Where do you get the data?
𝐵 = 𝑤1 𝐴 + 𝑤2 𝐼𝑁 + 𝑤3 𝐷𝑁 ∗ 𝑤4 𝑃𝐵𝐶
Opinion Leader
Key Influencer
Key
Communic
ator
Alters
Informational
Conformity
Normative Conformity
Network Conformity
Social
Network
Analysis

18. ep.jhu.edu

19. ep.jhu.edu

20. ep.jhu.eduep.jhu.edu
Community Detection
• Cohesive clustering – community detection
- Newman grouping
- Louvain grouping
- Truss grouping
• Intuitively satisfying clusters
• Allows identification of distinct social groups

21. ep.jhu.edu

22. ep.jhu.eduep.jhu.edu
“When is a tourniquet applied to a
neck wound?”
• When it is a vein or artery.
• If it is spurting blood.
• Never
McCulloh, I. (2013). Social Conformity in Networks. Official Journal of the International Network for Social Network Analysts
Network Conformity Experiment

23. ep.jhu.eduep.jhu.edu
Centrality

24. ep.jhu.eduep.jhu.edu
Degree
Keyplayers-Pos.Between
Closeness

25. ep.jhu.eduep.jhu.edu
Different Leaders for Different Stages
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10
PercentAdopters
Time
Degree
Betweenness
Closeness

26. ep.jhu.eduep.jhu.edu
• Study of sociology
• Organizational behavior (leadership, management)
• Influencing groups (public health, propaganda, marketing)
• Increasing engagement with social media (computer science)
• Cool algorithms/heuristics (math, computer science)
Why Social Networks?

27. ep.jhu.eduep.jhu.edu
Social Media Analysis: Structure-Based Analytics
Social network construction using relational algebra
Let X be an association matrix of screen names by tweetID/image
Let Y be an association matrix of tweetID/image by MD5 hash
Then, YTXTXY is a hash network of images posted by the same people
Pro-ISIS
Anti-Assad
Shi’a
No Confidence
7,887 different hash values
3,583 hash shared by 2+ people

28. ep.jhu.edu

29. ep.jhu.eduep.jhu.edu
Time for the Workshop!
0
500
1000
1500
2013
2014
2015
2016
DailyDownloads
Package
igraph
sna
tnet
Daily Downloads (RStudio mirror) for
igraph, sna, and tnet
igraph (also in python)
• Social media, cluster, speed
statnet
• Statistics, longitudinal, egonet
Not compatible

30. ep.jhu.eduep.jhu.edu
• Centrality measures
• Diameter & density
• Clustering
• Social media context
• Network statistics (ERGM, SAOM)
What are the most common analytics?
It’s the social theory that gives life to analysis!
Let’s go to Rstudio!
We will use the igraph package
Python uses igraph

31. ep.jhu.edu © The Johns Hopkins University 2016, All Rights Reserved.