3. What is it?
• Social network analysis is a toolkit of approaches built on the
fundamental idea that a social relationship between two
people can be conceptualised as a link (‘edge’ or ‘tie’)
between two people (‘nodes’, ‘vertices’ or ‘actors’)
• Depending on the relationship, this can be directed or
undirected
• One mode or two mode networks
• Advantages of being able to visualise previously obscure
relationships, and use graph theory to model processes
Node NodeNode
4.
5. Frequently used metrics
• Network size: degree
• If directed, this can be considered in terms of in-degree and
out-degree
• Typically follows a power law distribution
Albert-Laszlo Barabasi, Linked: The
New Science of Networks.
6. • But how connected are the nodes within a
network?
• Density = proportion of possible connections
which do exist
• A clique = a set of nodes in which all
possible connections exist
• Smallest clique = a triad
• Clustering coefficient, community detection
methods
Frequently used metrics
7. Frequently used metrics
• Positions between communities are important – shortest paths
• Betweenness centrality, structural holes, brokerage roles
8. Origins
• Origins date back to early
20th century Sociology
• “[SNA] itself is neither
quantitative nor qualitative,
nor a combination of the
two. Rather, it is structural”
(Carrington, 2014, p.35)
• Interpretation of networks
depends on goals and
epistemology of studies
Image source: Bbuuggzz
https://en.wikipedia.org/wiki/File:15th_Century_Flore
ntine_Marriges_Data_from_Padgett_and_Ansell.pdf
9. Classic studies: Milgram’s small
world
• Sought to determine the average path length between two
nodes in a population
• Randomly selected people in Nebraska and Kansas
• Had to forward information to someone they knew personally,
with the goal of it reaching a target contact in Boston,
Massachusetts.
• 64 of 296 letters reached destination
• Hops ranged from 1 to 10; average number was six
• Origin of the phrase ‘six degrees of separation’
10. Classic studies: Granovetter’s
jobseekers
• First published in 1973
• Interviewed 100 people to find out how they used their social
networks to get new jobs
• ‘Strong ties’ are close friends, highly connected to ego and
often each other; ‘weak ties’ are less frequently met,
acquaintances
• Acquaintances more frequently the source of information
leading to new jobs; weak ties more likely to provide novel
information
• ‘The strength of weak ties’
11. Classic studies: Burt’s
brokerage
• Elaborated on links between structural characteristics of networks
and links to social capital
• Social capital: “networks together with shared norms, values and
understandings that facilitate co-operation within or among
groups” (OECD definition)
• ‘Structural holes’ as gaps between communities which could be
usefully exploited
• ‘Brokers’ as key nodes which mediate flow of information between
otherwise unconnected communities
• Nodes which are positioned between different communities can
have advantages and disadvantages in terms of social capital
12. SNA in the era of Big Data
• Networks everywhere?
• But how valid are the links?
• Automated network extraction does not account for
context.
• Unlike genes or hyperlinks, people have agency.
• E.g. are all your Facebook friends equally important to you?
• -> Importance of mixed methods to validate understanding
13. Some considerations
• Which level of network to focus on?
• Directed or undirected?
• One-mode or two-mode?
• Can learn from small networks too.
• If using statistical tests, bear in mind that many metrics don’t
follow a normal distribution (e.g. power laws).
• How relationships (edges) are defined, and how
confident you can be in the accuracy of what they
represent, is essential.
15. Benefits of using Gephi
• It’s free
• Works on both PCs and Macs
• Various plugins are available – e.g. export as web pages, fix
nodes to geographical co-ordinates
• Active community for support online
• Relatively user friendly
• Attractive visualisations
• Can export in various formats to other packages - .gexf or
.gml as a good lingua franca
16. What Gephi needs
• An edges table
• A nodes table (optional)
• You can enter this manually, or import data as .csv files
• An edges table is a .csv file with two columns: ‘source’ and
‘target’
Editor's Notes
Plus excellent MOOCs – Easley & Kleinberg on edX, courses from Stanford, Michigan and Pennsylvaina on Coursera.
Although I’m saying social network analysis, it has also been applied to other settings such as networks of proteins, or the internet, and some of the core characteristics of networks hold up (more on that later). By thinking about relationships in network terms, it gives us a way of comparing and modelling a range of different networks.
Examples of 50-node undirected random graphs generated using Gephi to illustrate a range of network densities from zero to one. (0, 0.05, 1)
But in reality networks aren’t uniform – you get groups of more highly connected people
A corrollary of the idea that networks aren’t equally distributed but have highly clustered communities within them is that individuals will be more or less embedded within particular groups
Now will introduce some of the classic studies where the commons metrics have been demonstrated
E.g. Families in Florence – small sample but very illuminating
Stanley Milgram
A series of experiments in the latter half of the 1960s – results first published in 1967
Average path length now arguably shorter – on Facebook it is 4
Mark Granovetter
Doctoral research
List people who put them in touch with information which led to a new job, and categorise them in terms of whether they see them frequently (at least once a week), occasionally (more than once a year but less than twice a week; rarely (once a year or less).
Related to Granovetter’s work…
Social capital as the social value in network structures – underpinned by concepts like reciprocity and trust