Transcript exclusive of “Anonymous” and self-replies

KM4DEV Discussion Group
Social Network Analysis 2012
Prepared by:
Correct at:
A/Prof Graham Durant Law CSC, PhD
6:00 am AEST on the 24th of June 2012
Caveat Emptor

The diagrams presented in this document are derived from just over 11 years of
data provided by KM4Dev in an XML format.

Data was dirty and required substantial cleaning to remove identified
pseudonyms, aliases, and duplicate names, as well as non-printing characters.
HyperEdge cannot guarantee that all pseudonyms and aliases have been
removed.

"Anonymous" posted about 18% of all posts. It was necessary to remove
"Anonymous" from the dataset because HyperEdge could not be sure
"Anonymous" was a single person.

Removing identified pseudonyms, aliases, and duplicate names, and
"Anonymous" leaves 703 identified individuals in the network. Collectively,
these people comprise the node-set for the public bounded or contained
network, for which activity and various network measures have been applied.
Copyright © 2012: HyperEdge Pty Ltd
2
Caveat Emptor (continued)

There is no way to measure the private interactions or shadow network from the
data supplied.

Mathematical approaches to network analysis tend to treat the data as
‘deterministic’. That is, measurements are viewed as an accurate reflection of
the ‘real’ or ‘final’ or ‘equilibrium’ state of the network. Clearly this not the case
as the network is constantly changing and evolving. Local knowledge should
therefore be applied to the interpretation of data.

The diagrams presented in this document are drawn from filtered data. Data can
be filtered in multiple ways resulting in differing diagrams. It is also possible to
present other forms of diagrams from the same filtered data.

Occasionally the results appear contradictory. Local knowledge should therefore
be applied to the interpretation of the maps.
Copyright © 2012: HyperEdge Pty Ltd
3
Mapping Objectives

HyperEdge was asked to conduct a social network analysis in order to allow
the KM4Dev Core Group, and the wider community, to:
– better understand KM4Dev in terms of issues such as identity,
relationships, function and role;
– better understand the relationship of SIWA and SA-Ge to each other and
the main community;
– better understand where KM4Dev sits in comparison with other networks
to help KM4Dev appreciate where they are unique, where they are in terms
of a traditional life-cycle, and how they might evolve; and
– implement practical methods to support the growth and enrichment of the
KM4Dev community.
Copyright © 2012 – HyperEdge Pty Ltd
4
A Quick
Social Network Analysis Lesson
Copyright © 2012: HyperEdge Pty Ltd
5
Typical Measures

Ties (links): in ties and out ties
represent the number of connections to
and from a node.

Density: the percentage of connections
that exist out of the total possible that
could exist.

Distance: degrees of separation or the
diameter of a network.

Reciprocity: the number of bidirectional links expressed as a
percentage.

Centrality: the extent to which a
network is organised around one or
more central nodes.
Copyright © 2012 – HyperEdge Pty Ltd
6
Organisational Dynamics
Providers and Seekers
Transmitters and Receivers
degree centrality
closeness centrality
Reveals how much activity is
going on and who are the
most active members by
counting the number of direct
links each person has to
others in the network.
Does not necessarily describe
power or influence.
Copyright © 2012 – HyperEdge Pty Ltd
People at the centre of the
network:
• are the connector or hub of
the network,
• may be in an advantaged
position in the network.
• are usually less dependent
on other individuals.
• are often a deal maker or
broker.
Highlights people with the shortest paths to other people, thus
allowing them to directly pass on and receive communications
quicker than others in the organisation.
Is strongly correlated with organisational influence if the
individual is a skilled communicator.
These individuals are often network brokers. They are often the
‘pulse-takers’ of the organisation.
7
Organisational Dynamics (continued)
Bridges
Influencers
(brokers and gatekeepers) - betweenness centrality
eigenvector centrality
Reveals individuals who:
• connect disparate groups
within the network.
• hold a favoured or
powerful position in the
network.
• have great influence over
what is communicated
through the network.
• act as intermediaries
Copyright © 2012 – HyperEdge Pty Ltd
Identifies the bridges within
the network. They may act as
the true gatekeeper deciding
what does or does not get
passed through the network,
or as the “third who benefits”
by passing information to
others to secure advantage.
.
Measures how well connected a person is and how much direct
influence they may have over the most active people in the
network
Measures how close a person is to other highly connected
people in terms of the global or overall makeup of the network
Is a reasonable measure of “network positional advantage”
and/or perceived power.
8
Brokerage Roles (B is the Broker).
Coordinator - a person who brokers connections within the
same group or team.
A
B
C
Gatekeeper - a person who transmits information and other
resources to the same group or team from sources
external to that group or team.
A
B
C
Representative - a person who transmits information and
other resources from their group or team to an external
group or team.
A
B
C
Consultant - a person who intermittently takes the central
lead by connecting others in the same group or team, but
who belongs to another group or team.
A
B
C
Liaison - a person who transmits information and other
resources from one group or team to another group or
team, whilst themselves belonging to a different group or
team.
A
B
C
Copyright © 2012 – HyperEdge Pty Ltd
9
Some Basic Activity Statistics
Copyright © 2012: HyperEdge Pty Ltd
10
Total Posts per Annum
(inclusive of “Anonymous” and self-replies)
Note: 2000 and 2012 are incomplete years
Copyright © 2012: HyperEdge Pty Ltd
11
Total Posts per Annum
(exclusive of “Anonymous” and self-replies)
Note: 2000 and 2012 are incomplete years
Copyright © 2012: HyperEdge Pty Ltd
12
Total Posts by Month
(inclusive of “Anonymous” and self-replies)
Copyright © 2012: HyperEdge Pty Ltd
13
Total Posts by Month
(exclusive of “Anonymous” and self-replies)
Copyright © 2012: HyperEdge Pty Ltd
14
Total Posts by Day
(inclusive of “Anonymous” and self-replies)
Copyright © 2012: HyperEdge Pty Ltd
15
Total Posts by Day
(exclusive of “Anonymous” and self-replies)
Copyright © 2012: HyperEdge Pty Ltd
16
Engagement Pattern - Reciprocated Posts
(exclusive of “Anonymous” )
This measure means the link is reciprocal: that is, there is an arrowhead at both ends. It does not measure the weight of the link:
that is, the number of reciprocated posts.
Copyright © 2012: HyperEdge Pty Ltd
17
Participating People and Posts by Time
(exclusive of “Anonymous”)
Dunbar’s Numbers are an indicator of meaningful relationships and the maximum effective number of people in a network. The
usually accepted number is 153. There is an mega-band number of around 500, and an upper limit of about 1,500 – see Dunbar,
R 2010, How many friends does one person need? Dunbar's number and other evolutionary quirks., Faber and Faber, London.
Copyright © 2012: HyperEdge Pty Ltd
18
Unique Participants in the Network
(Dunbar’s and Wellman’s Numbers)
Wellman’s Number
Dunbar’s Number
Dunbar, R 2010, How many friends does one person need? Dunbar's number and other evolutionary quirks.,
Faber and Faber, London.
Wellman, B 2011, 'Is Dunbar's Number up?', British Journal of Psychology, pp. 1-3.
Copyright © 2012: HyperEdge Pty Ltd
19
90-9-1 Community Participation Heuristic

A 2010 study by Dr Michael Wu, using
ten years of data from more than 200
online communities, found that:
– 90% of all users are “lurkers” who
don’t actively contribute.
– 9% of all users are “occasional
contributors” providing less than
50% of the content.
– 1% of all users are “hypercontributors” providing greater
than 50% of the content.

Using this heuristic the predicted size
of the KM4Dev discussion group is
2,420 people.
http://lithosphere.lithium.com/t5/Building-Community-the-Platform/The-90-9-1-Rule-in-Reality/ba-p/5463
Copyright © 2012: HyperEdge Pty Ltd
20
Comments

Removing “Anonymous” and self-replies has not changed the basic pattern of
network activity.

Reciprocal posts appear low given this is a “knowledge exchange” network. It
is perhaps indicative of a “knowledge seeking” organisation rather than a
“knowledge sharing” organisation.

For the previous four years the active or public network appears to have
stabilised around Wellman’s number.
– The literature suggests growth beyond this number may be difficult, unless
a number of smaller sub-groups or special interest groups are formed, or
the group is actively nurtured.

The predicted current size of the KM4Dev Discussion Group is 2,420 people,
comprising 2,178 lurkers, 218 occasional contributors, and 24 hypercontributors.
Copyright © 2012: HyperEdge Pty Ltd
21
The Growth of the Network
Copyright © 2012: HyperEdge Pty Ltd
22
The Growth of the Network - 2001
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network.
Copyright © 2012: HyperEdge Pty Ltd
23
The Growth of the Network - 2002
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
24
The Growth of the Network - 2003
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
25
The Growth of the Network - 2004
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
26
The Growth of the Network - 2005
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
27
The Growth of the Network - 2006
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
28
The Growth of the Network - 2007
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
29
The Growth of the Network - 2008
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
30
The Growth of the Network - 2009
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
31
The Growth of the Network - 2010
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
32
The Growth of the Network - 2011
(exclusive of “Anonymous” and self-replies)
1. All people who participated from 2001 to 2011 are in the network
2. This diagram does not include the previous years links.
Copyright © 2012: HyperEdge Pty Ltd
33
Complete Network Metrics
(exclusive of “Anonymous” and self-replies)
People
703
Unique Links
2,981
Reciprocated Links
1,770
Total Links
4,751
Maximum Geodesic Distance (Diameter)
8.000000
Average Geodesic Distance
2.931082
Graph Density
0.007325
1. Geodesic distance means theoretically everyone in the network can reach each other in a maximum of 8 steps. Most can reach
everyone in three steps: that is, there are three degrees of separation in the network.
2. Graph density might be low given this is a “knowledge exchange” network and less than 1% of all possible links are present.
Copyright © 2012: HyperEdge Pty Ltd
34
Comments

Mathematical approaches to network analysis tend to treat the data as
‘deterministic’. That is, measurements are viewed as an accurate reflection of
the ‘real’ or ‘final’ or ‘equilibrium’ state of the network. Clearly this not the case
as the network is constantly changing and evolving.

Noting the caution above, the network appears to have stabilised and reached
an equilibrium state, with more or less the same participants.

Given the number of people in the network and its obvious complexity, a way
of attributing and partitioning the network for further analysis is necessary.

One way of attributing data might be to use posting frequency or posting
patterns.

Data and visualisations from this section, coupled with the previous section,
suggest the a logical data partition is the 1st of January 2008 to the 31st of
December 2011.
Copyright © 2012: HyperEdge Pty Ltd
35
Attributing the Complete Network
Copyright © 2012: HyperEdge Pty Ltd
36
Gloor’s Contribution Index
messages sent – messages received
messages sent + messages received
If an individual only sends messages and receives none then their contribution index is +1.000
If an individual only receives messages and sends none then their contribution index is -1.000
If the communication behaviour is balanced then the contribution index is 0.000
Sender +1
Envoi
Expediter
Escort
Contribution
Index
Contribution
Frequency
Expert
Receiver -1
Gloor, P 2006, Swarm creativity: Competitive advantage through collaborative innovation networks, Oxford University Press, Oxford.
Copyright © 2012: HyperEdge Pty Ltd
37
Frequency of Sending and Receiving
(exclusive of “Anonymous” and self-replies)
Key Participant
Groups
High Posting Frequency >= 59 posts being two standard deviations from the mean
Medium Posting Frequency >=12 but <59 posts being one standard deviation from the mean.
Low Posting Frequency <12 posts.
Copyright © 2012: HyperEdge Pty Ltd
38
Frequency of Sending and Receiving
(exclusive of “Anonymous” and self-replies)
1. Red circles are people with high
frequency activity >= 59 posts being two
standard deviations from the mean.
2. Blue squares are people with medium
frequency activity >=12 but <59 posts
being one standard deviation from the
mean.
3. Green triangles are people with low
frequency activity <12 posts.
4. Noting the list is incomplete because of
space reasons, the most active people in
order are:
Copyright © 2012: HyperEdge Pty Ltd
a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
l.
m.
n.
o.
p.
q.
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
39
[email protected]
Role Types by Contribution Index
(exclusive of “Anonymous” and self-replies)
Key Participant Groups
No Role people are low frequency senders and receivers with a contribution index between -0.499 and +0.499.
Experts are go to people, but are low frequency receivers with a contribution index between -0.500 and -1.000.
Envois are low frequency senders with a contribution index between +0.500 and +1.000.
Escorts are medium frequency senders and receivers, with a contribution index between -0.499 and +0.499.
Expediters are high frequency senders, with a contribution index between 0.000 and 1.000.
Copyright © 2012: HyperEdge Pty Ltd
40
Role Types by Contribution Index
(exclusive of “Anonymous” and self-replies)
No Role
Envoi
Expert
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
1. The 14 people with the “highest” index for their role type have been selected. For the “No Role” people this
means an index of -1, and for the “Expert” people it means an index of 0.
2. This list is incomplete because of space reasons. The complete list could be provided as an annex to the report.
Copyright © 2012: HyperEdge Pty Ltd
41
Role Types by Contribution Index
(exclusive of “Anonymous” and self-replies)
Escort
Expediter
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
1. The 14 people with the “highest” index for their role type have been selected. For the “Escort” people this means
an index of 0, and for the “Expert” people it means an index of 1.
2. This list is incomplete because of space reasons. The complete list could be provided as an annex to the report.
Copyright © 2012: HyperEdge Pty Ltd
42
Engagement Pattern by Role Type
(exclusive of “Anonymous” and self-replies)
No Role
Envoi
Escort
Expert
Expediter
1. The links inside the “circles” are posts between like roles. Note there are no posts between Experts.
2. The thicker curves linking groups are consolidated exchanges between groups. They do not show frequency, or links from one
individual to another.
3. Note the relative density in the Escort and Expediter groups.
Copyright © 2012: HyperEdge Pty Ltd
43
Engagement Pattern by Role Type
(exclusive of “Anonymous” and self-replies)
Copyright © 2012: HyperEdge Pty Ltd
44
Comments

There are about 116 “key” active members in the network.
– These people are either Escorts or Expediters.
– 116 falls within the range of a Dunbar number, with room for marginal
growth.

There are 21 Expediters. This is very close to Wu’s Heuristic prediction of 24
hyper- contributors.

Gloor’s index provides a useful way of attributing the network, and other
potential way of partitioning it.
Copyright © 2012: HyperEdge Pty Ltd
45