Social network analysis for modeling & tuning social media website

Social Network Analysis (SNA) for Modeling and
Tuning an Online Social Media, Combating
Terrorism Community

Edward B. Rockower (MOVES)
Sean F. Everton (Defense Analysis Dept.)

MOVES Research and Education Summit 2011
Session 1
12 July 2011

Ph: 831-402-4881
Email: ebrockow@nps.edu
http://movesinstitute.org

Agenda
• Introduction & Disclaimer
• Global Community of Combating Terrorism (CT)
Professionals
• Social Media Websites
• Social Network Analysis (SNA) + Feedback
• The “Plan”
• Potential Outcomes
“If I had 8 hours to cut down a tree, I’d spend 6 hours
sharpening my saw”
-- Abraham Lincoln

4/14/2012 2

Interdisciplinary War on Terror
• Social Network Analysis: The study of network representations of
social interactions

• Web 2.0, Social Media, User-generated content, Distance
Learning, Virtual Classrooms, games, multi-media, …

•Establish, sustain a global network of CT experts &
practitioners
• Web-delivered knowledge management and
collaboration platform
• Significantly extend our reach to international
communities
• Support sustained global CT partnerships

“The greatest challenge to any thinker is stating the problem in a way that will allow a solution
-- Bertrand Russell
4/14/2012 3

Global Combating Terrorism (CT) Challenge
• A Program for Planning Future Directions for Analysis,
Modeling, and Feedback, in its initial planning.
• How to foster collaborations among ‘neighborhoods’
of CT professionals
– diverse international cultures and organizations.
• How to measure cohesiveness within the global CT
community
– from behavior on a social media website
• How to encourage formation of
– bridging ties and tight-knit clusters

4/14/2012 4

Measures of Effectiveness for
CTFP* Continuing Engagement
Existing Capital Investment Keeping Connected
Cost per interaction is dramatically reduced
~$200 Million  ~20,000 Alumni e.g. If ~ 20 - 80% use CTFP-ECCO*
~$10 K in-class cost per Fellow  ~ $200 - $900 per user/yr
~$5 K per in-person Continuing Engagement (for ~3,000 to 13,000 Alumni continually)
(for only ~ 400 Alumni/year)

Extending the Reach: Metrics “Network Effect”
% of all CTFP graduates in our database Metric: speed at which the website
% of all graduates using the CTFP ECCO website
% of new graduates regularly using ECCO
'scales up' to a critical mass
% of the total set of countries using website

Number of and growth in connectivity of sub-networks of
Metric: Value of a network >= N2
Alumni interacting on the website (but too simplistic)
Reduced numbers of disjoint networks having low
connectivity
increased number of bridging links between relatively
homogeneous sub-networks of users (inter-agency or
inter-nation)

* CTFP = Combating Terrorism Fellowship Program, ECCO = Education and Collaboration Community Online 5

Hierarchy of MOEs & MOPs
Operationalize the
CT Network

CT Capacity
Building
Build a Network
Education To
(& training) Fight the Network
Information
News
DB’s

Games

M&S
Data IT, CMS, LMS, Web, Stakeholders, Social Media Palette
SNA
6

Goals
• Instantiate in-class Alumni relationships online
• How to develop, tune, and test models of
online collective behavior for military and CT
“capacity building”.
• How to measure ‘success’
– Plethora of potential MOPs & MOEs
• How to foster ‘success’
– Survival of one community vice another!
– Observe, analyze, and tune the website to
increase important metrics
4/14/2012 7

Connecting Disjoint Networks:
‘N’ individuals, ‘m’ disjoint sub-nets
Metcalfe‟s Theorem Value of Network ~ N * (N-1) ~ N2
Value of Network ~ m * (N/m) * (N/m - 1) ~ N2 * 1/m

‘Network Effect’ vs # disjoint sub-networks
Value 

For collaboration does too much connectivity
hinder effectiveness? We need “boundaries”.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

m
8

Create Searchable* Networks
• But respect boundaries, security, privacy
• “Good fences make good neighbors”
– each Fellow controls what information they share
– anonymity is an option
Emergent
diversity creative behavior

Each-1-teach-1
boundaries accessibility
predictable Info-glut
comfort
cliques

respect uniformity

* Fellows can find & access diverse sources of data, expertise
4/14/2012 9

Aggregate and add meaning and meta-data to information/news/reports, connect people to
actionable information, connect people to people… foster teamwork and empowerment.

Repositories, Libraries, Databases, Blogs, BBS's, Discussions, News-
feeds(RSS), Websites, games, chats, Reports, articles, ...

10

Two-Mode Affiliation Networks
Users may interact:

• Directly
-- contact via private messaging, comment on a blog or article
authored by 2nd person. “Single-mode”, i.e. person  person
-- connection in Social Network is stronger, more immediate
• Indirectly
-- comment on the same article, a 3rd person‟s blog, read same
RSS feed story, contribute to same wiki article, …
-- infer a connection, raw data is “affiliation”, personblog
, person  forum, …
-- Analysis yields person  person & affiliation-i  affiliation-j
-- Relationships identified among information
types, sources, articles, books*, …

* e.g. Amazon book recommendations

4/14/2012 11

Two-Mode  One-Mode

• “Affiliation”, person  blog , person  forum, …

• Visualize connections among site „content‟
• Visualize inferred connections among users

4/14/2012 12

What is Optimal Connectivity?
• To organize to fight a common enemy?
• Can a network be too connected
– Or too disjoint?
• Is there an optimal network connectivity (i.e.
not just max or min)?
• To share information (be ‘searchable’)
• To promote useful emergent behavior

4/14/2012 13

Can a “Community” be too disjoint?
• Can a group of people be too disjoint and cliquish
-- to heed important warnings
-- to enable responsive and effective emergent
behavior
-- to self-organize against a common enemy

• An influential 1973 study* shows exactly that:
-- Boston‟s West End Community** was too
cliquish, forming highly disjoint sub-networks
-- Failed to organize against Boston‟s “urban
renewal”
 was eventually destroyed

* Mark Granovetter, 1973 American Journal of Sociology

** “… widely known because a late 1950s urban renewal project
razed” neighborhoods to redevelop the area.” -- Wikipedia
4/14/2012 14

The Strength of “Bridging Links”
• In contrast to the West End Community
• A neighboring Boston community:
– Had numerous “bridging links”
– Enabled effective organization for defense
– Successfully withstood the threat
– Survived as a viable community

4/14/2012 15

Fast-forward to the Global CT challenge of today

• Granovetter study suggests important questions
• A global context, with “neighborhoods” of CT
professionals
– Diverse languages, cultures, professions, talents,
organizations
• How to measure current state of cohesiveness of
global CT community?
• How to encourage bridging links and cohesiveness to
enable effective learning & self-organization?
•  thus materially increasing capacity-building CT
partnerships!
4/14/2012 16

Social Network Analysis
• Concepts:
– “Centrality” (betweenness, degree, …)
– Influence
– Searchable networks
– Reach
• Model site activity, Analyze & Visualize
• “feedback loop”? To optimize (what MOE’s?)
ECCO SNA
Site Tuner

4/14/2012 17

Social Media Websites
• Goals
– Not optimizing page views
– Not “monetizing” traffic, nor selling Google Ads!
– MOE’s
• Usability
• Site engagement, stickiness, page views,
• Interactions among nations, organizations, individuals, races,
cultures, …

• Techniques
– ‘Social Equity’, reputation, recognition
– Users contribute content
– Users interact directly & indirectly
4/14/2012 18

Closing the SNA Loop
Website
architecture & Server logs Site metrics
parameters MySQL tables Soc. Net Structure
MOPs, MOEs

Optimize SNA CT MOEs Visual Analytics
Model
Continuing engagement Exploratory Data Analysis
By tuning
Capacity building SNA Interpretation
site Decision Var’s
Collaboration Longitudinal Studies
Emergent Behavior?

Tune the website, spiral development
4/14/2012 19

In-class Networks  Online (& bridges)

Network Searchability
Number of Bridges

Edge betweenness
Max. “Influencers” in Network

(Venture Capital Network)

4/14/2012 20

Thank You!

Questions?

Comments?

"Judge a man by his questions rather than by his answers"
-- Voltaire

21

Backup slides, materials
• Longitudinal Study of Venture Capital Network

4/14/2012 22

Longitudinal Studies
• Time evolution of networks
• Feedback to site architecture
• “time scales”
– For on-site networks to evolve
– For feeding back onsite behavior/analyses to site
architecture
– For users to respond to changes
– Identify “emergent behavior”
• Overall MOE’s for “continuing engagement”…
4/14/2012 23

CTFP: Combating Terrorism
Fellowship Program
Short- and long-term Combating Terrorism programs for
international students at 5 schoolhouses and 6 regional
centers around the world.

24

What kinds of usage data?
• Build:
 effective collaboration,
 knowledge & data -filtering, -ranking, and -fusion.
 supportive professional learning networks
 social structures (nets) through computer-mediated
interactions?

• Website features:
 wiki, messaging, user-ratings, and Forums, personalization
 generate that data help us learn how to foster effective CT
collaboration and learning community online.

CTFP ECCO Project
Needs, Gaps, Solutions

Needs Gaps
Build and strengthen a global Limited reach to international CT
network of CT experts and experts and practitioners, and
practitioners committed to
Limited, short-term and intermittent
supporting US efforts against
meaningful contact and collaboration
terrorists and their organizations
using traditional means
“The worldwide CT network does Solutions
exist! And all this thanks to you!”
--CISA Romanian Alumnus Establish and sustain a global network
of CT experts and practitioners using a
web-delivered knowledge management
and collaboration platform that will
significantly extend our reach to
international communities and support
sustained global CT partnerships
Create the networks
that fight the
networks

CTFP Education, Collaboration Community Online Project
Addressing the Gaps

Extending the Reach Legitimate Environment
We can reach any CT practitioner CTFP has created a legitimate
with an internet environment for international CT
connection, anywhere in the world collaboration with courses and outreach

Keeping Connected We extend this environment where
international exchange can take place
Global on-demand, long-term and without raising red flags
sustained meaningful contact using
social media web tools through online Proven Technologies
knowledge
sharing, collaboration, training and Accepted, established and matured in
education large private corporations and
government agencies (e.g., State
Department, WhiteHouse.gov)
Social media web technologies are
high leverage force multiplier tools for
global connectivity

Dynamic Network Processes

Key Analysis Data:
• Group formation and dissolution
• Opinion consolidation, polarization
• In-group vs. inter-group connectivity
• Presence of bridging ties and groups
• Hierarchy vs. global reachability

Foster efficient teamwork, information sharing and knowledge fusion, link
together individuals, groups with interests, expertise, information:
 effectively solve shared problems.

Visualization of large interaction network in Pajek

Talking Points for MOEs
Talking Points: Measures of Effectiveness (MOE's) & Measures of Performance (MOP's)

Ed Rockower February 28, 2010 DRAFT for discussion

MOE: "a quantitative measure of how well a System is accomplishing what our customer needs to
accomplish"
MOP: "a quantitative measure of how well individual sub -systems, components, or processes are
performing, as needed for the successful functioning of the System being developed and deployed"
Metrics: MOE's, MOP's, Figures of Merit (FOM's), Measures of Outcome (MOO's), ...
"Technical Performance Measures" (TPM's): the important metrics with high visibility being measured,
recorded and reported to the customer on a regular basis
===================================================================

CTFP Alumni:
1. % of all CTFP graduates with which we've established contact, i.e. are in our database
with at least contact information, current job function
2. % of all graduates who regularly use the CTFP Virtual University/Commu nity website
3. % of new graduates introduced to, and regularly using our website
4. % of the total set of countries with which we've established contact and introduced to
the website
5. Satisfaction ratings of users of our website
6. Number of and growth in connectivity of sub-networks of Alumni interacting on the
website
a. "6 degrees of separation" effect (cf measures LinkedIn uses for "your network")
7. Number of Alumn's participating in the yearly meeting in Germany, etc.
8. Number and usage of refresher courses created and deployed specifically for the
website
9. ...
CT Taxonomy of Expertise and Experience
1. measure completeness of taxonomy, and % of users who've identified their expertise
2. usage of the taxonomy data for individuals for team -formation, recruiting, and other
collaborative teams with appropriate expertise and diversity
3. reduced numbers of disjoint networks having low diversity according to measures from
the Taxonomy
4. increased number of bridging links between relatively homogeneous sub-networks of
users
5. reduced number of disjoint networks
6.
Information and Traffic
1. Measures of the 'age' of information in the website (timeliness and currency of
information)
2. Measures of number of users active in managing their own RSS feeds.
4/14/2012 Numbers of messages posted and numbers of blogs, chats, and other synchronous and
3. 29
asynchronous interactions among users.

Venture Capital Network

4/14/2012 30

Social network analysis for modeling & tuning social media website

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Social network analysis for modeling & tuning social media website (20)

Recently uploaded (20)

Social network analysis for modeling & tuning social media website

Editor's Notes