Information Flow and the Locus of
Influence in Online User Networks:
The Case of iOS Jailbreak *
Nitin Mayande & Charles Weber
Department of Engineering and Technology Management
Portland State University
12
thOUI Conference, Boston, Mass., USA
July 28 – 30, 2014
*
This material is based upon work supported by the National Science Foundation under Grant No. 0822062 (Enabling Timely Revolutions in Organizational Performance).Outline
•
Influence in online user communities
•
Motivation for Study
•
Extant theory
•
Quantitative Empirical Study
–
(setting, measures, criterion)
•
Findings
•
Implications
Understanding the Nature of Influence Online
Networks
•
Is becoming increasingly important
–
For the first time in history, Americans are expected to spend
more time online this year than watching television.
–
Many companies are reallocating their marketing resources
•
to specifically target social networking platforms
•
such as Facebook and Twitter.
•
For example, Proctor and Gamble spend between 25% and 35% of its
marketing budget on social media.
1–
They are not just doing this to push their marketing messages;
–
they also want to entice the user communities within their
user networks to participate in the innovation process.
•
(Mayande and Jetter, 2010)
1
http://online.wsj.com/article/SB10001424127887323681904578641993173406444.html
accessed on 04/01/2014.
Mayande & Weber -- OUI 2014 3 7/25/2014
Motivation for Study
•
Yet, the influence that individuals wield within
online user networks has yet to be quantified.
•
What factors determine influence within
online user networks?
Extant Theory on Social Networks
(e.g., Granovetter, 1973; Coleman, 1988; Powell, 1990; Burt, 1992; Podolny, 1993; Putnam, 1995)
•
Not particularly useful in characterizing the loci of influence in
online networks.
–
Primarily based on connectivity,
–
Derived from phenomena that are subject to real-world
constraints.
•
We cannot assume that it can explain the behavior of online
networks,
–
which are highly dynamic open systems
–
to which traditional constraints such as physical distance
(Allen,
1977)
and geography do not apply.
Mayande & Weber -- OUI 2014 5 7/25/2014
Extant theory does not show how influence within a
network can depend upon control of information
Quantitative Empirical Study
•
Investigates the nature of influence in an
online user innovation network
•
Characterizes how influence in this network
depends upon information flow.
•
Statistically analyzes the Twitter traffic
–
that transpired between the network’s members
Setting:
Apple’s Jailbreak Community
(
http://en.wikipedia.org/wiki/Jailbreak_(iPhone_OS)
)
•
Members collaborate online
–
to remove restrictions on iOS,
–
Apple’s operating system for mobile devices.
•
Observed throughout September 2012.
–
Three significant events took place during that month:
1. a YouTube video about Jailbreak became available on Sept. 1;
2. a press release about a jailbreak of iOS6 occurred on Sept. 13,
3. Apple released the iPhone5, which featured iOS6, on Sept. 21.
Mayande & Weber -- OUI 2014 7 7/25/2014
Measuring Information Flow
(See also Appendix A)
•
The interaction between members of an online
network represents a form of serial propagation.
•
(Mayande and Weber, 2012)
•
Number of paths
–
acts as a measure of the extent to which information
can spread virally within the network,
•
Number of geodesics
(shortest paths)
–
acts as a proxy for the speed at which information
spreads within the network.
Criterion for Influence:
Eigenvector Centrality (EVC)
(Bonacich, 1972)
•
Chosen because,
–
in contrast to well-known centrality metrics from graph theory
–
such as
degree
,
betweenness
and
closeness
•
(Freeman 1977, 1979),
–
it measures the relative importance of particular nodes.
•
(Mayande, Weber & Jetter, 2011)
•
Assumes that,
ceteris paribus
,
–
connections to high-degree nodes are more important
–
than connections to low-degree nodes.
–
(For example, Google's PageRank is a variant of the
eigenvector centrality measure.)
Mayande & Weber -- OUI 2014 9 7/25/2014
Key Finding 1
•
User networks on open platforms like Twitter
are activated by or in anticipation of specific
events.
•
Once the event passes the network dissipates.
•
People within the user network interact
•
even if they are not friends with or followers
of each other;
•
they simply share a common interest in a
Trend Chart of Normalized Data of Twitter Traffic
for iOS Jailbreak User Network
People--Total number of people involved in conversation about Jailbreak (daily max. = 27788 people) Tweets--Total number of tweets that form the Jailbreak conversation (daily max. = 30362 tweets) Network Size--The biggest connected network of people within the Jailbreak conversation
(daily max. = 2334 nodes)
Network Tweets—Total number of tweets that form the biggest network (daily max. = 3116 people) Graph Diameter--The longest geodesic (shortest path) in the network (daily max. = 29 ties)
(Graph diameter: everybody in the network has access to everybody else in the network.)
Mayande & Weber -- OUI 2014 11 YouTube Video Release
about Jailbreak
Press Release about Jailbreak of iOS6
Apple releases iPhone5 with iOS6.
Key Finding 2
•
Eigenvector centrality
–
correlates
well
with the number of paths,
–
but
not
with the number of shortest paths
(geodesics).
EVC vs. Total # of Paths
•
Daily correlation between total number of paths and eigenvector centrality
(TPvsEVP) for September 2012.
•
Statistical significance is given by p1.
Mayande & Weber -- OUI 2014 13 0.00 0.20 0.40 0.60 0.80 1.00 1 -2 SE P 2 -3 SE P 3 -4 SE P 4 -5 SE P 5 -6 SE P 6 -7 SE P 7 -8 SE P 8 -9 SE P 9 -1 0 SEP 10 -1 1 S EP 11 -1 2 S EP 12 -1 3 S EP 13 -1 4 S EP 14 -1 5 S EP 15 -1 6 S EP 16 -1 7 S EP 17 -1 8 S EP 18 -1 9 S EP 19 -2 0 S EP 20 -2 1 S EP 21 -2 2 S EP 22 -2 3 S EP 23 -2 4 S EP 24 -2 5 S EP 25 -2 6 S EP 26 -2 7 S EP 27 -2 8 S EP 28 -2 9 S EP 29 -3 0 S EP TPvsEVC (Corr.) p1 (Sig.) 7/25/2014
EVC vs. Total # of Geodesics
•
Daily correlation between total number of shortest paths (geodesics) and
eigenvector centrality (TSPvsEVP) for September 2012.
•
Statistical significance is given by p2.
-0.80 -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 1.00 1-2 SE P 2-3 SE P 3-4 SE P 4-5 SE P 5-6 SE P 6-7 SE P 7-8 SE P 8-9 SE P 9-10 SEP 10 -1 1 S EP 11 -1 2 S EP 12 -1 3 S EP 13 -1 4 S EP 14 -1 5 S EP 15 -1 6 S EP 16 -1 7 S EP 17 -1 8 S EP 18 -1 9 S EP 19 -2 0 S EP 20 -2 1 S EP 21 -2 2 S EP 22 -2 3 S EP 23 -2 4 S EP 24 -2 5 S EP 25 -2 6 S EP 26 -2 7 S EP 27 -2 8 S EP 28 -2 9 S EP 29 -3 0 S EP TSPvsEVC (Corr.) p2 (Sig.)
Implication of Findings
•
The extent to which information can spread
virally,
•
rather than the speed of information
propagation,
•
determines the loci of influence in online user
networks.
•
Alternatively, eigenvector centrality may not a
universal measure of influence in online user
networks
Mayande & Weber -- OUI 2014 15 7/25/2014
Further Research
•
Better measures of influence in online user
networks to be found.
•
This study is a work in progress,
–
which will hopefully lead to a behavioral theory of
online user networks
–
that encompasses influence, networks structure
and information flow.
About the Authors
• Nitin Mayande has been an independent consultant in the area of virtual social network design since 2008. Currently, he is the Chief Scientist at Tellagence Corporation, a social media analytics firm. He received his training in communication engineering in India, and holds an MS in Technology Management from Portland State University, where he is pursuing his doctorate in the same field. • Charles Weber received (among other degrees) a B.S. degree in Engineering Physics from the University of Colorado, Boulder; an M.S. degree in Electrical Engineering from the University of
California, Davis; and a Ph.D. in Management from MIT's Sloan School of Management. He joined Hewlett-Packard Company as a process engineer in an IC manufacturing facility. He subsequently transferred to HP’s IC process development center, working in
electron beam lithography, parametric testing, microelectronic test structures, clean room layout, and yield management. From 1996 to 1998, Charles managed the defect detection project at SEMATECH as an HP assignee. In December 2002, he joined the faculty of Portland State University where he is an associate professor of engineering and technology management. In the winter semester of 2014, he held the Fulbright/Kathryn and Craig Hall Distinguished Chair for Entrepreneurship in Central Europe at the Vienna University of Economics and Business.
17 Mayande & Weber – OUI 2014
Mayande & Weber -- OUI 2014 7/25/2014
References (1)
Allen, T. R. (1977). Managing the Flow of Technology. Cambridge, Mass., USA: MIT Press.
Bonacich, P. (1972). Factoring and weighting approaches to clique identification. Journal of Mathematical Sociology, 2, 113-120.
Borgatti, S. P. (2005). Centrality and network flow. Social Networks, 27, 55-71.
Burt, R. S. (1992). Structural Holes: The Social Structure of Competition, Cambridge, Mass, USA: Harvard University Press.
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95-S120.
Franke, N. and Shah, S. (2003). How communities support innovative activities: An exploration. Research Policy, 32, 157–178.
Freeman, L.C. (1977). A set of measures of centrality based on betweenness. Sociometry, 40, 35-41. Freeman, L.C. (1979). Centrality in social networks: Conceptual clarification. Social Networks, 1, 215-239. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78, 1360-1380.
Lakhani, K. R. and von Hippel, E. (2003). How open source software works: “Free” user-to-user assistance. Research Policy, 32, 923–943.
Mayande, N. and Jetter, A. (2010). Business models, governance structures and network structures in virtual communities. 8th Workshop on Open and User Innovation, Boston, MA, USA, Aug. 1-2, 2010.
Mayande, N., Weber, C. M. and Jetter, A., “A novel approach to analyzing online user innovation networks” Mass Customization and Personalization Conference, San Francisco, California, USA, November 19, 2011.
References (2)
Mayande, N., and Weber, C. M. “Structure, information flow and synchronized innovation in online open innovation communities,” 10th Workshop on Open and User Innovation, Boston, MA, USA,
July 29-August 1, 2012.
Podolny, J. M. (1993). A status-based model of market competition. American Journal of Sociology, 98, 829–872.
Powell, W.W. (1990). Neither market nor hierarchy: Network forms of organization.In Barry M. Staw and L.L Cummings (eds.), Research in Organization Behaviour, 12, 295-336.
Putnam, R. D. 1995. Bowling alone: America's declining social capital. Journal of Democracy, 6(1), 65-78.
Von Hippel, E. (2001). Innovation by user communities: Learning from open-source software. MIT Sloan Management Review, 42(4), 82-86.
Von Hippel, Eric (2002). Open source projects as horizontal innovation networks - by and for users, Working Paper 4366-02, MIT Sloan School of Management, June 2002.
Wasserman, S. and Faust, K., (1994). Social Network Analysis: Methods and Applications, New York, NY, USA, Cambridge University Press.
Wellman, B., Boase, J. and Chen, W. (2002). The networked nature of community: On and off the Internet. IT and Society, 1(1), 151-165 (http://www.ITandSociety.org).
West, J. and Lakhani, K. R. (2008). Getting clear about communities in open innovation, Industry and Innovation, 15(2), 223-231.
Mayande & Weber -- OUI 2014 19 7/25/2014
Online User Communities
•
Featured prominently in the OUI literature
–
(e.g., von Hippel, 2001; Lakhani and von Hippel, 2003; Franke and
Shah, 2003; West and Lakhani, 2008).
•
“… interpersonal ties that provide sociability, support,
information, a sense of belonging, and social identity.”
–
(Wellman, 2002, p. 154)
•
Special case of
online user networks
,
–
which have been defined as “user nodes interconnected by
information transfer links”
(von Hippel, 2002, p. 1)
.
–
that do not necessarily exhibit all of the abovementioned
attributes of online user communities.
Extant Theory (continued.)
•
Extant theory does not show how influence within a
network can depend upon control of information flow.
•
In principle, every person within a social network can
decide whether or not information should be passed on
and who should receive it.
•
Yet, existing methods for social network analysis do not
take these preferential attachments into consideration.
•
They may consequently ascribe a degree of influence to a
manager that may not reflect his/her actual influence,
•
potentially causing managers to choose incorrect channels
to get things done.”
(Mayande, Weber & Jetter, 2011, p. 1)
Mayande & Weber -- OUI 2014 21 7/25/2014
MEASURING INFLUENCE AND
INFORMATION FLOW
Various Types of Information Flow
•
Distinguished by two properties:
–
the routes through which the traffic flows
–
and the method by which the flows are propagated.
Serial Duplication
•
Path
•
Any possible way you can get from one node to another
•
Paths emphasize the impact information spread and exposure from
multiple sources.
•
Geodesic
•
Shortest possible path
•
Geodesics emphasize speed at which information spreads.
Serial duplication Geodesics Information Spread
Paths Information Spread and