A FRAMEWORK FOR INFORMATION DIFFUSION OVER
SOCIAL NETWORKS RESEARCH
Outlining Options and Challenges
Juan Yao and Markus Helfert
School of Computing, Dublin City University, Dublin, Ireland
Keywords: Information Diffusion, Social Networks, Social Influence.
Abstract: Information diffusion is a phenomenon in which new ideas or behaviours spread contagiously through
social networks in the style of an epidemic. Recently, researchers have contributed a plethora of studies,
approaches and theoretical contributions related to various aspects of the diffusion phenomenon. There are
many options and approaches. However there are only rare research articles consolidating and reviewing the
various options. In this paper, we aim to contribute an overview of the most prominent approaches related to
the studies of the diffusion phenomenon. We present a framework and research overview for this area. Our
framework can assist researchers and practitioners to identify suitable solutions and understand the
challenges in the information diffusion over social networks research.
1 INTRODUCTION
Information diffusion within social networks has
known an increased interest in recent years. It is a
ubiquitous process in human social networks.
Recently, the emergence of mobile, email and online
social networking have gradually transformed
communication among people. Social interaction
through these platforms leaves extensive digital
traces by its very nature. The availability of such
rich data has lead to a number of empirical and
theoretical studies on information diffusion process
at a more precise, quantitative level. These studies
have shed light on several important principles of
such phenomenon. The objective of this paper is to
review various recent studies related to information
diffusion process in social networks, summarize
their findings and possible extensions, and present
directions for future research.
2 MODELS & FORMAL
DESCRIPTION
While the outcomes of the diffusion processes are
easily visible, their inner workings have remained
elusive. Models of diffusion process aim to provide
a clear picture of how such dynamic process unfolds
by defining the way information flows between
individuals. Here we consider a collection of
probabilistic and game-theoretic diffusion models;
summarize some of the challenges faced when
modelling the diffusion phenomenon and common
approaches to address these challenges.
2.1 The Dynamics of Adoption Process
One of the most common approaches of modelling
the adoption process is to explicitly represent the
step-by-step dynamics of adoption. Typically, it
assumes the dynamic process unfolds in discrete
time unit with each individual following certain rule
when making adopting decision. A set of individuals
are chosen to be initial active set which corresponds
to the early adopters of the behaviour. Starting with
the initial active set the process then unfolds as
follows: at each time step, individuals who were
active at previous time step remain active; an
individual will be activated according to certain
rules such as a threshold number of his neighbours
have been activated. The dynamic process continues
until no more activation is possible.
In the economics literature, diffusion models
have been studied from a game-theoretic perspective
(Morris, 2000). This approach builds on work
investigating how a new technology A might be
spread through a social network of individuals who
491
Yao J. and Helfert M. (2010).
A FRAMEWORK FOR INFORMATION DIFFUSION OVER SOCIAL NETWORKS RESEARCH - Outlining Options and Challenges.
In Proceedings of the 5th International Conference on Software and Data Technologies, pages 491-494
DOI: 10.5220/0002926204910494
Copyright
c
 SciTePress

are currently users of technology of B. The diffusion
process is modelled as the dynamic of a coordination
game played on the social network, in which the
adoption of a common strategy between players has
a higher payoff. In particular, in every time step,
each player in a social network has two available
choices A and B. Each player receives a positive
payoff for each of his neighbours that has the same
choice as he does, in addition to an intrinsic benefit
that he derives from his choice.
2.2 Modelling Social Influence
An underlying premise many diffusion models build
on is that people are influenced by their neighbours
in the social network when making adopting
decision. Social influence determines to a large
extent what we adopt and when we adopt it. One
simple way to capture this effect is to assign each
individual in the network a threshold value
(Granovetter, 1978). The threshold value indicates
the personal tendency of an individual to adopt the
behaviour when his neighbours do. In addition to the
number of adopted friends, how those friends are
connected to one another could also have an impact
on individual’s propensity of adopting. It is argued
that (Burt, 2005) if two actors related to the same
individual are also related to each other, they have
greater power over that individual than if they were
unrelated. A primitive approach (Katona, Zubcsek
and Sarvary, 2009) to incorporate this idea is to
postulate the adoption likelihood of individual
increases as a function of the density of relationships
among their adopted neighbours.
Another way to encode neighbour influence is by
using infection rate, inspired by the epidemic
models. In the Independent Cascade Model
(Goldenberg, Libai and Mullen, 2001), every time
an individual contacts with an active neighbour, he
has a constant chance of getting activated.
Obviously, a constant infection rate seems not
accurate enough. Kempe, Kleinberg and Tardos
(2003) refined the Independent Cascade Model to
interpret the idea that an individual's receptiveness to
influence depends on the past history of interactions
with his neighbours. Contrary to a constant rate, an
individual's probability of being activated is a
function of the set of neighbours have already tried
and failed to influence him.
It’s commonly accepted that social influence
affects adopting decision. However, it is not the only
factors that drive information diffusion. Van den
Bulte and Stremersch (2004) argue that S-shaped
diffusion curves can also result from heterogeneity
in the intrinsic tendency to adopt. In the context of
new product diffusion, individuals are different with
respect to subject matter expertise, strength of
opinion, personality traits, media exposure or
perceived adopting costs. Efforts have been made to
capture this effect. For instance, Hartline, Mirrokni
and Sundararajan (2008) model a buyer’s decision to
buy an item is influenced by the set of other buyers
that own the item and the price at which the item is
offered.
2.3 Modelling Diffusion of Competing
Technologies
Most models discussed above typically focus on the
diffusion of one behaviour or technology. What is
often ignored is that several different behaviours
may coexist in a system at the same time and
possibly compete with each other. Some may have a
better chance to survive and spread than others. In
fact, this scenario frequently arises in the real world.
In consumer market, producers of consumer
technologies often must introduce a new product
into a market where a competitor will offer a
comparable product, which makes them vie for sales
with competing word-of-mouth cascades. Thus it is
important to examine the diffusion of competing
technologies in social networks.
As mentioned earlier, such phenomena can be
modelled as a coordination game played on the
edges of the social network with multiple equilibria.
An influential paper by Morris (Morris, 2000)
provided a set of elegant game-theoretic
characterizations for when these qualitatively
different types of equilibria arise in terms of the
underlying network topology and the quality of
technology A relative to technology B. In recent
work, Immorlica et al. (2007) incorporates
compatibility between technologies and discuss how
this effects the diffusion. Their results show that in
some cases, for one technology to survive the
introduction of another, the cost of adopting both
technologies must be balanced with a narrow,
intermediate range.
3 EMPIRICAL STUDIES &
FINDINGS
Leskovec, Adamic and Huberman’s (2006) argues
that while above models address the question of how
influence spreads in a network, they are based on
assumed rather than measured influence effects. The
ICSOFT 2010 - 5th International Conference on Software and Data Technologies
492

wide variety of rules theoretical diffusion models
posed on individuals’ behaviour, even if plausible,
are often lacking empirical support. Furthermore,
most of the diffusion models take a single snapshot
of the evolving network and then build upon this
static network topology. As such, it becomes unclear
how accurately existing models render real-world
diffusion phenomena. Cointet and Roth (2007)
suggest that future investigations of the diffusion
mechanisms should begin with adequate empirical
protocols, then propose adapted modelling
frameworks. Most of the recent empirical studies on
information diffusion focus on addressing the
following fundamental questions:
 What are the characteristics of social influence?
 What are the patterns of information cascade?
 How does network structure affect diffusion
process?
3.1 Characteristics of Social Influence
If we view the diffusion process as a cascade of
social influence, a natural starting point is to
understand the local mechanism of the influence.
Social influence can be described as the actions of
an individual can induce his or her friends to behave
in a similar way. Backstrom et al. (Backstrom et al.,
2006) investigated the membership problem in
online communities and measured how propensity of
individuals to join a community depends on friends
already within the community. Specifically, they
measured the joining probability as a function of the
number of friends already in the community. The
adoption curve exhibits kind of a diminishing returns
pattern in which it continues increasing, but more
and more slowly, even for large numbers of friends.
This kind of diminishing returns pattern has also
been observed in many other studies. For example,
Christakis and Fowler (Christakis and Fowler, 2007)
studied the spread of obesity in large social network
over 32 years and found out that people were most
likely to become obese when friends became obese.
However, Kleinberg (Kleinberg, 2007) argued that
the dependence of adopting probability on the
number of friends adopted expressed in this way
reflects an aggregate property of the full population,
and does not imply anything about an particular
individual’s response to their friends’ behaviour.
Anagnostopoulos et al. (Anagnostopoulos et al.,
2008) argues that while these studies have
established the existence of correlation between user
actions and social affiliations, they do not address
the source of the correlation. There are factors such
as homophily (Mcpherson, Lovin and Cook, 2001)
or unobserved confounding variables that can induce
statistical correlation between the actions of friends
in a social network. Homophily is the tendency for
people to choose relationships with people who are
similar to them, and hence perform similar actions.
Recently, Anagnostopoulos et al. (Anagnostopoulos
et al., 2008) proposed a statistical test to distinguish
social influence from correlation using time series
data of user actions. Crandall et al. (Crandall et al.,
2008) developed techniques for identifying and
modelling the interactions between social influence
and social selection process and found an elaborate
interplay between the two factors.
3.2 Patterns of Information Cascade
Although above studies have shed light on the
mechanisms of social influence, the overall patterns
by which the influence spreads through social
networks have been a mystery. Several recent
studies have been conducted to illustrate the
existence of cascade and observe patterns of
cascading behaviour. In particular, Leskovec et al.
(Leskovec, Singh and Kleinberg, 2006) consider
information cascades in a recommendation network.
According to their observation, the distribution of
cascade sizes can be approximated by a heavy-tailed
distribution. Generally cascades are shallow but
occasional large bursts also occur. Another recent
study on cascading behaviour in large blog network
(Leskovec et al., 2007) found that blog posts do not
have a bursty behaviour; they only have a weekly
periodicity.
Liben-Nowell and Kleinberg (Liben-Nowell and
Kleinberg, 2008) traced the information cascade
process on a global scale by using methods to
reconstruct the propagation of massively circulated
Internet chain letters. Contrary to predictions that
large-scale information spreads widely and reaches
many people in very few steps, their results show
that the progress of these chain letters proceeds in a
narrow but very deep tree-like pattern, continuing
for several hundred steps.
3.3 Role of Network Structure
Researchers have long emphasized the important
role played by the network structure in determining
properties of information diffusion. However, the
way such dynamic process is affected by network
structure is still poorly understood. How widely
does information spread? Does it spread only in
local region? Does it spread quickly on a dense
network? Several studies on real-world network data
A FRAMEWORK FOR INFORMATION DIFFUSION OVER SOCIAL NETWORKS RESEARCH - Outlining Options
and Challenges
493

have been conducted to address questions like these.
Granovetter (Granovetter, 1973)’s weak ties
hypothesis states that weak ties typically act as
connectors between different communities or circles
of friendship. Using mobile call records, Onnela et
al. (Onnela et al., 2007) have observed a coupling
between interaction strengths and the network’s
local structure, confirming the weak tie hypothesis.
Specifically, they found that weak ties appear to be
crucial for maintaining the network’s structural
integrity, but strong ties play an important role in
maintaining local communities. In addition, they
investigated how the dynamics of different tie
strengths influence the spread of information in the
network. They show that the coupling between tie
strength and network structure significantly slows
the diffusion process, resulting in dynamic trapping
of information in communities and find that both
weak and strong ties have a relatively insignificant
role as conduits for information.
4 CONCLUSIONS
In this paper we summarize three major challenges
faced when modelling diffusion phenomenon. We
review recent studies that measure diffusion
phenomenon empirically. We believe our framework
can assist researchers and practitioners to understand
the challenges in studying information diffusion
over social networks and identify suitable solutions
to address those challenges.
REFERENCES
Anagnostopoulos, A., Kumar, R., and Mahdian, M., 2008.
Influence and correlation in social networks. In KDD
'08, pages 7-15, New York, NY, USA. ACM.
Burt, R. S., 2005. Brokerage and Closure: An Introduction
to Social Capital. Oxford University Press, USA.
Backstrom, L., Huttenlocher, D., Kleinberg, J., and Lan,
X., 2006. Group formation in large social networks:
membership, growth, and evolution. In KDD '06,
pages 44-54, New York, NY, USA. ACM.
Cointet, J. P. and Roth, C., 2007. How realistic should
knowledge diffusion models be? Journal of Artificial
Societies and Social Simulation, 10(3).
Christakis, N. A. and Fowler, J. H., 2007. The spread of
obesity in a large social network over 32 years. N Engl
J Med, 357(4):370-379.
Crandall, D., Cosley, D., Huttenlocher, D., Kleinberg, J.,
and Suri, S., 2008. Feedback effects between
similarity and social influence in online communities.
In KDD '08, pages 160-168, New York, NY, USA.
ACM.
Granovetter, M. S., 1973. The strength of weak ties. The
American Journal of Sociology, 78(6):1360-1380.
Granovetter, M., 1978. Threshold models of collective
behavior. The American Journal of Sociology,
83(6):1420-1443.
Goldenberg, J., Libai, B., and Mullen, E., 2001. Talk of
the network: a complex systems look at the underlying
process of word-of-mouth. Marketing Letters,
12(3):211-223.
Hartline, J., Mirrokni, V., and Sundararajan, M., 2008.
Optimal marketing strategies over social networks. In
WWW '08, pages 189-198, New York, NY, USA.
ACM Press.
Immorlica, N., Kleinberg, J., Mahdian, M., and Wexler,
T., 2007. The role of compatibility in the diffusion of
tecfhnologies through social networks. In Proc. 8
th
EC,
pages 75-83.
Kempe, D., Kleinberg, J., and Tardos, Ã., 2003.
Maximizing the spread of influence through a social
network. In KDD '03, New York, NY, USA. ACM.
Kleinberg, J., 2007. Cascading Behavior in Networks:
Algorithmic and Economic Issues. Cambridge
University Press.
Katona, Z., Zubcsek, P. P., and Sarvary, M., 2009.
Network effects and personal influence: Diffusion of
an online social network. Working paper.
Leskovec, J., Adamic, L. A., and Huberman, B. A., 2006.
The dynamics of viral marketing. In EC '06:
Proceedings of the 7th ACM conference on Electronic
commerce, pages 228-237, New York, NY, USA.
ACM.
Leskovec, J., Singh, A., and Kleinberg, J., 2006. Patterns
of influence in a recommendation network. pages 380-
389.
Leskovec, J., McGlohon, M., Faloutsos, C., Glance, N.,
and Hurst, M., 2007. Cascading behavior in large blog
graphs.
Liben-Nowell, D. and Kleinberg, J., 2008. Tracing
information flow on a global scale using internet
chain-letter data. Proceedings of the National
Academy of Sciences, 105(12):4633-4638.
Morris, S., 2000. Contagion. Review of Economic Studies,
67:57-68.
Mcpherson, M., Lovin, L. S., and Cook, J. M., 2001. Birds
of a feather: Homophily in social networks. Annual
Review of Sociology, 27(1):415-444.
Onnela, J. P., Saramäki, J., Hyvönen, J., Szabó, G., Lazer,
D., Kaski, K., Kertész, J., Barabási, A.-L., 2007.
Structure and tie strengths in mobile communication
networks. Proceedings of the National Academy of
Sciences, 104(18):7332-7336.
Van den Bulte, C., Stremersch, S., 2004. Social Contagion
and income heterogeneity in new product diffusion: A
meta-analytic test. Marketing Science, 23(4):530-544.
ICSOFT 2010 - 5th International Conference on Software and Data Technologies
494