IEEE SIGNAL PROCESSING MAGAZINE [77] MARCH 2012
H
ow can we model influence between individuals
in a social system, even when the network of
interactions is unknown? In this article, we
review the literature on the “influence model,”
which utilizes independent time series to esti-
mate how much the state of one actor affects the state of
another actor in the system. We extend this model to incorpo-
rate dynamical parameters that allow us to infer how influ-
ence changes over time, and we provide three examples of
how this model can be applied to simulated and real data. The
results show that the model can recover known estimates of
influence, it generates results that are consistent with other
measures of social networks, and it allows us to uncover
important shifts in the way states may be transmitted between
actors at different points in time.
INTRODUCTION
The concept of influence is extraordinarily important in the
natural sciences. The basic idea of influence is that an out-
come in one entity can cause an outcome in another entity.
Flip over the first domino, and the second domino will fall. If
we understand exactly how two dominoes interact—how one
domino influences another—and we know the initial state of
the dominoes and how they are situated relative to one anoth-
er, then we can predict the outcome of the whole system.
For decades, social scientists have also been interested in
analyzing and understanding who influences whom in social
systems [1], [2]. But the analogue with the physical world is
not exact. In the social world, influence can be more complicat-
ed because internal states are often unobservable, intentional
Digital Object Identifier 10.1109/MSP.2011.942737
Modeling
Dynamical Influence
in Human Interaction
[
Wei Pan, Wen Dong, Manuel Cebrian,
Taemie Kim, James H. Fowler, and Alex (Sandy) Pentland
]
Date of publication: 17 February 2012
[
Using data to make better inferences
about influence within social systems
]
©iSTOCKPHOTO.COM/ANDREY PROKHOROV
Signal and Information Processing
for Social Learning and Networking
1053-5888/12/$31.00©2012IEEE