Fall 2004
Tuesdays
12:00 GWC 604
Seminar
Schedule:<http://math.la.asu.edu/~tom/cognition/math+cogsched.html>On
Tuesday, October 5, at 12:00 Noon
in GWC 604,
the Mathematics and Cognition Seminar will
present brief discussion with Root Gorelick
of the School of Life Sciences and SIRG
on the topic of
"Quantifying Social
Interactions"
Abstract
| The ratio of mutual to
marginal entropy (which asymptotically has an F-distribution) provides
a measure of social interactions and social structure. This
statistic
can be used to measure the ensemble of all pair-wise interactions
within a social group (where rows and columns of the data matrix are
individuals and where the matrix components are the time each pair
spends interacting) division of labor within a social group (where rows
of the data matrix are individuals and columns are tasks and the matrix
components are the time each individual spends on each task). By
broadening the concept of what constitutes a social group, we can apply
this methodology to many non-obvious applications. With the
social
group consisting of species in an ecosystem, normalized mutual entropy
can be used to quantify levels of predator-prey interactions or levels
of specialization of pollinating animals and the plants they
pollinate. With rows of the data matrix being females, columns
being
males, and entries in the matrix being number of matings, we can
quantify the continuum amongst monogamy, polygyny, and polyandry, or
the intensity of sexual selection. With the social group
consisting of
all trinucleotides and tasks consisting of all amino acids, we can
estimate the change in specialization over the evolution of the modern
genetic code. The one limitation on this methodology is the data
structure, a matrix. Otherwise there is a myriad of biological
applications for normalized mutual entropy. The use of normalized
mutual entropy broadens our concept of social groups (and division of
labor) to any biological systems that can be represented by a matrix of
behaviors, interactions, or relationships. |