# Emergent (Non)Majors:

# Communities and Connections in an Interdisciplinary College

## Samuel Heller

### June 2, 2008

Over the years there has been much discussion of and concern about
the structure, or lack thereof, of a College of the Atlantic (COA)
education. Some maintain that since all COA students design their
own majors, all students have a unique course of study. These voices
typically praise students for taking a diverse set of courses and
charting an individualized and special, if not unique, path through
the curriculum. Others suggest that COA has de facto majors.
People with this point of view suggest that students form clusters or
cohorts that concentrate in a relatively a focused area, re-creating
the disciplinary majors that the college has sought to avoid. Those
with this view may criticize students for taking a narrow set of
courses and criticize faculty for offering classes which are only of
interest to this narrow set of de-facto majors.
The mathematical tools in the emerging area of complex network
analysis make possible an empirical examination of these concerns.
In this talk we will present results from such an analysis. We have
implemented and applied a standard algorithm for discovering clusters
or communities in networks. We consider a network of courses formed
by students' course selections. Two classes are linked together
if there were students who took both classes. The strength of the
link is related to the relative number of students in both classes.
The algorithm we use for discovering clusters is non-parametric, in
the sense that we do not specify the number of clusters for the
algorithm to seek, nor do we use any information beyond the collective
course choices of students.

Our results reveal five distinct groupings or clusters of classes.
While the clusters consist of a fairly diverse set of courses, they
nevertheless seem to each have a particular academic flavor. Our
results are statistically significant in so far as it is
astronomically unlikely that such clustering could occur if students
were selecting classes at random. Nevertheless, the overall
clustering effect is not extremely strong at the level of individual
courses and students, and we caution against over-interpreting our
results.

Although there will be some technical components in our presentation,
it is intended to be accessible to all community members. There will
be ample opportunity for discussion and critique.