Internet Learning Volume 3, Number 2, Fall 2014 | Page 93
Internet Learning
Figure 10. A sample view of the discussion shown in Figure 9, with author corpora added.
with a diminished weight for the author’s
own posts. At each level of the tree at which
an author posts, they are assigned one point
for their own post, and one point for each
subsequent response that is not their own.
The process is repeated until there are no
more posts by that author in the tree. For
a given thread, a person with a higher DiscussionRank
score can be said to have generated
more discussion than a person with
a lower score. This metric does not claim
to evaluate the quality or relevance of discussion.
Just as PageRank (Page et al, 1999)
considers a link to a web page as a vote of
importance without otherwise judging the
quality of the page, DiscussionRank considers
a response to be a vote of importance
for a conversation. The resulting metric
serves as a consistent, replicable yardstick
for investigating what happens when multiple
individuals enter into conversation together,
and against which we can compare
other quantitative and qualitative measures.
Figure 11 describes the DiscussionRank
counting method. In the example shown,
DiscussionRank flips the ‘scoreboard’ upside-down
as compared with a basic measure
based on the raw number of posts an
author contributes.
In its basic application, Discussion-
Rank is assigned to a person: the author
of the initial post. Thus we could compare
Renlit’s thread to Kerrad’s thread (as we will
do in 7.3. RQ2 Example), to assess which
author’s thread produced the most discussion
activity. However, this metric can be
extended and repurposed in interesting
ways. First, the initial response node need
not be the lead post. Imagine that both a
student and a teacher post questions at the
third level of a thread. One could measure
DiscussionRank from each point to determine
which person’s post generated the
most subsequent discussion. One could also
analyze data over longer periods of time,
in various situations, and under different
activity structures, to see which individuals
are more or less highly ranked under
specific conditions. Secondly, Discussion-
Rank can measure not only the generative
92