Internet Learning Volume 3, Number 2, Fall 2014 | Page 108

Visualizing Knowledge Networks in Online Courses
C. Tools and Platform
Another area of future research and
development concerns Learning
Management Systems and other
platforms in which learning-focused discussions
are hosted. The traditional linear,
threaded discussion forum might make the
effective facilitation of discussion difficult.
Consider the case of Jakata’s entry into the
week 1 discussion: Jakata responds to all
visible posts in a brief timespan but receives
no notification of new posts after two new
students respond. Further, these new posts
are pushed to the bottom of a chronological
display, meaning that when Jakata logs
in, these responses may not be immediately
visible. Rich opportunity lies in investigating
the kinds of layouts, signals, entry
points, notifications, and recommendations
that give rise to more expressive and efficacious
social learning environments.
D. Data Science, Automation, and Algorithms
The numerical, categorical, text, and
other attributes of each response in
a corpus or a discussion are available
within the native graph structure of the data
for detailed statistical, graph-structural, and
other analyses, as well as for visualization.
This enables a combination of high-level visual
survey and detailed data analysis that
we hope can help speed the research-into-practice
cycle for online social and cooperative
learning environments.
Of course this does not mean we
have discovered how to reverse-engineer
deep, digital-ethnographic descriptions
from course or discussion data. Most attributes
for this study were manually coded by
human experts. However, if over time we
can develop the capabilities to automatically
apply some or all of these, or other, codes,
we believe it will lead to valuable new ways
of designing, describing, navigating, supporting,
and evaluating social and cooperative
learning activity in online courses
at scale. Therefore, the Pearson team continues
to evolve, scale, and automate this
research-based graph database system for
social and cooperative learning and discourse.
For example, we have implemented
experimental versions of: NLP-based question
and citation identification; a preliminary
topicSpread metric; a conversation
influence metric; an ontology comparison
model for understanding conversation concept
structures; a measure of response reciprocity
among a community of learners and
instructors; and visualization components
for viewing participant conversations and
corpora in ways similar to those presented
in this paper. Some of these features are currently
available in experimental alpha release
form to individual students and instructors
using the OpenClass LMS platform, on the
Learner Intelligence alpha page.
E. Closing Thoughts
We have suggested here that the
confluence of data-driven decisions
in education and the proliferation
of social media tools make the
time right for a deep exploration of how
knowledge is constructed in online social
learning spaces. Our goal, in particular, was
to define a set of individual, conversational,
and content-based attributes and behaviors
that might support the formation of thriving
social knowledge networks.
We have accomplished something of
our goal, in that we have been able to identify
and visualize trends and behavior in those
three areas. We recognize, however, that the
work is far from complete, and we hope that
this paper serves as a catalyst for additional
research into this important, emerging field.
107