This study focusses on the role of highly active participants in online learning communities on Facebook. These people, often known as “power users” in the literature on social computing, are a common feature of a wide range of online learning groups, and are responsible not only for creating most of the content but also for getting discussion going and providing a basis for other’s participation. We test whether similar dynamics hold true in the context of online learning.
Based on a transactional dataset of almost 10,000 interactions with an online community of 32 postgraduate students who were following the same online course, we find evidence that power users also exist in the context of online learning. However, whilst they do create a lot of content, we find that they are not fundamental to keeping the group together, and in fact are less adept at creating content which generates responses than other “normal” users. This suggests that online learning communities may have different dynamics to other types of electronic community: it also suggests that design efforts should not be focused solely on attracting a small core of “power learners”. Rather, diverse types of users are needed for online learning communities to survive and prosper.
Authors:
Cristóbal Cobo, Center for Research - Ceibal Foundation, Uruguay
Monica Bulger, Data & Society Research Institute, United States
Jonathan Bright, Oxford Internet Institute, United Kingdom
Ryan den Rooijen, Oxford Internet Institute, United Kingdom
Presented at the LINC Conference (MIT, 2016) Digital Inclusion: Transforming Education through Technology.
2. Introduction
Unique challenges of studying online learning
communities
Blur between online and offline
Our students--physical interaction, context
collapse
This study focus on the concept of power users (term taken from the
field of social computing) individuals who make a disproportionately
large contribution to online activities. Studies show that retaining these
power users should be a key focus when designing online activity.
3. keyquestions
1) Are power learners naturally committed to the learning community, or do they decide to become more
involved progressively?
(Studies have suggested that power users can also learn as they develop. Does the same hold true for online
education?)
2) Are power learners crucial for starting and maintaining online education communities?
(i.e. Wikipedia editing, active early involvement of a small mass of committed participants is crucial for
starting collective activities. )
● This study explores a ‘small’ sampling (active consent of participants).
● Data describe interactions of 32 masters and PhD level students in a Facebook group specifically set
up to promote interaction amongst a real life masters cohort during their year-long interdisciplinary
study programme (from October to June) at a large UK university.
● Dataset records: time and kind of interactions (post or comment).
4. Methodology
Figure 1 – Activity in the Facebook group over time
● High Activity:
Almost 10,000 contributions made
during the observation
(4 posts per day, 18 responses).
● Contribution were higher during
term time: Levels declined from
the start of the course in
October.
Still attracting a non-trivial
amount of traffic even six months
after the course finished
(September 2012).
5. Distributionofpostsandcommentsperuser
Figure 2 - Distribution of posts and comments per user
● Most members made a high level of
contributions (median user made 285
posts or comments).
● A clear evidence of “power learners”,
outliers in terms of the number of
contributions made.
● Outliers (red): Two members are clear
outliers (~800 comments) and two others
(with over 150 posts).
6. Results What explains the emergence of Power Learners?
Lit. Rev. two competing perspectives:
1) They are naturally predisposed to take on this high activity role;
2) They are normal users, but gradually adopt the role of power learner as time goes by.
● Cumulative contributions made by each user.
● Key moment: ~75 and 100 days after the initial posting, they were active but no outliers (in terms of
contributions made). Does not appear that people were “natural” power learners when they arrived.
Figure 3 – Cumulative contributions by user type. Chart shows the first 50 days of activity. Power learners in red.
7. Figure 5 – Posts created and response rates (left panel).
Although Power Learners contributed a significant amount to the group,
they generate less responses than normal users.
Local regression line showing the evolution of
posts and response rates (right panel)
2nd question: Do power users drive the activity of the group?
8. When Facebook is considered as a “network” (each node is a user and each link is created when one user responded to
another). This is crucial to identify which users hold the network together.
Each set is a month of activity (from October to June).
Figure 6 – Community as a network with Power Learners in red (left panel) and without Power Learners (right panel)
Two findings:
a) The network shows little evidence of being fragmented into groups (a core of well-connected users).
b) Power Learners are almost always near the centre of the graph (well connected to all others). If they were removed the
network remains tightly connected (they do not seem to be fundamental in keeping the network together).
9. Conclusions
1. Power learners are naturally predisposed towards becoming high activity users (even from the
first day) along the course they truly distinguish themselves.
2. Power learners create a lot of contents. However they are less likely to receive a response than
other users.
3. Power learners can be central to enable communication in the network, but not absolutely vital
(network resilience).
Limitations of the study:
● The type (elite university) and size of the community (results can not be generalized).
● Better understand the type of communications within the group (in-depth qualitative analysis).
● On and offline educational networks need to be compared.
11. Figure 4 - Number of contributions by normal and power learners during the early parts of the group
Contributions online (first day, week and month):
● Power learners were more active than normal users during their earliest interactions.
● A decisive break emerged later on during the course.
12. Social media platforms (SMP) play an integral part
of the educational experience (groups of
students who use SMP as a complement of the
courses).
Whether online or off:
Learning is the result of an interaction between
individuals, (sharing, negotiation of roles,
communication norms and practices).
SMP form part of students’ lives outside of the
classroom (fewer barriers for its adoption).
While some classrooms boast vibrant online
discussions running parallel to the offline
learning activity, other communities attract little
participation.
This study focus on the concept of power users (term taken from the
field of social computing) individuals who make a disproportionately
large contribution to online activities. Studies show that retaining these
power users should be a key focus when designing online activity.
13. ● Betweenness centrality:
Measures the extent to which a node is
central to a network.
To identify which nodes facilitate
communication between other pairs of
nodes within a network (on the basis of how
many other pairs of nodes sits between).
● At the beginning Power Learner’s
betweenness centrality was higher.
In the last three months do the values are
considerably higher.
● Power Learners are more active in terms of
trying to hold the network together
towards the end as the group.
Figure 7 – Betweenness centrality for different user types over time
Each node in the network for each of the month's plots the distributions of these centralities as boxplots (Power Learners in red).
14. This finding is reinforced by table 1, which is a
logistic regression which seeks to explain why some
posts received a comment whereas others did not. It
includes three control variables:
whether the post contained a question mark (which
increased the likelihood of a response by 174%),
the number of words in the post (with each word
increasing the likelihood of a response by 1%),
whether the post contained a link (which decreased
the likelihood of a response by 49%).
It also includes a variable for whether the user was a
power learner or not, which shows that posts created
by power learners were in fact 22% less likely to
receive a response. All these findings were
statistically significant at conventional levels.
Table 1 – Logistic regression explaining the likelihood of a post receiving a
response