Learning with Strangers
The Value of Sets in Online Learning
Jon Dron and Terry Anderson
Technology-enhanced Knowledge Research Institute, Athabasca University, University Drive, Athabasca, Canada
Keywords: Social Media, Education, Learning Technology, Group, Network, Set, Collective, Lifelong Learning,
e-Learning.
Abstract: Most research and practice relating to online and distance learning to date has focused on the social form of
the intentional group, a named collection of people, typically hierarchically organized, with norms and/or
explicit rules of conduct as well as inclusion or exclusion, membership, pacing and shared goals. The group
provides a backdrop and infrastructure support for formal or informal learning activities. Since the last
decade of the 20
th
century a different social form, the network, has been the subject of much research in
informal and non-formal learning. Increasingly, however, we teach and we learn with and from countless
anonymous others that are not formed into either identifiable networks or groups. We describe a collection
of people who share little apart from interests or attributes but that none-the-less affect one another’s
learning as the Set. Under the right conditions, collective intelligence (or collectives) can emerge from such
sets that can actively guide learning. In this paper we explore the nature of set-based learning and the role
that collectives can play in helping or hindering learning.
1 INTRODUCTION
Much learning through the Internet involves
following or active engagement with strangers,
whether through sharing ideas and comments in
blogs and websites, editing a Wikipedia page,
contributing to a Q&A forum or posting to a listserv.
Traditional notions of social capital, group dynamics
and social contracts are significantly mutated when
we are not talking with people we know or
recognize, and we are in the open, away from the
safety of controlled groups of people with shared
purposes and norms. Beyond that, there are often
emergent and/or designed effects arising from large-
scale interactions that play an active role in shaping
the behaviours of participants in this partly
anonymous crowd. This paper is concerned with the
actual and potential value of these sets of minimally
connected strangers both purposefully and
inadvertently helping one another to learn. As well
as explaining how such sets differ from the more
commonly researched social forms of groups and
networks, we will be listing some of the common set
tools, some of the ways they can be used for
learning, some of the risks and dangers, and some
potential and actual solutions to those problems.
2 GROUPS, NETS, SETS AND
COLLECTIVES
2.1 The Group
The bulk of research into social learning, whether at
a distance or not, has so far focused on ways that
intentionally formed groups can be used to help
people to learn. The group (or often ‘team’ in
business circles) is a fundamental social form. It
plays out in myriad ways, from the most rigid
committee or court to the most informal study group
or family, but it has some common features. For
learning, there are familiar groups such as classes,
cohorts, tutorial/seminar/working groups, teams,
faculties, schools, houses and clubs. By and large
they have leaders and, beyond a certain size,
hierarchies of leadership. Almost all have names. All
have implicit or explicit rules and rituals that govern
how members should behave, how people become
members and, as importantly, who to exclude.
In a learning context, most are time-limited,
specify distinct goals and operate to a schedule.
Groups tend to go through phases of development,
such as forming, storming, norming and performing,
or Salmon’s five stages of e-moderation (Salmon,
99
Dron J. and Anderson T..
Learning with Strangers - The Value of Sets in Online Learning.
DOI: 10.5220/0004955700990104
In Proceedings of the 6th International Conference on Computer Supported Education (CSEDU-2014), pages 99-104
ISBN: 978-989-758-022-2
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
2000). Groups have explicit membership: it is
almost impossible to unknowingly become a
member of a group and it is at least in principle
possible to know the names of all the other group
members. The overhead needed to organize,
schedule and maintain a group is significant. Groups
require commitment and do not scale well in a
learning context to large numbers of people.
2.2 The Network
Over the past few decades there has been an
increasing amount of research into an equally or
more important social form for learning, the
network. Every individual’s networks are different
from every other’s, because networks are constituted
of the people we know. From the weakest ties of
recognition to the strongest friendships, we are
normally members of many overlapping networks,
often without even being particularly aware of it.
Networks are mostly emergent structures based on
the connections we make with others, and their
edges are typically fuzzy and constantly shifting.
Ideas, norms, behaviours and other forms of learning
can and do spread through networks, often with
amazing speed and effect. The Internet has played a
major role in making networks more tangible, most
notably through social networks like Facebook,
LinkedIn and Google+. However, many other
Internet-based systems from emails to instant
messaging to blogs enable the nurturing and growth
of social networks. The network is a fundamental
social form for learning, described by Wenger as a
community of practice (Wenger, 1998), later refined
to the notion of the network of practice (Wenger et
al., 2011) and providing the basis of Siemens’s
Connectivist model (Siemens, 2005). Indeed,
networks play a crucial role in groups, connecting
members within the group as well as sustaining the
exchange of knowledge beyond the group.
2.3 The Set
A third important social form can also be described
that extends beyond groups and networks, and that
has not received anything like as much recognition
in literature on learning: the set. Sets are simply
collections of people with shared attributes who
share the same virtual or physical space. In a
learning context, the most significant shared
attribute tends to be a shared interest in a topic but
others may matter too, such as prior knowledge or
location. In our non-virtual lives we can and do
make use of sets to learn. For example, when we
publish a book or a web page we normally provide
categories (tags) so that people with a particular set
of interests or attributes can find it. We do not know
who they are but, as authors, we are communicating
with and to the set of people who may find it
valuable. Equally, the set can communicate with us:
for instance, the fact that there is a set of people
outside who are carrying umbrellas tells me that I
should probably do the same when I go out. More
deliberate uses of sets are common: shows of hands
in a classroom, divisions of crowds by demographic,
gender, or other lines are a regular feature of our
lives, for better or worse.
Part of the reason for the lack of recognition of
sets for learning till now is due to the fact that, in
most social contexts before the advent of the mass
Internet, sets performed relatively little useful work.
The Internet makes it possible to interact with a vast
number of people with whom we have no shared
social connection at all. Much of the activity that
drives Wikipedia, for instance, is from anonymous
people whose only interaction is in editing one
another’s words. While networks and groups exist
on the Wikipedia site and can play a strong role in
the development of pages, there are at least as many
people contributing to the site who are helping one
another without ever being aware (or caring) who is
helping whom. Likewise, though networks and
groups exist on Q&A sites like Slashdot, Yahoo
Answers or StackExchange, much of the learning
that results from their use emerges from virtually
anonymous interactions between people unknown to
one another and not organized into groups or
networks. Sets are the basis of Google Search,
arguably the most significant learning technology
invented in the last millennium. Sets underpin
crowd-mining technologies such as Amazon’s book
recommendations, Netflix’s movie
recommendations and Pandora’s music
recommendations. Countless specialist sites cater for
particular interests that are, by nature and our
definition, set-oriented. Curation sites like Pinterest
and Learnist are largely set-oriented, focusing on
topics rather than communities. Twitter hashtags are
primarily concerned with sets, not networks or
groups. Usenet newsgroups and email listservs have
long been an important source of knowledge and
dialogue, often among strangers sharing nothing but
an interest in a topic or need for topic-specific
information. Despite the popularity of group-
supporting tools like learning management systems
and network-nurturing tools like Facebook,
Academia.edu and LinkedIn, set-based interactions
are the dominant social form in Internet-based
CSEDU2014-6thInternationalConferenceonComputerSupportedEducation
100
independent learning and may soon be in formal
education as well.
2.4 The Collective
Set-oriented systems can be wild places, full of half-
truths and falsehoods as much as rich and
meaningful information, not to mention abusive,
malevolent and mischievous contributions. This is
overcome in part through reification of the
conversation, so that individuals can choose the
most compelling solutions and arguments. More
significantly, almost all successful systems of this
nature incorporate crowd-sourced algorithmically
collated metadata like ratings, likes, reputation
measurement and filtering tools so that the crowd
can collectively guide its own members.
We refer to the outcome of this algorithmic
combination as a collective, using the term much
like the creators of Star Trek’s Borg to signify a
single entity made up of many independent entities
acting as one. Collectives combine the behaviours of
many people through one or more algorithms in
order to provide help, guidance or structure to
otherwise overwhelming or ambiguous content
generated by the crowd. The algorithms may be
provided by machines, such as in rating systems, or
collaborative filters, or by people, such as when
people are collectively drawn to active sites or
repelled from those that are too active, or both, as
we see in people’s reactions to the search order of a
Google search or the weightings of tags in a tag
cloud. In many cases, processing is split between a
machine and the heads or hearts of human beings,
the machine offering alternatives according to one
set of algorithms and people making choices using
others. A collective is not a social form as such, but
the emergent result of people interacting, directly or
indirectly, with one another.
2.5 Set Combinations
Social forms seldom exist in isolation. Sets may be a
supplement or a pre- or post-emergent form of
traditional group-based learning, existing networks
or conventional individual study. Equally, the social
forms we describe are not binary categories but are
more like primary colours that often occur in blends.
For example, at our own Athabasca University, our
individualized study model means that students are
self-paced, choosing when and how they work over
a six-month period. It is thus rare for two students to
be working on the same things at the same time.
Despite this, forums other social tools are normally
provided for each course. Although courses share
some group-like features including rules, shared
goals and hierarchies, students do not form teams,
seldom know others, do not collaborate and are not
expected to work together. Their interactions are
thus notably set-like. What they share helps them to
solve problems, alleviate a sense of isolation, and
discover different ways of seeing a subject. Many
large MOOCs, though they may have designs that
resemble those of conventional group-based
university courses, are more set-like in social form,
for similar reasons.
3 WHY DO PEOPLE
CONTRIBUTE TO SETS?
For many contributors to the public good, social
capital plays an important role: by providing help to
others, one is increasing one’s own social capital,
with consequent gains for all concerned (Nemoto et
al., 2011). This is equally true in a learning context
(Daniel et al., 2003). However, this is not the whole
story, even in tight-knit social networks, where
expectations of reciprocity may not play a dominant
role (Wasko and Faraj, 2005). A survey of frequent
contributors to Wikipedia found that five of the top
67 editors (those who have made at least 500 edits)
were known only by their IP addresses (Various,
2005). Amongst these anonymous contributors there
can be no expectation of reciprocal social benefits.
As a species, we have an evolved tendency towards
altruism that cannot be simply explained away by
assumptions that people rationally weigh costs and
benefits. We are genetically inclined to help one
another (Wilson, 2012). Beyond anonymous
contributions, many sets emerge as a side-effect of
other interactions. For example, academics may
publish blog posts primarily for the benefit of a
small subset of people in their own networks or
groups, while knowing that there is an added benefit
that their writings might be read by the set of others
with a similar interest.
4 SET LEARNING
Set-based learning tends to be appropriate when the
objectives of learning are already known. It is well
suited to information seeking, inquiry-based and
problem-based pedagogies, where goals are known
and the learner already has some subject knowledge.
While people often help one another in sets,
LearningwithStrangers-TheValueofSetsinOnlineLearning
101
there tends to be little or no deliberate collaboration
because there are few opportunities for sustained
interaction, no shared projects, and limited
scheduling of activities. Conversely, there are also
minimal temporal or spatial constrictions on
independent learning. Cooperation (not
collaboration) is the dominant form of working
together, in which learners working individually
contribute to the learning of others. There is more
sharing with others rather than direct dialogue and,
when dialogue occurs, it tends to be fleeting and
limited in scope. Where coordination does occur, it
is either through centralized methods like FAQs
compiled by individuals, or more sophisticated
structural processes such as the forking process used
in Github, that enables people working
independently to contribute to one another’s work.
Apart from sets that form around temporal
events, most sets tend to eschew schedules and
pacing. People tend to contribute as and when they
want or need to do so. For those seeking answers to
problems or discussions about issues, this can be
frustrating, unless the set is sufficiently large to
ensure a constant succession of contributors.
However, the almost ubiquitous reification of
previous interactions (including recommendations)
means that answers previously given at one moment
can continue to provide value to later-arriving
members of the set.
Sets have great value in forming and building
learning networks and even groups. For instance, on
sites that form around (say) support for a specific
piece of software, there is typically a caucus of
enthusiastic contributors who come to know and
respect or at least recognize the strengths and
limitations of one another, leading to what may often
be rightly described as a community. While there
may be hundreds or thousands of occasional
contributors in such spaces, and countless people
who do not contribute, but do read, such spaces
often contain rich social networks as well as sets.
The non-contributors in such set-oriented spaces are
often misleadingly referred to as ‘lurkers’. This is a
consequence of failing to recognize that sets are not
communities as such, mostly lacking the norms and
network bonds that hold communities together. It is
as meaningless to describe readers of books as
lurkers as to describe members of sets that way.
Sets are typically great for finding diverse views
and perspectives, inasmuch as the shared attributes
that bind the set together may have little to do with
any other shared values. There are typically few
dangers of group-think, nor of only connecting in yet
another network with like-minded people. Despite
the potential for this diversity they may also reveal
underlying homogeneity that can be used as a basis
of more intensive interaction. It should also be noted
that some shared attributes such as religious belief,
occupation or cultural origins, may be a shorthand
for a cluster of shared attributes or set memberships.
There is a world of difference between the set of
religious fundamentalists and the set of people
interested in learning to sail.
5 SET DISADVANTAGES
5.1 Focus
In order to learn in a set it is normally necessary to
know what one wishes to learn. Unfortunately,
knowing that is one of the most common challenges
faced by a learner. Until one has been immersed in a
subject, it is hard to know what questions to ask, -
what sets to align with. There are some solutions.
Many Q&A forums, for instance, are divided into
categories such as ‘help for beginners’ and
‘advanced topics’, creating subsets with a learning
focus. Similarly, every Wikipedia page supports and
is the focus of a different set. Wikipedia provides
plentiful links within each page to other pages, that a
learner can follow in order to gain a grounding in a
topic as well as to get foundational knowledge in
many areas – exhibiting the learning potential of the
set. However, set-based learning can be
overwhelming unless the learner already knows the
information he or she needs to seek. The paths
through potential answers are multitudinous, so set-
based learning can be circuitous and inefficient.
Moreover, the information that is available may
often be contradictory, and it can be hard for a
beginner to distinguish the good from the bad.
5.2 Depth
Related to problems of focus, set-based learning
typically tends to involve brief exchanges rather than
sustained dialogue. This is fine if one needs an
answer to a programming problem, but not great if
one is seeking to become a medical doctor, where
lengthy study crossing many disciplinary boundaries
may be needed, and where a sustained path may
need to be planned, with dependencies and
prerequisites at every stage. The set may be able to
provide help with constructing or advising on such a
path, but it requires a fair degree of self-discipline,
independence and self-determination to succeed.
Typically, sets may provide help and support but, for
CSEDU2014-6thInternationalConferenceonComputerSupportedEducation
102
longer learning journeys, are often best
supplemented by networks and/or groups. Sets can
provide the seed for these to emerge, with phases of
peripheral participation leading to stronger
involvement with networks of learning partners as
time progresses. However, for set-oriented
approaches like xMOOCs (sets with an interest in
subject X) that often use group-oriented methods
like tight schedules, there may be challenges of
insufficient time for networks to form and learners
needing group support may be set adrift.
5.3 Trust
One of the biggest problems faced by set-based
learners is that anonymity makes it more likely that
there will be trolls, spammers, scammers and other
undesirables. Even when intentions are good, sets
often contain members with limited knowledge as
well as those with too much knowledge, whose
attempts to help may be positively harmful.
Inaccurate or scanty knowledge may result in poor
foundations or wasted work, while excessive
complexity or jargon can be demotivating to
someone trying to make sense of the basics. Division
of set into subsets with greater focus can help here,
as can enthusiastic moderators, but more complex
collective tools are often needed to address this
problem.
5.4 Diversity
Part of the value of sets lies in the diversity of
opinions, skills and interests of set members.
However, this can come at a high price because
people with different cultures, different vocabularies
and different understandings may cause confusion,
upset one another, or fail to communicate
effectively. Sets are fertile ground for flame wars,
angry debate and what some set members will see as
irrelevant or unimportant. At best this can be
inefficient, at worst it will drive people away from
the set. Thus, this strength of diversity is also a
potential weakness of disharmony.
6 THE ROLE OF COLLECTIVES
IN SETS
Given the aforementioned difficulties, learning
within sets can be frustrating, misleading, circuitous
and poor for motivation. Collectives can provide the
missing pieces to replace some of the guidance roles
of the teacher and can make up for the lack of
personal connection and relatedness that occurs in
networks. The general principle behind any
collective is that the actions of many people are
combined, processed and represented, typically as
recommendations, or for filtering, or to structure
information or to suggest a path through it.
Collectives can filter and help make sense of the
information generated by the set (and, to a lesser
extent, the net and the group). For example:
An automated collaborative filter can find
others with similar patterns of interest or
behaviour, and recommend content that may be
of value.
A tag cloud can show topics of interest to a set,
helping to get a better sense of the overall
shape of a subject area and to make it easier to
know what to look for, suggesting other things
that may be of interest.
A reputation system can identify individuals
who have been found to be trustworthy or
knowledgeable within a subject area.
A rating system can help promote good
answers/solutions/recommendations and
demote bad ones.
A data visualization tool can graphically
display activities, actions or ideas of a set of
learners.
A crowd-sourced spam filter can help to
remove content that is injurious or irrelevant
Collectives based on sets can be an embodiment
of the wisdom of the crowd, with relatively few of
the problems that can arise when individuals are
connected or know what decisions others are making
(Surowiecki, 2004). Sometimes, sets of moderately
informed people can outclass experts when dealing
with a range of tasks (Page, 2008). Collectives are,
however, only as smart as the algorithms that
underlie them and the combined wisdom of the
crowds that feed them. This means that they tend to
be susceptible to some common flaws, including:
The Matthew Effect (Merton, 1968), in which
the rich get richer and the poor get poorer, an
out-of-control path dependency that makes it
hard for better novel solutions to gain a
foothold and the rewards priority and
familiarity more than quality.
Filter bubbles (Pariser, 2011), in which we tend
to see things that resemble what we have
already seen, limiting opportunities for
serendipity and discovery of novelty. This is
especially risky for learners who, by definition,
need to enter novel territory.
Lack of pedagogical model, so that it is not
LearningwithStrangers-TheValueofSetsinOnlineLearning
103
always value to learners or even learning that is
valorized in the results. Relatively few
collectives explicitly support learning and most
rely on some variation of popularity or
commonality measures albeit, in the case of
more sophisticated tools like collaborative
filters, with significant personalization.
Intentional abuse, in which mischievous or
malevolent people, especially when working as
in consort, can subvert or overly influence a
system. ‘Google bombing’ and search-engine
optimization strategies are good examples of
this.
Selection bias, in which a distinctive subset of
individuals provides a biased collection of raw
data on which to operate. For example, a
student or a set of experts may fail to consider
solutions to problems that are unconventional,
and so miss some important opportunities.
While collectives have been used to good effect
in an educational setting as well as offering a lot of
value to informal and non-formal learners, it remains
an important research area to find ways of adapting
them effectively to the distinctive needs of learners.
7 CONCLUSIONS
The set is an under-researched social grouping that
we have only recently begun to explore. The set has
increasing importance as we move away from the
familiar formal learning approaches of institutions
that worked well in an industrial face-to-face context
but that do not operate so well at Internet scales, and
that do not cater well for informal or just-in-time
learning. As well as being crucial in supporting day-
to-day lifelong learning, the social form of the set
dominates in large-scale MOOCs. However, many
MOOCs are designed as though they were groups of
a conventional academic variety, with schedules that
assume group-like engagement and commitment,
discussion forums that are often over-populated,
fuzzy in purpose or that assume collaborative rather
than cooperative pedagogies. As a result, they often
carry unrealistic expectations of trust and shared
intent that, in a large and diverse population, are
unlikely to be achieved. This paper has begun to
scratch the surface of how and why we might use
sets for learning, as well as some of the pitfalls that
await the unwary. We continue to research the
differences and to build tools to support sets for
learning. In our forthcoming book, Teaching Crowds
(Dron and Anderson, in press), we explore these
issues in greater detail.
REFERENCES
Daniel, B., Schwier, R. & McCalla, G. (2003) Social
capital in virtual learning communities and distributed
communities of practice. Canadian Journal of
Learning and Technology., 29,
Dron, J. & Anderson, T. (in press) Teaching crowds:
social media and distance learning. AU Press,
Athabasca.
Merton, R. K. (1968) The Matthew effect in science.
Science, 159, 56-63.
Nemoto, K., Gloor, P. & Laubacher, R. (2011) Social
capital increases efficiency of collaboration among
Wikipedia editors. Proceedings of the 22nd ACM
conference on Hypertext and hypermedia, 231-240.
Page, S. E. (2008) The Difference: How the Power of
Diversity Creates Better Groups, Firms, Schools, and
Societies (New Edition). Princeton University Press,
Pariser, E. (2011) The filter bubble : what the Internet is
hiding from you. Penguin, New York.
Salmon, G. (2000) E-moderating: The Key to Teaching
and Learning Online. Kogan Page, London.
Siemens, G. (2005) Connectivism: a learning theory for
the digital age. International Journal of Instructional
Technology and Distance Learning, 2,
Surowiecki, J. (2004) The Wisdom of Crowds. Little,
Brown, London.
Various (2005) What do Wikipedians do all day? 2013,
retrieved from http://en.wikipedia.org/wiki/User:
Statistics#Case_1:_Anon_Surprise.21 12
th
Februar
2014.
Wasko, M. M. & Faraj, S. (2005) Why Should I Share?
Examining Social Capital and Knowledge
Contribution in Electronic Networks of Practice. MIS
Quarterly, 29, 35-57.
Wenger, E. (1998) Communities of Practice: Learning,
Meaning and Identity. Cambridge University Press,
New York.
Wenger, E., Trayner, B. & de Laat, M. (2011) Promoting
and assessing value creation in communities and
networks: A conceptual framework. The Netherlands:
Ruud de Moor Centrum,
Wilson, E. O. (2012) The social conquest of earth.
Liveright Pub. Corporation, New York.
CSEDU2014-6thInternationalConferenceonComputerSupportedEducation
104
