You are on page 1of 16

18 & 19 March 2008,

The Riverbank Park Plaza,


London

Paper 14

CONFESSIONS OF A MODERATOR:
How web communities fail and how
marketers can stop that happening
Tom Ewing
Research International

CONFESSIONS OF A MODERATOR
1
CONFESSIONS OF A MODERATOR: How web
communities fail and how marketers can
stop that happening
Tom Ewing, Research International

Introduction: Unintended Consequences


If I was allowed to dedicate this paper, I would dedicate it to a baby called Harry, who was born last July to friends of mine
who first met on the web community I started in 20001. When I started a music discussion board it’s fair to say that I wasn’t
expecting, or hoping, that it would lead to a friend having a baby. Because online communities simulate and intersect with real
communities, they are subject to the same laws of unintended consequences.

Not all unintended consequences are as dramatic or delightful as a baby, but they are still what make web communities so
fascinating to run. A simple example is ‘off-topic’ content. Online communities tend to come together for a particular purpose
– usually it’s talking about something its members have in common. But as members get to know one another they’ll find
themselves talking about lots of other stuff too. At this point the people running the community have a choice – allow the
people to talk about the other stuff or not. But any successful community automatically has the potential to generate off-topic
material.

This sits outside the parameters of traditional market research - my day job while I ran and took part in web communities.
Research is most often about the generation of intended consequences – answers to pre-set questions; the deliverables and
outputs our clients pay for. Its primary tools, like the quantitative survey or discussion group, are geared to minimize off-topic
material: its participants are not generally encouraged to take their interaction beyond the moment of research.

Recently, marketers and researchers have become interested in web communities, in part because of the possibilities for
open-ness they offer. At last year’s conference, for instance, Mike Cooke and Nick Buckley presented the paper “Right Brain,
Weak Signals” which strongly argued that researchers needed to embrace this open-ness, these unintended consequences
and ‘fuzzy outcomes’, and explore the use of communities as research tools.

That paper indirectly inspired this one. It struck me that as I’d run a thriving online community for several years, I might have
something to say about how they work. So this paper is written from a moderator’s perspective, offering a look not only at the
potential that web communities hold for marketing, but also the stresses and opportunities that arise within most, if not all,
successful communities, since not all unintended consequences are positive.

Freedom And Control: Styles Of Community


First of all, let’s define our parameters a little more. The term “community” can refer to a myriad of web activities – from
auction sites to profile-based social networks.2 For the purpose of this paper I’m talking about communities whose purpose is
discussion – which tends to mean a forum, bulletin board or blog which encourages comments. These are also, I’d suggest,
the type of communities most likely to directly provide the user-generated ideas and insights researchers are looking for.

CONFESSIONS OF A MODERATOR
2
Even narrowing the definition down to this type of community leaves us thinking about sites with very different interfaces, feels,
and levels of content. One fundamental factor in determining the style of a community is the level of control its owners impose
over each of four elements:

• User participation: is the community open to all or invitation only? Do new members need community or moderator
approval?
• Topic: does the community have a central topic subject and how tightly does it enforce that? (One important way of
controlling this, for instance, is to limit who can start new discussions.)

• User content: how much free speech are users allowed? What is the procedure for editing or deleting controversial
material?

• Time: is the community time-limited or is it expected to operate or grow indefinitely?

These elements can change as a community evolves and there are no ‘right combinations’: different communities have
different purposes and needs.

For example, the Guardian’s popular ‘Comment Is Free’ community3 is designed as a space which celebrates and privileges
freedom of speech. It has a high level of freedom in terms of user participation and content, and is designed without a specific
expiry date. However the level of topic control is moderately high – users cannot suggest topics, they can only respond to
postings by the site’s paid writers.

Research Communities
How do communities built for research purposes generally approach these design factors? There’s a split between project-led
and client-led communities – ones built for a specific research objective (like testing new concepts or products) and ones built
as a more general user or consumer panel for a client. In these, consumers of a client product or category are brought together
and discuss their experiences: insights ensue.

Project-led communities are generally time-limited with tight participation and topic control. Depending on the project,
however, the level of freedom allowed for user content may be very high (certainly, censoring material produced in a research
project would be counterproductive!).

Client-led communities are often designed to be permanent and have a lot more open participation. The level of topic control
varies – contributors are often allowed to start discussions but may well be required to stay on-topic. The level of freedom
over user content will also vary – possibly depending on the brand the community presents a public face of. A family oriented
bulletin board run by a food company, for instance, may have a stricter approach to censoring language than one run by MTV
or Harley Davidson.

The effectiveness of these research communities is becoming widely and publically accepted. A recent issue of Marketing
Week, for instance, quoted the example of Del Monte, whose dog owner community convinced the manufacturer that there
was a market for specialist ‘breakfast’ dog snacks.4

Communities Create Cultures


The phrase “client-led” is a slightly loaded one to describe these communities, though. After all, aren’t they really “consumer-
led”? Yes, but the use they are put to is based on a business agenda. As the website for community-building firm Liveworld
puts it, “All communities form a culture, even if left to themselves. The best of them develop cultures pro-actively guided to
engage the members, engage your brand ethos, and meet particular goals. “5

There has been vigorous debate over the level of “pro-active guidance” needed for useful consumer communities to develop:
advocates of radical co-creation would argue that letting the consumer’s voice be heard uncensored is paramount. The
consumer may be critical and angry with a brand, but that’s when she’s most productive to the attentive brand owner.

But consumers can also be angry with one another, or more passionately involved with a competitor. They can be silly or
flippant. They can be entertaining enough that the quality of their rhetoric obscures the paucity of their insights. They can talk

CONFESSIONS OF A MODERATOR
3
about something else entirely, and they can outright mislead and lie. These things happen especially when what someone is
getting out of a community is the pleasure of participation – the culture – as well as the content.

Liveworld’s statement – “All communities form a culture, even if left to themselves.” – is one of the simplest but most powerful
truths I’ve read about communities. The question is – what sort of culture, and are there ways in which the left-to-themselves
cultures might be more useful to researchers than the guided ones? Marketers building a web community are generally most
concerned with getting the set-up right and encouraging initial participation. The question of what happens after this has been
achieved often goes unasked. What happens to a community after a year, or two, or five?

In the next section of the paper I look at the community I founded and its success, and go on to explore what it taught me
about how web communities work in the longer term.

Introducing ILX
I set up the I Love Music message board in August 2000, because I wanted somewhere I could discuss articles from the music
website and blogs I ran. I had no intention of establishing a successful web community – if Blogger comments had been
available at the time, I would have used those. Early in 2001, the high level of off-topic conversation on I Love Music led us to
start a second board for talk not about music – we called this one I Love Everything – and over the years other users started a
whole series of other, usually much smaller boards. The collective name for this whole community of boards is ILX, and unless
I’m referring to a specific message board this is the term I’ll use.

The interface I chose was a free one, on servers maintained by Philip Greenspun, who had created forum hosting software as
an ongoing design experiment.6 I never had any contact with Greenspun, though his design aesthetic led to the distinctively
bare-bones look of ILX which subsequent coders and programmers have kept.

Whether intentionally or not, Greenspun’s ‘LUSENET’ software dispensed with a lot of the features most web communities
consider basic – no icons to quickly visually identify posters; no pictorial ‘emoticons’ to signify mood; no community
hierarchies and easily visible post counts; no division of threads into pages of posts; no facilities to edit or delete one’s posts;
no spellcheckers and HTML checkers. It was also extremely difficult to ban anyone or suspend privileges, though if you were a
moderator you could delete or edit posts and threads.

So I quite unintentionally learned the valuable lesson that function defines content. As a newcomer to ILX, you would
immediately notice several things. You would have no means of knowing who was in charge. You would have no means of
establishing a visual identity for yourself, and no profile page. All you could do was read or get involved. And when you did get
involved, you couldn’t go back and delete foolish posts or edit typos: what you said endured, unless you could persuade a
moderator to alter it. If you could find a moderator.

In other words, thanks to Greenspun’s design choices, ILX was a low-regulation environment. (It still is: though moderators
have more powers, they tend to agonise publically before using them and any banning or deletion is nearly always
accompanied by much spirited debate). This made it a joy when it worked and a headache when it didn’t. It also, as the site
grew, let me observe the development of a web community in a relatively unfettered state. We didn’t interfere much because
we couldn’t: we could only lead by example and try to encourage good behaviour. In terms of the four factors discussed
earlier, we had minimal controls over user participation
and content, we were not time-limited, and we didn’t
Figure 1
enforce on-topic posting with any strictness.

How And Why ILX Grew


A further consequence of Greenspun’s design is that
ILX loaded and ran very quickly. When the site began
growing it became an advantage, allowing it to handle a
greater volume of traffic and new posts.

This graph (figure 1) shows the number of posts per


month to ILX from August 2000 to the present7. It’s worth
pointing out that posts per month certainly doesn’t equal

CONFESSIONS OF A MODERATOR
4
views or the number of times each page would have been loaded – as threads developed, all participants would check them
many times. Generally though, a thread would need to hit several hundred individual posts – all displayed on the same page –
before it would significantly slow or fail to load.

The graph shows a steep upward curve and then an initial levelling off, then more steady growth until a peak in June 2004
of over 118,000 posts in a single month. At this point we have one of several major service outages and the total declines –
though for the last few years it has been stable at between 80,000 and 100,000 posts a month, which seems to be the site’s
natural level.

This level of posting is very large indeed for a web community, the equivalent of eighty 50-post discussions every day. As a
point of comparison, examination of a sample week suggests that the Guardian newspaper’s nationally advertised ‘Comment
Is Free’ site gets around 50,000 posts per month8. Big is not always beautiful in online communities, though, and ILX became
very hard to manage.

ILX reached this level of activity for two reasons. Firstly, its content was compelling enough that existing users wanted to read
more of it and post more often – a very common reason for users leaving the community was that ‘it takes up too much time’.
Secondly, ILX attracted a continuous stream of new users. The site never advertised or solicited for members, so where did
these users come from?

Some were individuals, arriving via search engines or following links elsewhere to interesting threads. Others were people
who knew existing users either offline, or elsewhere on the internet. From early on there was a strong component of offline
interaction among ILX members – regular meetings in pubs; club nights or parties. Members brought friends along, who
started reading the site – so there was a primitive and undesigned ‘social networking’ element to the site’s growth.

Finally and most interestingly, entire groups of people who used existing networks would occasionally migrate en masse to ILX.
The cult pop artist Momus, for example, started spending time on the site and instructed his voluble fanbase to do likewise.
An American message board dedicated to indie rock music, The Fake Matador Bulletin Board, shut down and many of its core
users chose ILX as their new home.

These ‘grafts’ of users who had already formed connections obviously increased the site’s traffic, but they also brought fresh
perspectives and new life to the site’s ongoing discussion. When the FMBB posters arrived, for instance, they quickly identified
an attitude of disdain among existing posters towards a lot of the indie music they liked. They pushed back against this hard,
and the resulting arguments and disputes became part of community folklore.9

In my opinion these arguments – which generated an exchange of ideas as much as an exchange of insults - would probably
not have happened had the posters from the FMBB arrived at ILX individually: they would either have been discouraged and
not begun posting, or been outnumbered and quit. Being part of an existing network, though, gave them the backup they
needed to push through a point of view (though the discussions also brought out their own disagreements and differences).

The Evolution Of ILX


This paper’s focus is on the evolution of communities and of cultures within communities using ILX as an example, rather than
on the specific content of ILX. It’s worth talking briefly, though, about the style used by ILX posters, as it will help illustrate
some more general points about communities. I should stress at this point that while I’m enormously proud of it and the people
who made it so interesting, I am in no sense upholding ILX as a model of how communities should operate – we got a lot of
things right but also a lot of things wrong. Also, ILX was founded and initially used by people who didn’t grow up with the
Internet – a community created from the ground-up for younger users would look and feel different in ways beyond the scope
of this paper to discuss.

The basic unit of ILX, like most discussion boards, is the thread. One poster starts a thread and other posters give their
responses and discuss the topic with one another. Threads on ILX might begin with a question or set of questions, an
observation or theory, a link to a news story or another site, or sometimes just a picture. Some threads acted as their own
miniature sub-communities: threads about the current football season or new R&B releases, for instance, might be continually
‘revived’ (posted to after a period of inactivity) and updated.

CONFESSIONS OF A MODERATOR
5
ILX, unusually, evolved a tradition of reviving and continuing old threads on a topic rather than starting new ones. This means
several ILX threads see multiple periods of activity, which makes tracing the evolution of communication styles within the
community much easier. A thread about The Smiths, for example, was started in 2001 and then saw periods of activity in 2003,
2004 and 200610. Its general trend – which applies across most threads - is towards shorter, less in-depth answers as time
goes on, and more conversational content, with long-term posters commenting on their own earlier replies rather than the main
question, and a greater level of assumed familiarity.

ILX is studded with threads comparing the current state of the community to its earlier days, often unfavourably: the boards
now are supposedly less intelligent, or less friendly, or less interesting than they used to be. The terms “old ILX” and “nu-ILX”
are common currency, though nobody agrees on specific cut-off points for either. These threads tend to be very long and
argumentative and almost nothing gets concluded11 – but they testify to the level of engagement the community fosters and
serve as a vent for individual and collective frustrations. They also demonstrate the fact that the community has evolved a
degree of self-awareness – it thinks of itself as a community. Self-awareness, I’d argue, is part of the way in which a community
evolves a culture, whether it’s a freeform group like ILX or one set up for marketing purposes.

Life Cycles Of Online Users


A major difference between offline and online communities is that barriers to entering and leaving online ones tend to be very
low. So the “life cycle” of an online community is better understood as the aggregate of a series of individual life cycles within a
community.

Amy Jo Kim, author of Community Building On The Web, has proposed a model of these life cycles based on participation
in learning communities. A community member begins as a lurker, who is aware of and reads the community but does not
participate. They then become a newbie – a new member of the community, then a regular, who participates fully and is
recognised and accepted by other community members. The next stage is becoming an elder (or expert) – someone with a
level of authority within the community, perhaps with moderating powers. And the final stage is becoming a legacy member –
someone who is in the process of leaving the community, or who has in fact left it. A growing community is simply one which
has more newbies (people arriving) than legacy members (people leaving).

Kim’s model provides a useful framework for thinking about online life cycles, though it should not be treated as linear. A
community member can move from any stage to ‘legacy’ – not all newbies become regulars and not all regulars become
elders. The model also doesn’t address individuals who leave communities and then return to them – as the founder of ILX, for
instance, I was undoubtedly an elder on that site. But having not regularly contributed for a year, if I was to start posting again,
I would still have elder status for some regulars (who knew me) but not for others.

In fact, the newbie-regular-elder divisions are quite problematic, because posters can move through these stages at wildly
differing rates, and be accepted at different times by different other posters. A lot of community designers seek a practical
solution to this by coding hierarchies like these into the community, with publically displayed counts of how often a poster
has contributed, or public ranking systems. This can exacerbate other problems, as we’ll find when we look in more detail at
community elders.

Kim’s model is most useful because it reminds us that the experience of a community is non-uniform. A regular poster and
a newbie will react differently to new stimuli. On a music board, for instance, a question about which Beatles record is your
favourite may well excite a newbie, who has not given their opinion before, and bore or annoy a regular, who has seen similar
topics many times. In fact there’s obvious potential for conflict between the stages in any long-term community: handling this
conflict is one question anyone running a community needs to consider.

Content Motivation And Social Motivation


There’s another way of looking at participation in communities, which is to consider what the participants get out of it.
Community theorists have detailed several benefits that successful participation in a community offers its members, but from
my experience I would propose that these can be grouped into two main areas: content based motivations and social based
motivations.

The first can be roughly boiled down to “these discussions are great!”; the second to “these people are great!”. Generally,
content motivations are what attracts someone to read and initially contribute to a community – I’m counting the desire to

CONFESSIONS OF A MODERATOR
6
add one’s own expertise or opinion as a content motivation. Social motivations begin to arise when the user has started
contributing.

The two are obviously linked rather than necessarily opposed – it’s not a case of stressing one over the other when developing
a community. On ILX the I Love Everything bulletin board was essentially started to cater for the social motivations of posters
on I Love Music. But the socialisation depended on them producing interesting content for one another too. And the content
motivation to provide expertise is clearly related to the social motivation of increasing one’s reputation in a community and
winning friends. There are also communities – for new parents, for instance, or online boards for existing offline communities
like towns or districts – where content rests on the understanding of a shared social context, and there’s a strong social
motivation to participate.

But I would argue that for most members of online communities, the life cycle of their participation involves a shift in emphasis
from content motivations to social motivations. On the one hand a member gets to know the other people in the community
and begins to enjoy their virtual company. On the other – depending on how narrow the core topic is – she becomes more
familiar with their opinions and the information they offer, so is less attracted to the content they provide.

This is at the heart of what I mean when I talk about communities evolving a culture: the pleasure of participation comes to
lie not solely in the content but in the community itself, which in turn leads to the community developing self-awareness, and
reinforces the culture that makes it a community.

One way in which this can happen is through in-jokes and ‘memes’ – repeated ideas that spread virally through a community.
In the early days of ILX, one prominent poster would occasionally semi-ironically adopt the “hacker-speak” slang used by
some gamers and coders. This included adding the modifer “-XoR” to words, so “It rocks” became “It roXoR”. This occasional
usage caught on across the community, and when ILX came to acquire its own domain name, it picked ilxor.com.

Jokes and memes may well build and encourage participation in a community – using them is a way for new members to
show they understand the culture. But they can also be alienating – too much socially motivated posting can turn off content-
motivated newcomers, which is one reason a lot of communities discourage off-topic posting.

How Culture Effects Content


In any community that isn’t strictly time- or goal-bound (and in many that are) a culture based on social motivation will start to
emerge. The question for communities set up by brand owners or researchers is: what does this do to content? After all, they
generally didn’t start a research community in order for people to make friends, they started it to get worthwhile insights and
information for their clients.

The emergent social culture may have an impact on the amount of content produced, as more of contributors’ energies go
towards interaction with other members – but this is balanced by the fact that socially motivated participants may well be more
engaged with the community, and spend more time there.

Can an emergent community culture affect the quality of content? I would argue that the answer is yes, in quite subtle ways.
The political theorist Cass Sunstein, in his book Republic.com 2.0, details an experiment where two groups of Americans,
one self-identified as liberal and one as conservative, were convened to discuss various social issues. They wrote down
their opinions before and after the discussion took place: in both cases the groups showed markedly less diversity after the
discussion, and in both cases the direction of agreement had moved towards extreme rather than centrist viewpoints.
Sunstein’s point is that lack of exposure to opposing viewpoints means that discourse within a strand of opinion tends to
become homogenous and extreme, and he sees this happening within online political communities. Does something similar
apply within non-political communities, especially as the community develops a level of self-awareness?

It’s worth pointing out that if you’re a marketer owning a community dedicated to a brand, this homogenising effect is a
potential benefit of a community, a way in which it can increase customer loyalty. But if you’re a researcher trying to run a
community as a way to produce consumer insight or innovation, it might be a curse.

Is Metal Machine Music The Best Lou Reed Album?


The I Love Music board on ILX provides an example of how a self-aware community can generate surprising or extreme points

CONFESSIONS OF A MODERATOR
7
of view. When ILX was redesigned in 2006 the new programmers brought in a “poll” function – allowing users to tick simple
multiple-choice options, with the results concealed until the poll ended. Sample sizes are very small and no quantitative
researcher with a hint of self-respect would go near these polls but as indicative data they can be interesting – they give a
broad sense of where the community’s sympathies lay.

A particularly intriguing poll asked which the best Lou Reed album is. Lou Reed, founder of the Velvet Underground, has made
solo records for almost 40 years. His best known album is Transformer, which contains his only hit song, “Walk On The Wild
Side”. He’s made a raft of other critically acclaimed albums, though, like Street Hassle or Berlin. Music connoisseurs – like ILX
members – might pick one of those instead.

They did not. Instead they chose Metal Machine Music, a 1975 record Reed reputedly made in an attempt to break his
EMI contract. MMM is four sides of raw electronic feedback and noise. It’s been called the worst record of all time, but it
comfortably won the ILM poll.12

Imagine you are a researcher working for Lou Reed’s record label, who is observing this community looking for insight. What
did this indicative result actually indicate? It might, from an insight hunting perspective, have meant one of several things.

i The community believes MMM is Lou Reed’s best album, because it is (i.e. their views are representative).

ii The community believes MMM is Lou Reed’s best album, but their expertise in the topic is distorting their
preferences (i.e. their views are legitimate but unrepresentative).

iii The community does not necessarily believe MMM is Lou Reed’s best album but thinks it is more interesting,
provocative or amusing to say it is.

If you were in the – perhaps thankless – position of marketing Metal Machine Music, it would matter very much which of these
was the case. If (i) was true, you might promote the record across the whole potential audience. If (ii), you might promote it to a
specialised audience. If (iii), you might try to forget the bloody thing existed.

If you’re an observer of the community you could probably work out quite quickly whether to dismiss option (i). But to decide
whether (ii) or (iii) is the case you would need to know the community in a degree of depth – how it defines itself and who
the prominent contributors are. Your interpretation would be enhanced by knowing, for instance, that there had been several
previous discussions about Metal Machine Music13, that a lot of contributors listen to other ‘noise’ music , or that a significant
proportion of regular contributors find the polls inane compared to longer discussion and so see value in sabotaging them.

My general point, though, is that once a community reaches a level of self-awareness, an observer can no longer take the
good faith of members or respondents completely for granted. But if this is so, why is it?

Communities And In-Group Bias


Researchers are very familiar with the idea that an individual’s preferences and opinions also contain a social element – when
we buy a brand, our perception of it also includes the other people we believe buy it and the ones we believe don’t. This also
applies at community level, particularly online where the boundaries of a community are generally non-physical. There is a
need to differentiate between people in the community and people outside it, even if initially that difference is simply “we like to
discuss this thing more than most people do”.

Psychologists recognise this phenomenon as in-group/out-group thinking, and as a culture emerges from the community, it
becomes more pronounced. On ILX, for example, the stock figure of the ‘12-CD guy’ emerged14 – individuals who only buy a
small number of new records in a year, stereotypically including people who bought music by the likes of Coldplay who most
ILM posters disapproved of. Market researchers and marketers have long known to take in-group bias into account when
thinking about attitudinal responses among different segments of society, but the rise of ongoing communities as a research
tool means that it has the potential to seriously affect results across a project, as “those taking part in the project” itself
becomes an in-group.

CONFESSIONS OF A MODERATOR
8
So is the choice of Metal Machine Music a result of in-group bias – the ILM community wanting to differentiate itself from
external opinions? Unfortunately it isn’t that simple – if we revisit Amy Jo Kim’s model of an online community, we see that any
community also contains internal in-groups and out-groups – most notably newbies, who have just begun to contribute, and
regulars and elders, who have been contributing for a long time.

Internal And External Out-Groups


So an individual acting in a community can orient themselves
on multiple axes. There is the extent to which they identify with
perceived external opinions and norms, and also the extent to
which they identify with internal community opinions, defined
(or perceived to be defined) by community elders. This quadrant
illustrates these axes, and also suggests the different roles these
individuals might play in a community.

Users who identify with community opinions but not with external
ones will serve to reinforce the core, differentiated identity of
the community. Those who identify with both community and
external opinions may want to build bridges with those outside
the community, and minimise the perceived differences: you are more likely to find users in this category when the differences
between the community and the non-community out-group are small or cosmetic.

Those who identify with external opinions but not with community ones are likely to be hostile to the community and question
its validity (we will talk more about this group when we come to consider ‘Trolls’). And finally, those who do not identify with
external opinions or with community opinions may be hostile, or may seek to influence the community opinions into ones they
do agree with.

Earlier I referred to how the growth of ILX was fuelled partly by ‘grafts’ from other communities, which sometimes caused
arguments with the existing membership. Where this happened, it’s an example of the fourth type of contributor, identifying
neither with community nor external norms and helping to refresh and shift discourse in the community.

However, these grafts were themselves small and tended to occur while ILX was still a small community. The larger the
community, the more significantly different in-groups may be able to form within it rather than arrive as immigrants. A turning
point in the history of ILX came with the formation in 2004 of the ‘Noize Board’, a separate message board set up by a group
of newer, mostly younger regular posters who wanted their own space within the ILX set-up.15 ‘Noize Board’ threads were more
anarchic and often a lot funnier and more vibrant than other ILX threads, with shorter responses and a much higher proportion
of pictures and cutting one-liners. The new board had a divisive effect – almost all of its members still posted on the main
boards, which gradually became closer in style to the ‘Noize Board’. But at the same time several threads on the new board
were devoted to mockery of activities and posters from the main boards, which upset a lot of other regular posters.

The ‘Noize Board’ is also a collective example of the “loose cannon” contributor who can effect change in a community. They
set up their board in a brief period when any ILX member could start a whole new sub-forum, rather than simply new threads.
I know of no other online community that has tried this and I wouldn’t recommend it. But the “Noize Dudes” are only unusual
in that they were allowed to successfully formalise their status as an alternative in-group: such groups are likely to arise in any
community that reaches a certain size. Alternative in-groups, whether internal or grafted, can have immense benefits in terms
of keeping dialogue fresh and questioning in-group thinking, but they also carry risks.

In-group thinking, alternative in-groups, and the potential distortions and disputes that arise are issues a community moderator
needs to deal with. But what do we mean by a moderator, and how involved should they be in a research community?

Power And Authority


Within any community, there are individuals with power and individuals with authority. Often – but not always – the same
individuals.

Power in this formulation refers to the ability to make and enforce changes within the community. Authority refers to an

CONFESSIONS OF A MODERATOR
9
individual’s ‘standing’ within a community – the extent to which they set the tone for interaction and the extent to which their
commentary or recommendations may be listened to.

In offline communities the lines between powerful individuals and authority figures tend to blur. A member of the police
certainly has power, but the level of authority they have depends on other community norms. A vicar or social worker has very
limited power but is expected to exercise a degree of authority upon those they interact with.

In online communities the distinction is more obvious. Some powers – the ability to post and reply, and commonly the ability
to start new topics – are general. Some, though, are exclusive – the ability to edit or delete others’ posts, the ability to block
discussion topics, the ability to ban users completely.

Authority is a different matter. It certainly operates in online communities: in Kim’s model, authority is held by “elders”, or
‘experts’ or ‘core users’ – high-level users who are more respected, who reply regularly to other discussions and whose own
posts generate more attention and replies.

Where Elders Come From


But where does this respect come from – what makes an elder? The word implies length of service in the community, which is
often but by no means always the driver of online authority.

The factors which I believe play a part in creating community ‘elders’ online include:

Length of service/number of posts: These figures are public knowledge on many discussion-based communities, appearing
on a profile page or alongside every post. This is partly because they are easy to measure, but the weight of contributions they
indicate seems also to imply a commitment to a community.

Quality of posting: Frequency of posting need not correlate to quality of posting – some users may post less but the quality of
their contributions is such that they win a level of respect and influence. A particularly incisive newcomer, for instance, may be
able to win recognition quickly and an alternative in-group may even form around them.

External reputation: It is possible to advance straight to elder status, if your offline reputation in a topic area warrants it. If
Steve Jobs were to start casually posting to Apple sites, for instance, his low post count would hardly prevent him acquiring
instant respect. Of course, this is precisely why he would anonymise himself if he did decide to do such a thing – this thought
experiment helps illuminate the potentially distorting effect of elder status.

Idiosyncracy of posting: Finally, I would contend some people can achieve influence without necessarily commanding respect.
A regular poster with particular idiosyncrasies of style or opinion can become a core figure in a community without necessarily
posting a lot or being taken seriously.

The big question for people running an online community is – to what extent should the people with the power also be the
people with authority? In other words, should moderators participate in a community and attempt to command elder status, or
should they simply be administrators and not actually take an active part in discussions?

This is a particularly acute question in a research community. Here researchers will generally need to do some of the things
elders do – such as starting discussions – in order to more efficiently get the information they need and help keep the
community vibrant. On the other hand, to participate too fully would compromise their objectivity. But there’s another risk – on
communities where moderators don’t actively participate, they run the risk of becoming seen as an out-group: “the mods” who
hinder conversation as much as help it, and exist to be snuck around.

My opinion is that moderators should be fully involved in their communities as elders – not only to avoid this alienation from
group practise, but also because immersion will allow a researcher to better understand the in-group and out-group dynamics
which develop as a community culture evolves and will help in interpretation (as in the Metal Machine Music problem). An even
better, though more labour intensive solution would be to employ different researchers – some fully immersed in the community
to act as discussion leaders and stimulators and gain authority, others with power to enforce community rules. This might free
the ones with authority from accusations of bias.

CONFESSIONS OF A MODERATOR
10
But a researcher within a research community mustn’t simply be an elder – there is another community figure we need to
consider, one outside the five stages in Kim’s model of participation: the troll.
Trolls And Trolling

As the name indicates, the troll is a somewhat demonised figure in online culture: the word refers to individuals who post
to communities in the attempt to elicit an emotional or instinctive response and the verb “trolling” is the act of doing this. A
very simple form of trolling would be to visit a thread on a Research 2008 bulletin board and suggest that the conference is
being secretly run by the Freemasons – a lot of trolling is simple shock tactics, often personally abusive or racist. Almost all
administrators will react quickly to this kind of troll and delete their postings.

More complex, and more difficult to deal with, is trolling where it isn’t immediately obvious that the poster is a troll. As we’ve
seen, communities develop in-group opinions and thinking, so a poster arriving who challenges this thinking can often be
labelled a troll. If a poster on the Research 2008 board posted messages stating that all market research was bunk and
business people should rely on transactional data and intuition, they would not be stating a self-evidently indefensible opinion
but they might well be trolling: looking to annoy and provoke regular posters who have clearly agreed to agree that research is
useful.

Trolling is frowned upon because it encourages posters to get angry, it encourages them to repeat themselves or waste time
retreading old discussions (the successful troll never concedes an opponent’s point) and it diverts communities away from
content-led questions and into the pointless question of “Is X a troll and what can be done about them?”. This is why most
moderators will act quickly to ban trolls, and where banning them proves difficult will encourage other users to ignore them.
Trolling is a persistent issue on every open community. On communities run by invitation they are a lot rarer but still worth
thinking about, since not all trolling is entirely negative.

The Power Of Positive Trolling


As with every other element of online community, a troll’s activities can have unexpected consequences. On ILX, an early
thread about the rock band Killing Joke16 provoked an invasion of our message board by denizens of a Killing Joke board,
who were unimpressed both with our dismissive attitude towards their favourites and our appreciation of R’n’B and Top 40
pop. One of the Killing Joke fans hung around when his fellow trolls had departed, and would pepper threads with their slogan,
“Honour the Fire!”, and with hostile invective against our favourites. Gradually, as he got to know community members, this
individual became a regular poster, started more productive discussions about the other music he liked, and – while not
changing his original opinions – is now certainly an ILX elder.

In another example, a poster calling themselves “Becky Lucas” appeared on ILX in early 2003, posting innocuous questions at
first and then very sexually frank ones, while abusing anyone who replied or challenged her17. Banning her proved technically
very difficult, so the moderators took the extreme step of placing a message warning posters not to interact with her on every
single page of ILX – a message visible to all but “Becky”. Eventually being ignored caused her to drift away. The incident
entered into board legend and served as a kind of unifying, bonding experience for a usually-fractious community.

No research community would or should allow a Becky Lucas situation to happen, as banning troublesome users would be a
lot easier. But the two examples suggest firstly that trolls can ‘reform’, and secondly that troll attacks help a community evolve
a culture and history. Trolling can also be harmless and creative – one individual on ILX posed for a long time as a 50-year-old
female Chris De Burgh fan, unfailingly positive and polite, partly in order to expose the hypocrisy of other posters’ prejudice
towards that artist.

Positive trolling is rare, but it’s worth discussing the topic because the lines between trolling and non-trolling are so blurred. At
one level, almost all community behaviour is attention-seeking and performative: posters post partly because they anticipate
reciprocal posting. So community owners should be careful before making a definitive identification of a troublesome poster as
a troll.

It’s also the case that the role of a researcher on a discussion board can be very similar to that of a troll. The researcher’s
motivations are more benign, but they are also attempting to elicit emotional reactions, sometimes challenging assumptions
and making disruptive interventions. Toleration of trolls is probably a bad idea for a research-oriented community, but
researchers should look carefully at how trolls operate and successful trolling tactics, in order to modify them to create more

CONFESSIONS OF A MODERATOR
11
stimulating and productive discussions.

When Elders Turn Troll


One trolling tactic (which is explicitly designed not to lead to productive discussion) is to post a question on a topic which
the community has ‘discussed to death’ many times before, but that is known to cause arguments. On a Microsoft board,
for instance, the question “PC or Mac?” might prompt most regulars to roll their eyes and ignore it. But some newer posters,
excited by an apparently fertile discussion topic, will respond. This was known as “trolling for newbies”, because it was
specifically practised by community elders on some very early web communities – in fact these were some of the first trolls to
be described as such.

So the respected elders and the despised trolls have, in some circumstances, been two faces of one coin. This shouldn’t be
particularly surprising, if we remember that social motivation, rather than content motivation, is likely to motivate many core
community members, and “trolling for newbies” is a practise expressly designed to reinforce social connections – it’s funny
because we all know that anyone answering is new – at the expense of content discussion.

Elders are important for the functioning of a long-running community but they are not an unalloyed good. They can become
jaded, uninterested in the content of the community but continuing to commentate and contribute because they still enjoy
the social elements of the community. This leads into the fifth phase of Amy Jo Kim’s model of community participation: the
legacy stage, in which the member reduces and then terminates his involvement. Community elders can have extremely “long
goodbyes” and the contributions they make during the legacy stage are not always helpful – when divisive threads appeared
on ILX about how the community was declining, they tended to be started by long-term posters. The legacy phase is the
reason why I personally distrust a very common design feature on communities – publically visible membership lengths or post
counts as a way of measuring reputation. A poster in an unconstructive legacy phase, after all, will still have a very high post
count.

Making Communities Work


I argued earlier that in order to fully understand community discourse a moderator needs to fully participate in it. High levels of
moderator participation will also allow researchers to spot potentially insight-rich conversations more quickly and intervene in
conversations in productive ways, potentially counterbalancing many of the problems in community evolution I’ve outlined.

What kind of interventions work best? Any researcher used to moderating group discussions will find that they have an
armoury of techniques and exercises which can be ‘ported’ online quite easily: some will work better than others. Community
discussions, though, are not time-limited and often take place simultaneously across multiple threads – a participant can hop
easily from one topic to another, responding to the latest information on many different threads. So techniques designed to
refocus a conversation or move it along more quickly should be used with particular care.

This is a list of some of the ways I (and others) sparked conversation most successfully on ILX, which might well work on other
communities. A lot of them involved standardised thread formats we created to push conversation in productive directions.
Almost all worked best when the questioner would lead by example, giving their own answer and reasons to encourage others.

“Classic Or Dud”18 – asking whether something (like Led Zeppelin, or tomato ketchup) was ‘classic’ or ‘dud’ created a forced
binary. It segmented board users immediately into opposing camps on a topic, opening up more possibilities for disagreement
and discussion. It also annoyed posters who had more nuanced or moderate views, which seemed to provoke them into
explaining why a particular thing was neither classic nor dud, rather than just shrugging or ignoring the thread.

“Taking Sides”19: Another forced-binary tactic, which worked best when two related things were compared. Again, direct
competition seemed to make individuals examine their opinions more closely and go into more detail. The format could also
work as a way to draw comparisons between apparently dissimilar things.

“Search and Destroy”20: Asking contributors to decide what to search out and what to destroy within a category (eg. Madonna
singles). The initial rule was that if you gave an opinion on the one you had to give one on the other, and so responses like
“Search: all of it!” were frowned upon. Variants of this, like “Pick Only Five” were less successful because they didn’t involve
the ‘Destroy’ element of negative selection, and so quickly became lists with no justification offered.

Concern Trolling21: An example of where a moderator can learn from troll techniques: concern trolling, often used on political

CONFESSIONS OF A MODERATOR
12
messageboards, is a tactic where the troll poses as a supporter of something while asking “just a few questions” designed
to undermine or challenge genuine supporters. Professing indecision works just as well: “I’m not sure about this, what’s the
appeal?” generates more insightful responses than simply asking what’s good about something, because it humanises the
questioner and creates another reason to respond.

Events and Rituals: Special events and regular features can both stimulate discussion and reinforce community ties. An
example of this comes from the community I’ve founded since leaving ILX, a much smaller music group called Poptimists22,
whose focus tends to be on the discussion and discovery of individual new tracks. For this community I’ve run events like the
Pop World Cup, where individual members are given a country from the FIFA World Cup and tasked with finding music from
that country for the community to vote on. (England, inevitably, went out in the quarter finals).

Polls: Most online community software now includes polling facilities, and these can seem an easy way of stimulating
participation. My experience, from observing the ILX community before and after the introduction of polling facilities, is that
while they increase participation they dampen the depth and quality of response compared to forced-binary exercises –
members are happy to register their preference anonymously and leave without discussing it. This is a poor trade-off for some
indicative data so my advice is to use polling functions with great care. An exception to this is where a community can see the
poll results unfold in real time – this can be very engaging and spark insightful advocacy for posters’ preferred choices.

On-Topic And Off-Topic Content


In this discussion of how long-form communities evolve unique cultures, you might argue that I’ve painted a fairly negative
picture. I’ve stressed how communities lead to in-group thinking, how the stages of community participation create
divisiveness, how trolling undermines communities and how even community elders can’t necessarily be relied upon to enable
worthwhile content. I’ve also underlined how keeping a community producing fresh content involves a great deal of work and
direct participation. So you might be forgiven for asking: why start one in the first place? What makes long-term communities
any kind of worthwhile investment for a researcher?

I’ve purposely focused on the downsides of communities not because I think they don’t work as a research tool – I believe
in them passionately and love working with even very limited ones – but because the benefits have been explored more
extensively. Research communities offer the opportunity to communicate with and understand consumers in a much more
natural, interactive space than the more artificial zones created by standard techniques. They can produce a wealth of
information very quickly, and they can enable powerful co-creative processes.

There’s another benefit, though, which emerges directly from how a community develops its own culture. Right at the
beginning of the paper I talked about off-topic content, which I’d argue is a by-product of a community’s evolution into a
culture: the stuff people talk about in order to get to know fellow members better. Not all web communities allow off-topic
discussions, and even when they do, not all communities find that it becomes a large part of the community’s activity. What
good is it?

That depends on your perspective. Imagine you’re a food brand owner and the people on your user community suddenly stop
talking about recipes and have a fascinating conversation about mobile phone tariffs. There’s no reason for you to stop this
conversation, but there’s not a huge amount you can do with the information either.

Now imagine you’re a research agency in the same situation, and the research community you’ve set up, which you allow
clients to access, goes off-topic in similar fashion. The conversation about mobile phones may not be of any interest to your
immediate client, but you can tag it and analyse it later in case it contains insights that another client might appreciate. One
person’s off-topic is often another’s key area of interest.

Do long-term communities often go off-topic in productive ways? My experience with ILX suggests that they might. The reason
lies in the way they stimulate consumer involvement.

Involvement And Talkability


One reason marketers have become interested in online communities is the apparent solutions they offer to the problem of low
consumer involvement. In their stimulating paper on co-creation from Research 2007, Co-Creation Rules, Cherkoff and Moore
listed as one of their rules – “Listen to your 1%ers” ­– those being your most passionately loyal and involved consumers, the

CONFESSIONS OF A MODERATOR
13
ones you know your product inside out and will advocate for it. The example they give – and the source of the 1%er name - is
Harley Davidson, who transformed their commercial fortunes when they embraced, rather than demonised, their most loyal
users. The same impulse lies behind a lot of online user communities.

But do all brands and products create 1%ers? Harley Davidson is a great case study but vehicles are a category with a very
high average level of consumer involvement. So are food ingredients (consumers love to swap recipes), sports, music and
fashion brands, and technology brands: all these will certainly have 1%ers, but will a brand of batteries, white goods, or
toothpaste?

If they do, online communities might be the way to find them. But involvement is only the first hurdle. I am heavily involved in
my lunchtime sandwich purchase – I think about it a great deal but I don’t usually talk about it in the way I might talk about a
new gizmo I’ve bought. There’s a difference between involvement and talkability. Even if you are a 1%er for Fairy Liquid, what
sort of conversations would you hold about it?

The obvious way to get round this by building a community around social roles or activities in which their brands come into
play – a community for homemakers, for instance, might well bring out laundry insights. But only if its members are engaged
enough with the community to want to share those insights, which is why building long-term communities, and allowing a
community culture to emerge, is so important.

My experience with ILX, and other communities I’ve been involved in, is that they hit a sweet spot between content and social
motivation which leads to both involvement and talkability increasing for almost any given topic. The path runs:

1. These people talk interestingly about stuff that I’m involved with,
2. The more I interact with them the more I like them.
3. So because I like them I am happy to talk with them about stuff I don’t usually care about.

Practically speaking, this meant that ILX threads on mobile phone tariffs, cereal eating, personal grooming rituals or a host of
other topics received the same detailed and intelligent answers as threads on music or football, or other more typically high-
talkability subjects23. Reading ILX would often feel like reading reports from fifty or so funny, smart focus groups, all happening
simultaneously and spontaneously.

Of course, this required a level of tolerance for off-topic discussion in the first place. A community devoted to a single brand
may well reinforce involvement and talkability for that brand, but the benefits of increased involvement and talkability on any
topic would only emerge in a community that allows, or even encourages, wider topic posting.

Darts In The Air: The Lessons Of ILX


Was ILX a fluke? This question is the elephant in the chatroom. It was – and is – by no means a typical online community.
What I’ve tried to suggest in this paper is that the problems and virtues of the community I helped run arose from the way a
culture emerged within it, as well as from the specific and unusual moderating decisions we either made or had imposed on
us by technology. Culture emerging within a community is inevitable; the consequences of that emergence can and should be
managed.

The experience of ILX taught me some valuable lessons.


1. “General”, topic-free communities can be a source of insight: Though ILX began as a music community it came
into its own when its music fans were liberated to talk about everything else, and this produced much of its most
interesting and insightful material.

2. More freedom for users makes unexpected, insightful content more likely: in other words, the co-creation
movement is right, though not all unexpected content is beneficial.

3. Moderators will understand and influence community behaviour more if they are active participants:
as active, empathic participants they can gain more respect and (if necessary) set a desired tone for discussions.

4. Disruptive contributions can be positive: if new members, trolls and ‘grafts’ onto the community help to

CONFESSIONS OF A MODERATOR
14
challenge in-group bias rather than reinforce it.

5. A
 dynamic community balances the interests of competing in-groups: features that enforce hierarchy (like
reputation metrics) may well be counter-productive.

Research agencies, in my opinion, should be looking at communities not only as things they can build for clients, but as
proprietary resources like panels: they should be building research communities of their own which can add value to individual
projects and ongoing relationships. The practicalities of these potential communities – their platforms, their incentives, their
funding, the resources needed to run them – are beyond this paper’s scope, but I’ll end with two final thoughts on the kind of
members and moderators they might need.

The first is that members of a research community should be carefully recruited. Just as you might recruit particularly loyal
users, or category experts, or highly creative consumers for particular project needs, so at least some of a community’s
membership should be recruited for communication skills. Articulacy, humour, imagination, empathy, and tolerance: not
everyone in the community needs to have these traits but some do. Luckily, most existing communities already include such
individuals. Recruiting people with experience of online communities would also mean recruiting people which might have
more spare time or opportunity to contribute – always an issue with community projects.

A moderator needs all these traits too, of course, and more standard research skills: the ability to interpret, analyse and spot
patterns. They also need patience and an appreciation of surprise. My fellow ILX moderator, Mark Sinker, enjoys referring to a
Roadrunner cartoon in which Wile E Coyote begins by hurling a host of darts into the air, hoping to spear Roadrunner. These
darts fall back to earth throughout the episode, affecting the action when Wile E least expects it. It’s a good metaphor for
running a community: throw darts in the air, all the time, and wait for the unexpected to happen.

CONFESSIONS OF A MODERATOR
15
BIBLIOGRAPHY
Cherkoff, James and Moore, Johnnie, Co-Creation Rules (2007, MRS Research 2007 Paper)
Cooke, Mike and Buckley, Nick, Right Brain, Weak Signals (2007, MRS Research 2007 Paper)
Duparcq, Patrick, Community Design: The Four Principles Of Community Management (paper, Kellogg School Of Management,
Northwestern U)
Greenspun, P, Philip And Alex’s Guide To Web Publishing (1998, self-published)
Kim, A J, Community Building On The Web (2000, Peachpit Press)
Sunstein, Cass R., Republic.com 2.0 (2007, Princeton University Press)
ILX www.ilxor.com (2000-2008) – most of the specific citations from this web community are given in the endnotes.
Further explorations of these and related topics can be found at http://www.blackbeardblog.com

ENDNOTES
1 www.ilxor.com. In this paper I have chosen not to provide extensive quotations from ILX threads or from any other
communities, preferring to cite the URLs of specific threads in the endnotes.
2 Patrick Duparcq’s short paper on designing virtual communities gives a good overview of this wider definition.
3 http://commentisfree.guardian.co.uk/index.html
4 http://www.brandrepublic.com/InDepth/Features/777205/Market-research-Listen-learn/
5 http://www.liveworld.com
6 Greenspun is still involved with community management and design – see http://philip.greenspun.com/research/
solar-magnitude-forum for his latest ideas.
7 Source: ilxor.com “Statistics” page, January 28th 2008.
8 Calculation based on posts to the Comment Is Free site made week ending January 26th 2008.
9 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=6747 (“Indie Guilt”)
10 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=438 (“The Smiths: Classic Or Dud”)
11 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=5488 (“Has ILE Gone Down The
Dumper?”) gives a flavour of these conversations.
12 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=57296 (“Best Lou Reed Studio
Album”)
13 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=10521 (“Metal Machine Music, C
or D”) is one of the 11 threads about it.
14 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?showall=true&bookmarkedmessageid=15&boardid=41&th
readid=652 (“ABBA – Classic Or Dud?”) is its first appearance.
15 http://www.ilxor.com/ILX/NewAnswersControllerServlet?boardid=60
16 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=453 (“Killing Joke – Classic Or Dud,
Search And Destroy”).
17 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=13771 (“Becky Lucas – Classic Or
Dud”) is an excellent example of how not to deal with an attention-seeking troll.
18 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=130 (“Tomato Ketchup: Classic Or
Dud?”)
19 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=1037 (“Taking Sides: Pubs vs Bars”)
20 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=166 (“Fast Food: Search And
Destroy”).
21 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=41&threadid=27989 (“Help Me Understand Why
Reissues And Remasters Are So Important To So Many?”)
22 http://community.livejournal.com/poptimists/
23 http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=23884 (“Mobile Phone Networks
– S/D”), http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=53251 (“This just in: Rice
Krispies too buoyant”), http://www.ilxor.com/ILX/ThreadSelectedControllerServlet?boardid=40&threadid=8446
(“Cleanse, Tone, Moisturise”)

CONFESSIONS OF A MODERATOR
16

You might also like