Content uploaded by Andreas Girgensohn
Author content
All content in this area was uploaded by Andreas Girgensohn
Content may be subject to copyright.
ABSTRACT
Technology can play an important role in enabling people to
interact with each other. The Web is one such technology
with the affordances for sharing information and for con-
necting people to people. In this paper, we describe the
design of two social interaction Web sites for two different
social groups. We review several related efforts to provide
principles for creating social interaction environments and
describe the specific principles that guided our design. To
examine the effectiveness of the two sites, we analyze the
usage data. Finally, we discuss approaches for encouraging
participation and lessons learned.
Keywords
Awareness, common ground, communication, cooperation,
participation, practice and experience, social interaction,
usage data, usability, Web-based collaboration.
INTRODUCTION
Developments in information and communication technolo-
gies have increased the opportunities and means for cou-
pling technology with certain social practices (e.g., town
hall meetings, support networks) that would influence peo-
ple’s ability to act together. In addition to the immediate
outcome from such technology-mediated interactions,
Resnick [21] has suggested that, as a side effect, such inter-
actions could create productive resources — socio-technical
capital1. The capital may consist of artifacts created from
the interactions or relationships and practices developed
through repeated social interactions. Such capital can enable
future social interactions. Use of the capital by a group to
act collectively may not simply consume the resource but
can reproduce more of it.
The Web is one such social technology. It enables people to
share experiences, to expose tacit knowledge, to make rec-
ommendations, and to discuss a wide range of topics [22,
26, 27]. It also enables people to get to know other people
and to seek out affiliation, companionship, and support [5,
17, 16, 22]. The Web enables people to work together, to
conduct economic transactions, and businesses to build per-
sonalized relationships with their customers [13, 15]. Many
of these social by-products are achieved by leveraging affor-
dances of the technology for information dissemination,
communication, support, self-service transactions, and
mediation. The combination of these Web functions and
social processes can jointly influence the ability of people to
act together in productive ways such as the creation of
socio-technical capital [21].
Digital cities [9], community networks [24], portals such as
eBay and Slashdot, and older technologies based on elec-
tronic bulletin board systems, newsgroups and mailing lists
[22, 23] are just some of the examples of social interaction
spaces. This interest in social interaction spaces is not lim-
ited to work contexts but extends to commercial and non-
commercial contexts involving large and small groups of
non-technical end-users [1, 5, 10, 22]. For each successful
site, there are many more that fail for lack of interest, partic-
ipation, usability, and sociability. Design principles for cre-
ating such social interaction sites are beginning to emerge
[11, 12, 15, 20] but there is a scarcity of case studies of
designs applying these principles or analyses of the effec-
tiveness of designs. 1
In this paper, we present a case study of the design of two
social interaction Web sites: CHIplace and Portkey. In the
next section, we briefly characterize the target user groups
of these two social interaction sites. We then review several
related efforts to provide principles for the design of social
interaction environments. We describe the specific princi-
ples that guided our designs and how they were instantiated
in a technical solution. We present an analysis of the pat-
terns of use to examine the effectiveness of the Web-site
designs for facilitating social interactions. In conclusion, we
discuss the design implications of our findings.
TWO TARGET SOCIAL GROUPS
Both Web sites described in this paper had the goal of fos-
tering interactions among participants. The first Web site,
CHIplace, was developed for the ACM CHI 2002 confer-
ence to extend the interaction opportunities and the interac-
tions among people both in time and space. The second site,
Portkey, was developed for the summer interns at IBM TJ
Watson Research Center to enable the interns to exchange
helpful information and experiences and to develop the
social networks — personal and professional — that they
needed to function effectively in their new environment.
While both projects shared similar design goals for partici-
pation, social interaction, collective action, and develop-
ment of social networks, they had different requirements
and interactional constraints. For example, the CHIplace
group was expected to be large and diverse whereas the
Portkey group was fixed and characterizable. CHIplace was
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
CSCW’02, November 16–20, 2002, New Orleans, Louisiana, USA.
Copyright 2002 ACM 1-58113-560-2/02/0011…$5.00.
1Socio-technical capital is a resource produced as a side effect of
technology-mediated social interaction. It can be accumulated
and made available for people to create value for themselves or
others [21].
Making Web Sites Be Places for Social Interaction
Andreas Girgensohn Alison Lee
FX Palo Alto Laboratory, Inc. IBM TJ Watson Research Center
3400 Hillview Avenue, Bldg. 4 19 Skyline Drive
Palo Alto, CA 94304, USA Hawthorne, NY 10532, USA
andreasg@fxpal.com alisonl@us.ibm.com
designed to be an open-access system while Portkey was
restricted to registered users; the interns, researchers, and
staff involved in the 2001 summer internship program.
CHIplace Users
Of the 1176 CHIplace users that were registered at the end
of April 2002, 49 were CHI conference committee mem-
bers. Many unregistered users visited CHIplace but we have
no information about them. Among those who provided
country information, 52% were from the US and 48% from
44 non-US countries. The most popular of the 11 roles that
people could select from were researcher (44%), designer
(26%), student (22%), consultant (18%), and interface eval-
uator (17%).
Portkey Users
Roughly half of the Portkey members (738) were summer
interns (343) who were responsible for 95% of the site
usage. The summer interns came from 130 different univer-
sities located in 13 countries. Most students came from uni-
versities in the US (84%), France (4%), and Germany (3%).
Their backgrounds were in computer science (48%), engi-
neering (27%), sciences (11%), business (5%), mathematics
(3%), and other disciplines (6%). Among the students who
indicated their country of origin, 43% were from the US and
57% from non-US countries. 54% of the students were
enrolled in a Ph.D. program, 15% in a Masters program,
29% in a Bachelors program, and 1% in high school. These
demographics illustrate the diversity and international
makeup of the students.
DESIGN PRINCIPLES FOR A SOCIAL WEB SITE
How might one build Web sites that have a focus on people
and social interactions? Kim [11], Kollock [12], and Preece
[20] have independently presented three related sets of
design principles and strategies. Kim’s nine design strate-
gies provide the most systems-oriented approach of the
three [11]. Her strategies cover purpose, people, gathering
places, evolving roles, leadership, rules and policies,
planned events, rituals, and support of sub-groups as well as
a variety of technologies for each strategy. She also presents
three underlying design principles: a) design for growth and
change, b) create and maintain feedback loops, and c)
empower members over time. Taken together, the nine strat-
egies and three design principles provide a technology-facil-
itated social-scaffolding approach to social organization.
Kim’s strategies overlap extensively with Preece’s two
design principles: a) designing for usability and b) support-
ing sociability [20]. The first design principle focuses on
good usability so that people can interact and perform tasks
easily and effectively. The second design principle focuses
on social interaction and three components of sociability: a)
purpose, b) people and roles, and c) policies related to gov-
ernance, membership, codes of conduct, privacy, security,
and copyright protection. Preece also examines software
selection options and technology development approaches.
Her principles draw from an HCI perspective of understand-
ing individuals, groups and their “work” environments in
the design and development of such systems.
Finally, Kollock’s proposal focuses on the key sociological
challenges with building online communities [12]. He
argues that technological and user interface challenges exist,
but the sociological challenges are greater because of the
poor understanding of how they impact a successful online
community. His design principles are derived from work on
cooperation and social dilemmas and focus on fostering
social interaction, cooperation, collective action and social
order [2, 19]. The implications of his proposal do not differ
radically from Kim [11] or Preece [20] but are more
grounded in a sociological perspective. His principles
emphasize the need for a) repeated social interaction, b) a
persistent identity built on information about the person and
their behavior until now, c) well-defined group boundaries,
d) the evolution of norms and rules regarding collective
resources and e) a means to monitor and sanction member’s
behavior.
CHIplace and Portkey: Key Design Challenges
The designs of our two sites were, first and foremost, driven
by the need to identify the key sociological challenges, of
which there were three [12]. “Encouraging user participa-
tion” is invariably an important goal and a key challenge in
any socio-technical system. Without it, the people, their
presence, and their activities which define the social context
would not exist, social interaction would not be possible,
and social organization would not emerge [25]. This chal-
lenge was made even harder because both projects had
scarce human resources for creating and evolving content
and for leading and breathing life into the social system.
“Fostering social interactions” is a key focus of all three
sets of design principles. Kim discusses the importance of
bringing people together and the importance of scheduled
events in reinforcing a sense of belonging. One of Preece’s
two design principles focuses on supporting social interac-
tion. Kollock points to the importance of two individuals
meeting again in the future as a condition for cooperative
relationships to emerge and persist. Repeated social interac-
tions are important for increasing familiarity, developing
relationships, reminding people about what they have in
common, strengthening ties to an organization and helping
to develop a positive regard for the social system [2, 14].
“Promoting visibility of people and their activities” is a key
challenge because, without people, there is no community
[11, 20]. In making them visible, a number of important
issues can be addressed. First, Kollock [12] points out that a
persistent and up-to-date identity is an important condition
for cooperation both in letting individuals identify each
other and for knowing how people behaved in the past. It is
also a crucial element for any community governance sys-
tem to operate in which people need to be able to identify
people, to monitor inappropriate behavior, and to hold
accountable those who misbehave. Second, people provide
help to and interact with people they know, people they like,
people who are similar, and people who have helped them
[4]. Therefore, it can help “people to get to know each
other.” Third, theories of pro-social motivation suggest that
people provide help to strangers even when this help is per-
sonally costly [4]. The reasons are that helping others can
increase self-esteem, personal identification with organiza-
tion, stature within the organization, feelings of commit-
ment, helping to strengthen the social organization, and
helping to promote generalized reciprocity. Therefore,
attributing and recognizing the kindness and efforts of indi-
viduals can help motivate people to help others. Finally, pro-
moting the visibility of people and their activities can
encourage people to become active and to prolong participa-
tion. Related work with positive reputation systems have
shown the value of a good reputation in economic transac-
tions and in preventing people from defecting [13].
Aside from the three key sociological challenges, the design
of our two social interaction Web sites had two additional
design challenges. The first was “designing for usability.”
Clearly, if the system did not have good usability, it would
raise the bar for participation. Also, given the challenges
with motivating people to participate, bad user experiences
could turn away many potential participants.
Our second design challenge was “maintaining the site with
minimal resources.” Both projects had very few human
resources; CHIplace was a community service by the two
principals and Portkey was largely a two-person research
project. This meant that we needed to rely on using technol-
ogy to help with the management of the sites. Wherever
possible, the Web sites needed to grow and evolve through
the contributions of their participants.
DESIGN OF TWO WEB SITES
We provide some examples of and some approaches used to
address the five key design challenges for CHIplace and
Portkey. A seeding, evolutionary growth, and reseeding
approach was applied [7]. Periodic analyses (formal and
informal) were conducted to identify how the sites were
evolving and how socio-technical solutions could facilitate
and nurture their evolutionary growth.
Promoting Visibility of People and Their Activities
Both sites provided a number of mechanisms that enabled
participants to construct and evolve a persistent and verifi-
able identity. First, each registered participant had a profile
containing known and user contributed information that was
presented through individual people pages. Various options
were considered in the design of the profile; real-life details
about the person, a verifiable pseudonym as in eBay; or reg-
istered anonymity as in MOOSE Crossing [3]. The real-life
option was chosen in both cases because people were
expected to meet and interact both online and face-to-face.
Because Portkey was a closed community, all profiles were
created by the administrators. CHIplace profiles required a
valid email address to which the password was sent.
Profiles contained real-life information about each person
including their name, photograph, home country, email
address, and a link to a Web page with more information
(see Figure 1). The CHIplace people profile also contained
people’s affiliation, their HCI role (e.g., practitioner), possi-
ble role and goals for CHI 2002, their HCI interests, and
how they became involved in HCI. Portkey maintained addi-
tional information for students and for permanent research-
ers. This included information about their technical and
outside interests and their research project. Portkey users
were also asked to provide up to five questions that other
people could ask them. These questions were intended to
help Portkey participants initiate conversations when they
met face-to-face. In addition, the student’s profile included
their discipline, university and the researchers who made up
their management chain.
Information about a person’s activities and behaviors on the
site, known as traces, is an important element of a persistent
and up-to-date identity. This includes explicit and implicit
actions that identify both the individual’s interests and other
people’s interests. A clear illustration of traces can be found
in the CHIplace trivia (see Figure 2). Each trivia provides an
author attribution as well as an indication of how many they
have authored. This recognizes a person’s participation and
level of participation. Second, each trivia has an answer
component. The number of times the answer is viewed pro-
vides an implicit indication of interest in the question.
Finally, people can explicitly cast votes about what they
think of the trivia. The number of positive, neutral, and neg-
ative votes reflects interest in the answer. This series of attri-
butions empower authors and readers by recognizing the
author’s efforts and the readers’ interest. Furthermore, such
traces provide insights to potential trivia authors for what
interests the site participants and how to construct a trivia.
Figure 1: People page on CHIplace with member profile.
Figure 2: CHIplace trivia
Finally, both Web sites provided a number of social brows-
ing tools that enabled the participants to get to know the
other participants of the site.
• A list of people that joined most recently together with a
randomly selected picture of one of those people (see the
top of Figure 1). This publicly recognized new members
and provided a technology-facilitated means of introduc-
ing new members to the community [11].
• A gallery of eight randomly selected pictures of members
(see the bottom of Figure 1). For Portkey, four of the pic-
tures were interns and four were permanent researchers.
This used serendipity to acquaint people to other partici-
pants at the Web site.
• People directories sorted by last name or country. On
Portkey, additional directories sorted by such attributes as
first name, discipline, university, and technical and out-
side interests were provided. These directories provided a
means for people to learn about co-participants that
shared similar attributes.
• CHIplace also provided a lightweight people browser that
visually grouped the CHIplace participants with similar
HCI roles (see Figure 3). Each member was represented
as a dot. Rolling over the dot revealed the name of the
person and clicking on the dot led the user to the person’s
profile. Users could also select members by their roles.
Encouraging User Participation
We created awareness about the sites to drive traffic to them.
Invitations to participate were sent to various email distribu-
tions and groups. We also registered the CHIplace site with
various search engines and placed links at ACM Web sites.
Those links were moderately effective as 11% of the outside
CHIplace page accesses were referred by other sites. Search
engines accounted for much of the referred traffic (71%)
and the ACM Web sites for much of the rest (15%).
To sustain awareness, we kept participants apprised of new
additions and ongoing developments on the site. The home
pages of both sites devoted areas for presenting this infor-
mation (see Figure 4). Many of the participants were discre-
tionary users both in terms of goals and motivation for
participation. This turned out to be particularly true for
many of the CHIplace participants and Portkey permanent
researchers. In recognition of this, we sent out periodic
newsletters via email that summarized new additions and
ongoing developments.
Both sites had an incentive-based participation reward pro-
gram. For example, Portkey had monthly contests where an
eligible intern would be randomly drawn as a winner of a
$1000 gift certificate. To qualify for the monthly drawing,
the intern needed to provide a completed profile and to
make a minimum of three postings to the Portkey discussion
forums. Then, additional posts gave the eligible intern addi-
tional ballots in the drawing. This was successful in moti-
vating people to fill out their profiles (174), to provide a
photo (128), and to participate in discussion forums (144
students made at least one posting). However, this was not
the sole reason for participation as only 73 of the 343 stu-
dents ever met the eligibility criteria and no more than 47
people in any given month were eligible for the contest.
On CHIplace, users were not required to be registered to
view the site’s content. We believed that CHIplace partici-
pants would benefit if they registered as it would help to cre-
ate greater awareness of people interested in CHI and their
activities and to build a sense of community on CHIplace.
To encourage registration, parts of the site such as the
archive of old stories and articles and CHI 2002 full papers
were only accessible to registered users.
Recognizing, identifying and attributing participation has
been shown to be valuable for encouraging participation [4,
11, 13]. In particular, traces can be instrumental for user
participation [10, 12]. Aside from the trivia example, we
provided information such as page views, most viewed, and
links on people’s profiles, if applicable, to the number of
and the list of their authored posts. We only presented such
information selectively because limited resources prevented
making this pervasive, but we believe that such information
can both create awareness of interest as well as recognize
participation.
Supporting Sociability: Social Interaction Components
There has been extensive CSCW research on supporting
interaction among geographically distributed co-workers
[6]. We recently distilled the research results into a set of
Figure 3: People Browser.
Figure 4: Part of the Portkey home page providing a
calendar of events, news and discussion.
requirements and four basic social interface components
that can foster social interactions: common ground, aware-
ness, interaction enablers and mechanisms, and place-mak-
ing for building social interaction sites [15]. We briefly
describe each component and provide examples of what
these were in CHIplace and Portkey.
Common Ground
Common ground is the shared understanding among collab-
orators that permits interactions to proceed smoothly and
the shared context for the social interaction. To facilitate the
development of common ground among many of the partici-
pants who were strangers to each other, they were asked to
create profiles. All authored site content was annotated with
links to those profiles. Several topical content pages incor-
porated discussion forums to enable participants to share
and discuss issues with other participants.
In addition, we seeded content in areas that represented a
common need of the participants. For example, on CHI-
place, CHI retrospective consisted of folklore and trivia
about CHI’s history and a list of past CHI conferences, past
conference chairs, and locations. This provided a context
about CHI for newcomers. Another example is the elec-
tronic prepublication of CHI 2002 conference papers for
discussion (see Figure 5). The aim was to leverage the
shared interest that CHIplace members had to help shape the
authors’ presentations and the moderator’s remarks at the
conference. On Portkey, we used the official information
developed by the human resources group about the program
to seed the content of Portkey and supplemented it with
“tips” provided by summer interns from the previous year.
The lack of a common or prior history among Portkey mem-
bers meant that many of the community content opportuni-
ties (e.g., trivia) that CHIplace had did not exist on Portkey.
Instead, we solicited material to post on the site. For exam-
ple, photos from social outings and photos, abstracts and
snippets of student poster presentations organized by the
departments. We featured researchers at the corporation on a
weekly basis on the home page to help make contact with
the broader community (see bottom right of Figure 4).
Awareness
Awareness is the state of knowing or being informed that
enables collaborators to interpret others’ actions and to
guide one’s own actions. It consists of a combination of
people, activity and contextual awareness.
In addition to people profiles, activity traces, and annota-
tions of content with links to people profiles and activity
traces, participants were made aware of new contributions.
We provided listings of what’s changed to support return
visitors. All discussion threads were displayed in chrono-
logical order with the most recent at the top. As well, new
threads and messages since the last time the user visited
were highlighted with a different color. On Portkey, the
home page included a section on today’s active forums and
activities scheduled today and tomorrow (see Figure 4).
Social Interaction Mechanisms
Collaborators need both the enablers and the tools for social
interactions that allow them to act and interact. The enablers
provide cues of the availability and variety of social interac-
tion options. The social interaction tools support the social
and production-oriented aspects of interactions [14], the
synchronous and asynchronous elements of interactions,
and the ad-hoc and planned nature of interactions.
On both sites, we used quick polls to engage the audience
and to determine the course of action for controversial
issues such as the contest on Portkey. These polls provide a
a low-effort interaction mechanism that allows users to pro-
vide their opinions on a topic without having to cross the
hurdle of participating in a discussion forum. In that sense,
quick polls can ease users into taking a more active role.
Both sites had discussion forums for different topics. On
CHIplace, many of the discussion forums were attached to
specific pages. This allows the users to be engaged within
the context of what they are viewing to make suggestions or
to comment on stories. On Portkey, forums were organized
in several categories such as announcements, tips from
peers, and social and technical discussion forums.
CHIplace also featured two other social interaction mecha-
nisms: a) writing and voting of trivia (Figure 2) and b) paper
preview (Figure 5). These were efforts to engage the partici-
pants in discussion, in sharing experiences, and in providing
feedback about the value of people’s contributions.
Place-making
Place-making is the cultural and social understanding of the
norms and practices appropriate to an interaction setting that
emerges over time and through users adapting and appropri-
ating the setting [8, 11, 12, 20].
People profiles are examples of resources that had no
restricted or prescribed use. They were free-formed to pro-
vide flexibility in the kind of profile information people
could provide. There is quite a variety in how people used
Figure 5: Paper Preview.
the fields to express information about themselves. As an
illustration, some Portkey users chose to provide serious,
work-related “Ask Me” questions while others shared their
interests or aspects of their personality.
Much more of a sense of place developed on Portkey than
on CHIplace. One good example is the behavior of people
posting to discussion forums. The contest provided an
incentive for people to post many messages. Some of the
student interns felt that others abused the system and
attempted to demonstrate the absurdity of the situation by
starting a nonsensical discussion thread on the virtues of
Lotus Notes (see Figure 6). This led to a more serious dis-
cussion on proper behavior on the site. Another example is
the use of discussion forums for planning social activities
such as meeting in the park on the weekend. These episodes
illustrate that the discussion forums were more than simply
a vehicle for information sharing. They became a vehicle for
coordinating get-togethers and for teaching, discussing, and
formulating an initial set of norms for using the Web site.
Designing for Usability
In an effort to keep the barriers to participation low, both
sites provided a simple customized form for their users to
create and update their profile, a self-service mechanism to
reset forgotten passwords and email them a new password,
and to upload their photo for their profile. Some pages were
customized to display group-specific information to prevent
irrelevant information from being presented.
Also, we made the navigation structure visible on each page
and positioned it in a consistent location. People and list of
discussion forums were always accessible to the partici-
pants. On Portkey we also used a timeline to structure the
official content about the corporation, the geographic area,
work-related topics, and social activities. This allowed the
students to access the collection of information of concern
to them that was dependent on what phase of the internship
they were at (i.e., before arrival, first week of internship,
during internship, last week of internship). This design grew
out of early fieldwork that revealed that students needed to
access different information depending on what stage they
were at in the internship.
Browsing of the people pages and people directories was
very popular on both sites. People directories facilitated
easy browsing of the people profiles by a variety of
attributes. On CHIplace, we developed the people browser
(see Figure 3) to provide the more discretionary users with a
quick means of finding out who else was like them and who
other CHIplace members were.
Maintaining the Site with Minimal Resources
All pages on both sites were dynamically generated with
JavaServer Pages and a MySQL database. The use of tem-
plates and a database simplified adding new pages or modi-
fying existing pages. This approach also allowed us to
provide custom information for signed-in users.
We used an open-source Web usage analysis tool called
Webalyzer to generate daily and monthly traffic pattern
reports. The reports enabled us to note unusual traffic pat-
terns and to examine root causes for such patterns. For
example, we found heavy traffic at CHIplace on a number of
instances because of referrals by various people.
We developed several scripts that identified problems in
Web usage logs that were encountered by site visitors. For
instance, connection problems to our site, broken links that
we created or that our participants provided and problems
with our dynamically generated pages were brought to our
attention through these daily generated reports. We were
able to exploit Web crawlers that periodically crawled over
the site to identify similar problems.
We created scripts that ran frequently to scan for new photos
provided by participants. These scripts resized the photos to
a recommended size so that our people gallery would not be
populated with highly varying sized photos. Also, the
scripts cached externally-linked photos to enable our sites to
be buffered against vagaries of download and connection
problems. Scripts were also developed to provide scheduled
notification such as reminders to authors and moderators
about the imminent preview of their sessions.
Finally, we did not want to be involved in moderating con-
tributions. Therefore, we decided to only let registered users
contribute and to provide author attributions to all contribu-
tions. This allowed the sites to use a combination of self-
moderation and community moderation schemes. The latter
was particularly instrumental on Portkey when a few indi-
viduals tried to post numerous, low-value posts. It led to a
huge discussion about appropriate netiquette.
Figure 6: Portkey discussion thread with topic list.
USAGE DATA FOR THE WEB SITES
We analyzed the usage data of the sites to assess the effec-
tiveness of their designs. The data shows interesting trends
in response to the introduction of new features and to email
notifications sent to various groups. The data also highlights
similarities and differences between the two sites.
To find usage patterns, we grouped access to pages into ses-
sions. For our sites, sessions provide a fairly accurate esti-
mate for the number of unique users per day because very
few people visit more than once per day judging from the
client IP addresses. We also classified pages into meaning-
ful categories and used those categories to show transitions.
How the Data Was Collected
We collected usage data using a Web server access log that
included cookie information. With cookies we could associ-
ate page accesses with the servlet sessions maintained by
the Web application server. We also kept a sign-in log that
was synchronized with the Web server log via the session
identifier.
We only considered successful page accesses (and not
accesses to images and other auxiliary files). We also fil-
tered out page accesses that did not come from real users.
On Portkey where all users were required to sign in, we
ignored all page accesses that could not be associated with
an entry in the sign-in log and filtered out the entries associ-
ated with site administrators. Search engine accesses were
ignored automatically because there were no sign-ins.
Because sign-in was optional on CHIplace, we used a differ-
ent approach for filtering out pages. First, we ignored access
from the automatic display at the conference. We also
ignored access from a set of IP addresses that were used by
the site administrators. Unfortunately, that also filtered out
some entries by genuine users because one of those
addresses belonged to a “socks” server used by company
employees. Second, we filtered out all requests from user
agents that appeared to be search engine spiders. Finally, we
filtered out requests from browsers that did not return the
session identifier cookie because we could not properly
track those sessions. This last step caused the loss of about
6% of the usage data on CHIplace. However, as an added
benefit, any remaining spider accesses were filtered out
because spiders do not return cookies.
Weekly Usage
We tracked the usage of CHIplace and Portkey for 57 and 14
weeks, respectively. For CHIplace, we had 24,768 sessions
(62 per day) and 193,294 page accesses (excluding images
and other auxiliary files) and for Portkey we had 6,640 ses-
sions (68 per day) and 155,530 page accesses. Figure 7 plots
the weekly number of sessions over time. On Portkey, 95%
of all page accesses were initiated by the students.
When comparing the number of sessions over time, one can
see that the usage is fairly constant on Portkey with the
exception of the last week when many of the interns had
left. Except for the last 11 weeks on CHIplace, Portkey had
more than twice as many sessions. Only the start of the
paper preview on February 18 pushed the weekly number of
sessions above those of Portkey’s.
On average, Portkey sessions contain three times as many
page accesses as CHIplace sessions. This is mostly due to
the heavy use of discussion forums that is discussed later in
this section. Figure 8 shows that the average number of
pages per session stayed fairly constant on Portkey with a
drop in the last two weeks when the summer interns started
leaving and not much discussion took place. In contrast, on
CHIplace there was a noticeable downward trend during the
first four months before leveling off. We believe that this is
due to repeat visitors that found only few new pages and
0
200
400
600
800
000
200
400
600
800
4/2 4/23 5/14 6/4 6/25 7/16 8/6 8/27 9/17 10/8 10/2911/1912/1012/31 1/21 2/11 3/4 3/25 4/15
0
200
400
600
800
1000
1200
1400
1600
1800
5/14 6/4 6/25 7/16 8/6
Figure 7: Weekly number of sessions.
CHIplace Portkey
NNNN
NN
N
NN
C
N
Newsletters are marked with “N”. The CHI conference and a controversial change of the contest on Portkey are marked with “C”.
C
N
N
0
5
10
15
20
25
30
4/2 5/7 6/11 7/16 8/20 9/24 10/29 12/3 1/7 2/11 3/18 4/22
CHIplace
Portkey
Figure 8: Weekly average of page requests per session.
much less activity in the CHIplace discussion forums. Inter-
estingly, the introduction of new features such as the paper
preview only increased the number of sessions per week and
not the number of page requests per session.
Email announcements to participants and other interested
people caused spikes in usage on the day of the message.
For CHIplace, we sent nine newsletters to the members who
requested announcements (see the weeks marked with “N”
in Figure 7). Some spikes are noticeable in the weekly view
but the effect is much more pronounced on the day of the
message. On those days, the number of sessions was on
average 60% higher than on the previous business day and
the number of page requests was 86% higher. Announce-
ments are an effective means to provide awareness of the
site existence and new developments. Messages to mailing
lists such as all SIGCHI members and CHI announcements
proved to be good for drawing in new visitors.
Patterns of Use
To track patterns of use, we grouped pages into major mean-
ingful categories. While different categorizations of pages
are possible, we believe that the chosen categories represent
the different types of use well.
•Home. The entry page into the site.
•People.The main people page with random pictures, the
people directories, and the individual user profiles.
•Discussion. Pages related to discussion forums, threads,
and messages, and pages for posting messages.
•Content. The content specific to the site. On CHIplace,
the content includes CHI 2002 conference information.
On Portkey, the content includes information about the
organization and the intern program.
•Site. Information about the site (e.g., site map, about or
help pages) and tools for manipulating user profiles.
•Community. Information provided by the user community.
This category exists only on CHIplace and includes sto-
ries, articles, organizational overviews, and trivia.
•Preview. Advance information about conference events
such as the paper preview and the demo sign-up.
•Survey. User surveys and polls.
•Contest. Information about contests on the two sites.
We used the categories to illustrate usage of each category,
to summarize overall usage, and to track transitions between
categories (see Figure 9). The graphs are an adaptation of
the graph type described by Olson et al. [18].
Discussion Forum Usage
When comparing the two graphs, we immediately notice the
dominant use of discussion forums on Portkey; 144 people
wrote 1603 messages (16.4 per day). On CHIplace, the dis-
cussion forums were not used much in the 46 weeks before
the paper preview started. More than three quarters of the
messages were posted during the last 11 weeks and almost
two thirds of the discussion activity occurred during that
time. In the first 46 weeks, 43 people wrote 58 messages
(0.2 messages/day) while 89 people wrote 185 messages
(2.4 messages/day) in the remaining 11 weeks.
We believe that the much higher use of the discussion
forums on Portkey is due to two reasons. First, the chances
to win in a contest of Portkey were increased by posting
messages. Even for the people who were not motivated by
the contest, the threads started by posters provided fodder
for discussion. Second, the IBM summer interns were co-
located and faced similar problems. They used the discus-
Site
7.6%
Community
7.9%
Survey
1.1%
Contest
0.3%
Content
9.1%
Home
10.4%
Discussion
10.2%
People
41.4%
Outside
Preview
11.9%
Content
5.9%
Site
4.0%
Contest
2.4%
Survey
0.5%
Discussion
57.3%
Home
7.3%
People
22.6%
Figure 9: Page category usage and transitions.
CHIplace Portkey
The area of the circles is proportional to the percentage
of usage and the thickness of the arrows is proportional
to the percentage of transitions. Transitions below
0.7% are omitted for the sake of clarity.
sion forums to help each other, to share experiences and
interests, to find roommates or to organize activities outside
of work. In contrast, CHIplace participants were highly dis-
cretionary users with only common interests in human-com-
puter interaction and the CHI conference. The introductions
of paper preview and new CHIplace sections (HCI & Cul-
ture and HCI Techniques) with articles from CHIplace
members prompted more discussions over the last 11 weeks.
Similarities in Usage of the Two Sites
While the usage patterns for the two sites look very different
at first glance, they actually have much in common. On both
sites, most transitions stay within the same category or go
from the home page to one of the categories. When remov-
ing the discussion forum usage that dominates Portkey and
the CHIplace community and preview content, the usage
distribution for the two sites is very similar with the people
category dominating the usage. The transitions among the
remaining categories are also very similar. This similarity
might indicate that the users of the two different sites used
interaction elements in similar fashions and frequencies. It
also shows that the users of both sites were very interested
in finding out about other people on the site.
The proportion of usage of the different categories on CHI-
place stayed fairly constant over time. The discussion and
community categories had the most variable usage depend-
ing on the addition of new community content such as sto-
ries and organization overviews and occasional increases in
discussion activities. The conference preview activities
(mostly the paper preview) started in the last 15 weeks and
contributed 21% of the usage during the last 11 weeks.
New CHIplace Members
Another temporal trend concerns the addition of new CHI-
place members (see Figure 10). After a fairly fast increase
over the first two months, the rate of increase slowed down
for the next eight months. With the start of the paper pre-
view, we see a much faster increase, presumably because
the full versions of the previewed papers are only accessible
for CHIplace members and possibly because of the
approach of the CHI 2002 conference.
LESSONS LEARNED
The paper examined the design of the Web sites in terms of
the social functions of the infrastructure components. How-
ever, the underlying computational and content management
components were very important to reuse and to a low
maintenance and development effort. The use of dynami-
cally generated pages from templates and a database simpli-
fied adding and modifying pages and creating custom
information for signed-in users.
A supportive infrastructure alone is insufficient to attract
and retain user interests. New and frequent content addi-
tions, be it editorial or user contributions, must occur or else
user interest disappears quickly. Two other Web sites using
our infrastructure, CHIplace 2003 and CSCWplace, provide
little new editorial content and have little activity in the dis-
cussion forums. Perhaps, as a consequence, new users are
registering at a slow rate.
In the absence of a large editorial staff, user and community
contributions such as organizational overviews or articles on
HCI topics provided an alternative source for editorial con-
tent on CHIplace. Portkey was more successful in fostering
discussions because, in part, the Portkey members had more
interests and problems in common. As well, the contest on
Portkey helped to motivate a few to initiate some of the dis-
cussions that provided a critical mass of messages for others
to respond to. Using moderators in the CHIplace paper pre-
view turned out to be less effective. Lightweight social
interaction tools such as quick polls provide lower barriers
to participation and produces valuable user contributions.
Our usage analysis shows that information about other peo-
ple, discussions, and community content is of great interest
to the users. People pages were especially popular on both
sites. Some users kept reloading the page to obtain new ran-
dom picture galleries. Making people and their contribu-
tions visible is invaluable in capturing visitor interest.
Other than new content and discussion activity, email
announcements were good means for drawing in new users
and for renewing interest of existing users. Our data shows
that such notifications caused spikes in the usage. Interest-
ing members-only areas were also useful in encouraging
visitors to register as members (see for example the increase
in Figure 10 with the start of the paper preview in February).
CONCLUSION
The opportunities created by the Web open the doors to a
broad range of interactions, some involving repeated inter-
actions and some bringing strangers together. They can be
brought to bear on many application domains where there is
a need for human-to-human interaction mediated by socio-
technical elements, such as dynamic e-Business, customer
relationship management, support networks, and electronic
marketplaces [1]. Social interaction environments play an
important role in the interaction and task experiences in
these opportunities by coupling critical social relations with
communication and information technologies [21].
In this paper, we presented a case study of the design of two
social interaction Web sites: CHIplace and Portkey. We
described five key design challenges that influenced the
development of the Web sites, drawn from several related
efforts to provide principles for the design of social interac-
tion environments. We presented examples of how these
challenges were addressed with socio-technical solutions.
We analyzed usage patterns to examine the effectiveness of
the designs for facilitating social interactions. Our results
demonstrate that social interaction Web sites are effective in
sustaining interest and in fostering social interactions.
0
200
400
600
800
1000
1200
4/2 4/30 5/28 6/25 7/23 8/20 9/17 10/1511/12 12/10 1/7 2/4 3/4 4/1 4/29
Figure 10: Increase of CHIplace Members.
REFERENCES
1. A. Armstrong and J. Hagel III, The Real Value of On-
line Communities. Harvard Business Review, May-
June, pp. 134-141, 1996.
2. R. Axelrod. The Evolution of Cooperation. New York:
Basic Books, 1984.
3. A. Bruckman. MOOSE Crossing: Construction, Com-
munity, and Learning in a Networked Virtual World for
Kids, MIT Ph.D. Thesis, 1997.
4. D. Constant, L. Sproull, and S. Kiesler. The Kindness
of Strangers: On the Usefulness of Electronic Weak
Ties for Technical Advice. In S. Kiesler (ed.), Culture
of the Internet, Lawrence Erlbaum Associates, New
Jersey, pp. 303-322, 1997.
5. S. Farnham, L. Cheng, L. Stone, M. Zner-Godsey, C.
Hibbeln, K. Syrjala, A.M. Clark, J. Abrams. Hutch-
World: Clinical Study of Computer-Mediated Social
Support for Cancer Patients and their Caregivers. In
Human Factors in Computing Systems, CHI 2002 Con-
ference Proceedings, ACM, New York, pp. 375-382,
2000.
6. K.E. Finn, A.J. Sellen, and S.B. Wilbur (eds.), Video-
Mediated Communication, Lawrence Erlbaum Associ-
ates, New Jersey, 1993.
7. G. Fischer, J. Grudin, R. McCall, J. Ostwald, D. Red-
miles, B. Reeves, F. Shipman. Seeding, Evolutionary
Growth and Reseeding: The Incremental Development
of Collaborative Design Environments. In G. Olson, T.
Malone, and J. Smith (eds.), Coordination Theory and
Collaboration Technology, Lawrence Erlbaum Associ-
ates, New Jersey, pp. 447-472, 2001.
8. S. Harrison and P. Dourish. Re-Place-ing Space: The
Roles of Place and Space in Collaborative Systems. In
Proceedings of the Conference on Computer-Supported
Cooperative Work (CSCW’96), ACM, New York, pp.
67-76, 1996.
9. T. Ishida and K. Isbister (Eds.). Digital Cities: Technol-
ogies, Experiences, and Future Perspectives. Lecture
Notes in Computer Science. Springer, Berlin, 2000.
10. S.U. Kelly, C. Sung, and S. Farnham. Designing for
Improved Social Responsibility, User Participation, and
Content in On-Line Communities. In Human Factors in
Computing Systems, CHI 2002 Conference Proceed-
ings, ACM, New York, pp. 391 - 398, 2000.
11. A.J. Kim. Community Building on the Web. Peachpit
Press, Berkeley, CA, 2000.
12. P. Kollock. Design Principles for Online Communities.
Harvard Conference on the Internet and Society, 1996.
Also available at: http://www.sscnet.ucla.edu/soc/
faculty/kollock/papers/design.htm.
13. P. Kollock. The Production of Trust in Online Markets.
In E. J. Lawler, M. Macy, S. Thyne, and H. A. Walker
(eds.), Advances in Group Processes, 16, JAI Press,
Greenwich, CT, 1999.
14. R.E. Kraut, R.S. Fish, R.W. Root, and B.L. Chalfonte.
Informal Communication in Organizations: Form,
Function, and Technology. In S. Oskamp and S. Spaca-
pan (eds.), People’s Reactions to Technology: In Facto-
ries, Offices, and Aerospace, Sage Publication, pp. 145-
199, 1990.
15. A. Lee, C. Danis, T. Miller, and Y. Jung. Fostering
Social Interaction in Online Spaces. In Proceedings of
INTERACT 2001: IFIP TC.13 International Confer-
ence on Human-Computer Interaction, IOS Press, pp.
59-66, 2001.
16. K.D. Mickelson. Seeking Social Support: Parents in
Electronic Support Groups. In S. Kiesler (ed.), Culture
of the Internet. Lawrence Erlbaum Associates, New
Jersey, pp. 157-178, 1997.
17. E.D. Mynatt, A. Adler, M. Ito, C., Linde and V.L.
O'Day. The Network Communities of SeniorNet. In Pro-
ceedings of ECSCW'99, Kluwer Academic Publishers,
London, 1999.
18. J.S. Olson, G.M. Olson, M. Storrøsten, and M. Carter.
Groupwork up Close: A Comparison of the Group
Design Process with and without a Simple Group Edi-
tor. ACM Transactions on Information Systems, 11(4),
321-348, 1993.
19. E. Ostrom. Governing the Commons: The Evolution of
Institutions for Collective Action. Cambridge Univer-
sity Press, New York, 1990.
20. J. Preece. Online Communities: Designing Usability,
Supporting Sociability. John Wiley & Sons, Ltd.,
England, 2000.
21. P. Resnick. Beyond Bowling Together: SocioTechnical
Capital. In J.M. Carroll (ed.), Human-Computer Inter-
action in the New Millennium, Addison-Wesley, New
York, pp. 647-672, 2001.
22. H. Rheingold. The Virtual Community: Homesteading
on the Electronic Frontier. pp. 1-16, 1993.
23. T.L. Roberts. Are Newsgroups Virtual Communities? In
Human Factors in Computing Systems, CHI’98 Con-
ference Proceedings, ACM, New York, pp. 360-367,
1998.
24. D. Schuler. HCI Meets the “Real World”: Designing
Technologies for Civic Use. In J.M. Carroll (ed.),
Human-Computer Interaction in the New Millennium,
Addison-Wesley, New York, pp. 627-646, 2001.
25. J. Short, E. Williams, B. Christie. The Social Psychol-
ogy of Telecommunications. John Wiley & Sons: New
York, 1976.
26. L. Terveen and W. Hill. Beyond Recommender Sys-
tems: Helping People Help Each Other. In J.M. Carroll
(ed.), Human-Computer Interaction in the New Millen-
nium, Addison-Wesley, New York, pp. 487-509, 2000.
27. J.C. Thomas, W. Kellogg, T. Erickson. The Knowledge
Management Puzzle: Human and Social Factors in
Knowledge Management. IBM Systems Journal, 40(4),
2001.
