ABSTRACT
Abstract: Unmoderated
computer-mediated discussion groups suffer from the paradox that the more
successful they are, the more difficult they are to use and administer, because
of the increasing number of users and discussions. Traditional solutions to this problem, such
as creating static subgroups or requiring Web access, can change the nature of
the community in undesired ways. We
discuss the shortcomings
of
traditional solutions and present an approach to scaling online communities that we believe
increases, rather than decreases, communication opportunities without
overwhelming users or administrators.Systers is a 14-year
old online community for technical women to support each other and share
information. Explosive growth in the membership (from an initial group of 12 to
over 2300 today) and of the Internet
generally, has
changed the nature of this community. Systers has
always functioned as a self-moderated mailing list, to which any member may
post at any time, but this model becomes less tenable with greater size. We
report empirical research that we used to redesign Systers in a manner that
retains its best features –
intimacy, comprehensiveness, and diversity –
without the exponential increase in traffic that normally accompanies list
growth, allowing Systers to serve an increasingly diverse membership, balancing
the needs of newcomers, who desire more and broader discussion and services,
and long-time members, who tend to prefer limited interaction. We are beginning
the first field test of our new system and will have empirical results by the
camera-ready deadline.
Keywords
Electronic communities, diversity, networks
(social), networks (digital), web technology, computer-mediated communication,
electronic mail, groupware
INTRODUCTIONIntroduction
Systers is an unmoderated email list for women in
computer science, originally created in 1987 for 12 women (Borg 1993). The list has been
tremendously successful at creating a community for people who had felt
isolated, growing to include over 2300 members in 38 countries. As Systers has grown,
however, it has
lost hundreds of members, especially senior women, because of the increased message volume that has come with increased membership. Additionally, members have felt less free to post
to the entire list, instead responding directly to the sender of a message
or not at all, reducing the sense of community and the utility of the
list. Our goal is to
provide a better format for such a large community than an unmoderated email
list, while keeping the features that make Systers successful. Despite
programs and legislation that have opened professional opportunities to women,
women accounted for just 15% of the Ph.D. degrees in computer science in 1998,
and only 8% of full professors in computer science were women [1]. These
statistics reflect trends in the hard sciences and engineering more generally;
women have comparatively low participation rates and professional mobility.
The electronic community of Systers was begun to
encourage technical women’s persistence and advancement by helping technical
women help one another. Systers was started in 1987 as a small electronic
mailing list for 12 women who met at an academic conference on operating
systems, thus the name “Systers.” Today Systers is maintained by the Institute
for Women and Technology [2],
headed by Anita Borg, the founder of Systers. Online interactions
with other technical women through Systers has been of great benefit to
thousands of female computer scientists, addressing problems related to
isolation and adverse attitudes, increasing members’ access to job information
and contacts, and even mobilizing change.
The current challenge is to prevent Systers from collapsing under the
weight of its own success.
The current technical infrastructure of Systers
began as a research project in the use of databases for collaboration. With the
growth of the Internet and the Web, this infrastructure supported a substantial
growth of Systers to its current membership of over 2300 members from 38
countries, including Niger, Bolivia, and Russia. With growth came diversity of
interests and varying access to computer resources and colleagues. Growth also altered the sense of close
community that existed when members were few in number, met face-to-face at
meetings, and knew one another well. Norms of communication changed. Almost all
Systers members read all posted messages in email as they come to their
mailboxes. Today these messages arrive in a context of heavy Internet traffic.
Systers members we sampled receive on average 268 email messages each week and
send nearly 80. As a result of increasing communications and members, Systers
has had an informal policy of limiting discussion messages and topics. Limiting
communication, however, also constrains what a virtual community can do, since
all transactions are based on communication.
A redesign project is currently underway to improve
the experience of community within Systers. As part of the redesign, we have
been conducting research with Systers members to learn better who the members
are, what leads them to Systers, and what they need and want. We began with a
random sample cross-sectional survey of Systers members. We next instituted a
survey of new members as they subscribed to Systers. We plan to follow up with
these newcomers in a year. We also plan an exit survey of those who leave
Systers. We have completed a preliminary design and implementation of a new
structure for Systers. We report here
what we have learned about the Systers community and its members thus far and
describe the redesign. By the time of
the camera-ready deadline, we will describe the results of Systers’ adoption of
the new technology.
THE SYSTERS COMMUNITY
Although the business community today is interested
in creating “electronic communities” for the purpose of increasing revenue [3],
we believe that a successful electronic community should promote the welfare of
members [e.g., 4, 5].
As the introduction to Systers states, the intent of Systers is to be “a safe
and comfortable place [for its members] to discuss important issues.”
All electronic communities create and maintain
territorial boundaries to signal who is and is not welcome. Systers maintains
strong community boundaries through its membership rules. As it describes
itself:
Systers is a private organization of professional
women in technical areas of the field of computing – i.e., with a degree (or
currently a student) in Computer Science or Computer Engineering, or holding an
equivalent technical position in industry, academia, or government. This group
is not intended for those in marketing, writing, or administrative positions
within a computer company, or studying or reporting about women in the computer
industry, or just interested in the problems of female computer professionals.
It is for technical women. It is not intended for men, however sympathetic, as experience
has shown this tends to dilute the purpose of a special forum just-for-us,
where we don't have to spend energy explaining issues to a different point of
view.
(For the rationale behind limiting membership by
gender, see [6].) A feature distinguishing electronic communities
from narrower electronic groups is that the former
have more varied and differentiated functions and activities. Systers facilitates both unstructured and
structured communication (such as a weekly job postings compilation) both
online and off. Occasionally, members in a particular geographic location
(e.g., Bay Area Systers) or members attending a particular conference arrange
face-to-face get-togethers. Three, now separate, electronic groups first “met”
and coalesced on Systers: Systers-Students, Systers-Academia, and
Systers-Out. Much of the work
organizing the first Grace Murray Hopper conference was done via Systers.
Community norms and “rules of the road” create and
reinforce behavioral boundaries, signaling the kinds of behaviors that the
community encourages or discourages. Because the behavioral repertoire in
electronic communities is limited mostly to written communication, most of the
norms and rules are also related to written communication: What are appropriate
(and inappropriate) topics? Styles of discourse? Styles of interaction? Systers
is explicit about its rules for internal communication, as evidenced by
guidelines posted in its introduction:
DO NOT
FORWARD Systers list mailings without first obtaining permission from the
originator of a message.
DO NOT
USE OR DISSEMINATE information in Systers messages without permission of the
original sender.
Personal defamation or character assassination of individuals or by name
is an inappropriate use of the list. Information that is possibly libelous or
slanderous may not be posted.
With
the exception of the posting of resumes and certain job opportunities, Systers
may not be used for any advertising or commercial purposes.
Frequently Systers ask for opinions over the list, collect them and then
collect or summarize and redistribute them to the list. Collections should not be sent beyond the
list unless the collector states that the message can be disseminated. The
collector should get permission from every contributor before indicating that
the message can be passed along.
(http://www.Systers.org/mecca/cgi-bin/new- req.tcl.
Visited 3/3/00.)
As Systers and the Internet grew, the community
adopted two important normative mechanisms to organize and control
communication. Both norms are conveyed
primarily by example. The first norm is to answer requests for advice by
replying to the sender alone, who then collects and sends a summary of replies
to the community as a whole. This practice is also common in technical and
scientific distribution list groups, in which members post technical questions
or questions about the literature of the field. They then create summaries of
the replies they receive and post them to the entire list. This practice for
disseminating information contrasts with the social discussion and
give-and-take one finds in threaded forums, such as Usenet. Nearly 50 years
ago, experimental laboratory comparisons of analogous structures were a hot
scientific topic. So-called “wheel” or “star” networks often proved to be
efficient, but member satisfaction was highest in “all-channel” networks [e.g.,
7]. Thus, the norm for limiting communication is
not without cost.
A second important norm is that Systers topics
should be limited to issues related to being a technical woman. The following
gives a flavor of appropriate topics:
How
to negotiate a good maternity leave arrangement, if your company doesn't have
an official policy.
Barbie
software, and what it implies for getting girls interested in computing.
I'm
thinking of going back to school. These are my goals. What kind of program will
help me meet them?
Here
is a situation I am having with a male colleague at work; how do I deal with
it?
Let’s
create women-oriented versions of “real engineer” jokes. An example: Real engineers figure out algorithms to
minimize thread usage when doing counted cross stitch.
Keeping
up with technology vs. time off for family.
Off topic messages include such items as, “How do I
do X in perl?” (technical questions) and “Should I send my kid to private
school?” (general child care). When off-topic messages are posted, a member or,
occasionally, the founder, will post a response reminding the poster as well as
the entire community of topic norms.
Nevertheless, it remains an important problem, both in wasted bandwidth
and in frustrated members who disagree on what topics are appropriate.
Systers’ discourse norms substantially reduce daily
message traffic in comparison with that of groups whose norms are to tolerate
more digression and/or replies to the group as a whole. Systers’ daily message traffic ranges from 0 to 20, a
comparatively small number compared to that of many electronic groups and
communities. Indeed, many non-members
have trouble believing that such a large unmoderated mailing list has such a
small volume. Systers members in our
survey reported posting significantly more messages to other favorite electronic
groups, even to technical distribution lists.
The externally visible structures and processes of Systers described
above comprise the context for community interaction. The life of the community is to be found in its interacting
members, at least some of whom feel emotional affiliation, a sense of identity,
and a sense of obligation to the community. Our survey was meant in part to
capture these feelings and perceptions.
SURVEY OF MEMBERS
We obtained systematic information about Systers
members through a random sample survey. In December 1998, after obtaining
permission from the Systers membership, we randomly selected 1000 subscribers
from the rolls of Systers members. We obtained usable email addresses and
successfully sent email to 819 of these members, asking them to participate in
an online survey contained in the message. The survey consisted of 80 items,
nearly all presented in checkbox format. The survey requested information on
the member’s demographic attributes, technical environment, online groups,
social ties, work situation, and preferences on the redesign of Systers.
We received 367 completed surveys, a 45% response
rate. The respondents include women who were members at Systers’ inception as
well as women who joined through 1998; median tenure in Systers in the sample
was 3 years (mean = 3.8). We next report the demographic attributes of
respondents.
Mainstream with Diversity
Figures 1 and 2 show how Systers in the sample were
distributed by age and education; the median age was 38 years. Most Systers
were employed full time, married, white, U.S. professionals. As would be
expected, older, married, better educated,
fully employed Systers tend to have higher household incomes (rs = .16 to .28).
A large minority (42%) had household incomes above $100,000. However, some
lived in households earning less than $25,000 per year (5%), were students
(15%), single (25%), nonwhite (13%), or non-U.S. based (13%). These demographic
patterns were useful in predicting certain work situations and preferences of
Systers members, as we show below.
Social Ties
We measured Systers’ social ties using 5 survey
items: size of one’s local social circle, number of friends one talks with
about work, number of email messages received and sent per day, and number of
online groups. These factors were highly correlated (rs = .36 to .45). They
were also independent of demographic factors and work situation factors. Those
who are sociable and have many friends and colleagues would be likely to enjoy
communication, have more communication partners, and be exposed to more
opportunities for communication whether in the real world or online. There is
some evidence that those with fewer social ties in the real world seek them
online [9]. If so, more real world social ties might be
unrelated or negatively correlated with online communication, and posting
should be predicted by an absence of real world ties. This may be true in other
cases, but we did not see it in our data. Having more social ties and talking
with more friends about work was correlated modestly but positively with
posting in Systers and in other online groups (rs = .12 to .15). As we see
below, posting in Systers is indeed motivated by problems, but they are work situation
problems rather than an absence of people to talk with that seem to spur more
communication in Systers.
Work Situation
We measured four attributes of the work situation
of Systers members: size of the work group, percentage of female coworkers,
rated work satisfaction, and rated equity of the work environment using 4 items
adapted from the Equitable Work Environment scale [8].
Work environment equity was strongly correlated with more work satisfaction (r
= .40). Though having more female coworkers was associated with higher ratings
of perceived equity (r = .18), a higher percentage of female coworkers also was
characteristic of older Systers and those having less education and lower
income (see Figure 2). The statistical regression model using all demographic
variables to predict the percentage of women, and controlling for size of
workgroup, is a significant model: R2 =
.15, df [240], p < .001; where [less] education (F[1,240] = 10.7, p <
.01), [greater] age (F[1,240] = 20.1, p < .001); [less] income
(F[1,240]=6.9, p < .01) predict a higher percentage of women in the
workplace. In sum, our measures of the work situation tap two contradictory
facets of work situation quality. A higher percentage of women coworkers is
associated with higher perceived equity, but also with lower job salary.
Participation in Systers
Systers exists as an
active community insofar as members communicate. Among respondents, just 38%
had posted at least one message to the community in the past year; the mean was
1 message, with a range from 1 to 20. We tested a hypothesis that active
participation would be motivated by work-related needs, that is, a more
negative work situation or junior status in the profession. We used a series of
regression equations that also included demographic and social variables as
controls, successively deleting insignificant variables. A significant
predictive regression model of posting frequency in Systers (R2 = .04, F[4,293] = 3,
p = .05) included (lower) age (n.s.), (higher) number of people one talks with
about work (F[1,293] = 2.7 p < .10),
(higher) percent female coworkers (F[1.293] = 5.8, p < .05), and
(more) work environment inequity (F[1,293] = 4, p < .05). See Figure 3 for a
summary of the results across models we explored. Our results are consistent
with the hypothesis that more active participation in the online community is
motivated by work-related needs.
Attitudes about Systers
We asked respondents whether they felt Systers was
interesting, helpful, supportive, and a community. Respondents answered on
3-point scales (e.g., not interesting, somewhat interesting, very interesting).
Answers were heavily skewed towards the positive side of these scales, leaving
little variance. Those in a less equitable work environment tended to say that
Systers is more helpful (F[1, 327] = 5.3, p < .05); controlling for
education, those more junior in their career tended to feel Systers is more of
a community (F[2,311] = 3.1, p < .05).
Systers members with children felt less positively
about Systers on all these measures (ps < .10 to .001), possibly because
broad discussions of child raising and child care
not closely related to being a technical woman have been discouraged in the
group.
Figure 2. Percentage
of female coworkers as a function of age and household income. (Note: This
figure shows results for full-time employed respondents only.)
Preferences for
Systers Online Resources
We assessed
preferences and goals for Systers through ratings of 26 preferred activities in
using Systers. We examined each item’s distribution, setting aside those that were
skewed. We evaluated the remainder using principal components analysis and
varimax factor analysis constrained to three factors, accounting for more than 75%
of the variance. Factor 1 consists of the items related to interaction, such as
“find women to communicate with from all over the world.” Factor 2 consists of
broadcast information, such as “keep up with or make job/position
announcements.” We could not interpret the third factor, and did not use it. We
grouped remaining items logically: see Figures 4a-4f for details. Correlations
within the groupings ranged from .3 to .75.
The same work
situation variables that predicted posting in Systers also predicted
preferences. Those having more female coworkers and those in a more inequitable
work environment preferred more functionality for Systers and had stronger
preferences than other Systers for virtually all of the activities listed.
These trends were especially strong (statistically) when predicting mentoring
activities (R2=
.12, F[4,313] = 10.6, p < .001). This model, predicting who preferred to
give or receive mentoring within Systers, included a (higher) number of people
you talk with about work, (younger) age, (more) work environment inequity, and
(higher) percent female coworkers. Thus, age was negatively related to a desire
for mentoring activities (F[1,313] = 32, p < .001); work environment
inequity and more female coworkers were positively related to a desire for
mentoring activities (Fs[1,313] = 11.1 and 8.6 respectively, ps < .01).
Figure 3. Model of
Posting on Systers
Long-time Systers
A quarter of the
respondents of Systers had subscribed for more than 5 years. Since sustaining a
community requires not only active participation but also some stability in the
membership, we compared the behavior and preferences of long-timers with those
of other members. Long-time Systers are older, better
off financially and better educated than other Systers. Long-timers work in
larger groups but with fewer female coworkers. They rated their work
environment as more equitable than other Systers did. They read Systers as frequently and posted about average for the
community as a whole. However, their preferences and goals for Systers were
significantly more efficiency-oriented than those of the membership as a whole.
These data are exemplified in Figure 4, where we show the role of Systers tenure on preferences about the
redesign of Systers. All Systers rate communication and interaction highly, as
well as the exchange of news and information about jobs and technical meetings.
However, across all of these categories, long-timers were less likely to rate
these functions highly.
SURVEY OF NEWCOMERS
Recently, Systers has
been inundated with new members. Since June 1999, over 900 women have joined.
It was our hunch that these newcomers would hold stronger expectations of the
group to provide support. However, we did not know whether newcomers would be
looking mainly for broadcasts of job postings and information or for more
discussion and interactive communication. To investigate these alternatives, we
sent email surveys to 951 new members soon after they joined Systers. We
received 420 valid responses; a response rate of 44%. The survey was similar to
that of the original, but we added a few
measures and modified question wording to improve our response accuracy.
We added questions on health and stress
levels as well as more measures to assess how respondents’ perceived their
social ties and interactions at home and work.
How Newcomers Compare
In comparison to the
cross-sectional sample, newcomers to the Systers community are younger, less
well-educated, and more junior in their careers. Like their counterpart Systers,
they are predominantly full-time, white, U.S. professionals. However, larger
minorities of newcomers are unmarried (37%), are non-white (18%), and are
non-US based (20%). The median age of newcomers in the sample is 32 years old,
but the distribution in the right panel of Figure 1a shows us that a large
minority of newcomers (41%) are in their twenties. More than 50% of newcomers
have college degrees and 25% hold masters degrees, but very few have Ph.D.
degrees (see Figure 1b).
To measure women’s income
outcomes separately from that of their spouses, we added a new measure for
personal income to the measure of household income. Personal income correlated
highly with household income, however (r = .78) and, like household income, was
positively correlated with being older, married, better educated, and fully
employed (rs = .15 to .26).
Newcomers earn less
on average than the cross-sectional sample. Thirty-three percent of newcomers
reported household income over $100,00 as compared with 42% of the cross-sectional sample. Newcomers with household
incomes less than $25,000 were also more numerous (11%) than in the
cross-sectional survey (5%). The personal income measure provides us with a
better estimate of women’s levels of success in the job; 9% had personal
incomes over $100,000; 22% had personal incomes less than $25,000, which is
understandable given that 26% were full or part-time students.
Work Situation
Although the
newcomers differed demographically from our cross-sectional sample, we obtained
similar statistical results. The correlation between work environment inequity
and work satisfaction was still strong (r = .40). The newcomer data also showed
that higher numbers of female coworkers is associated with higher ratings of
equity (r = .22) despite the fact that a higher percentage of female coworkers
is indicative of lower incomes despite older age. A statistical model
controlled for size of work group and used all demographic variables to predict
percentage of women coworkers at a significant level. (The R2 =
.09, df = 295, p < .001 where [greater] age (F[1,295] = 6.9, p < .01) and
[less] income (F[1,295] = 6.9, p < .01) predict a higher percentage of women
in the workplace.
Newcomers’
Preferences
The newcomers were
asked to rate the same 26 preferred activities for using Systers. We grouped
items as we did for the cross-sectional sample. We used a multivariate analysis
of variance to examine comparative preferences of
Figure 4. Systers’
Preferences (see Appendix A).
newcomers,
long-timers, and other Systers. The mixed model test compared all preference
types at once (within-subjects), and controlled for age. This analysis showed a
significant effect of tenure in Systers across all preferences even controlling
for age (F[2,728]= 3.03, p < .05). Newcomers were more enthusiastic
generally. There was also a tenure X preference interaction indicating that
newcomers are especially looking for mentoring [t(1,649) = 4.4, p < .001],
multimedia functionality [t(1,649) = 3.2, p < .01], and discussing nonwork
topics [t(1,649) = 2.7, p < .01]. Newcomers who expressed interest in being
mentored also said they would want to mentor a less-experienced woman (r =
.38).
Technical Environment
Another area where we
find interesting contrasts between the original cross-sectional sample and the
newcomers is in their technical environment. These differences seem to
reflect both trends in the industry and seniority differences between the two
groups. The Systers in the original sample are primarily Windows users (56%)
with a significant minority (36%) using Unix. The newcomers are overwhelmingly
Windows users (84%), with only 8% using Unix. Nine percent of the original
sample and 5% of the newcomers use Macintosh computers.
Systers in the
original sample are more likely to have fast Internet connections, with 74% of
them connecting via a direct gateway vs. 52% of the newcomers. In fact, 39% of
the newcomers make their only Internet connection via telephone modem, whereas
only 15% of the original sample connects that way. However, the newcomers have
somewhat better modem connections, with two-thirds of them having 56K modems,
whereas the Systers in the original sample who do use modems are about equally
split between 56K and 28k modems.
Systers members use
many different email clients. Only four different clients, including Pine, a
terminal-based mail client, were mentioned by more than 15% of the original
sample. Almost half the newcomers use Microsoft Outlook, with Netscape
Messenger the only other mail client mentioned by more than 15% of the
Newcomers. Ninety percent of the original sample reported using the Netscape
browser. Only half of the Newcomers report browsing the web with Netscape and a
large group (40%) use Explorer.
The newcomers’
technical environment reflects what we know of the recent technical environment
of the general online population in the U.S. (e.g., overwhelming use of
Windows). The original sample is much more diverse in the technology used
by members, possibly due to their long-time use of computers and to the
original roots of Systers as a collection of women academics. That technical
diversity is a challenge for any infrastructure design. We have to assume a
significant (in fact, increasing) fraction are on low bandwidth connections,
and we can only make lowest common denominator assumptions about available
applications, web browsers and mail clients.
DESIGN FOR THE FUTURE
The core goals of the
Systers redesign are to:
Address
the growth of Systers. The current infrastructure was designed to cope with
500-1000 members; designs that may work well for 2500 Systers may similarly
fail when Systers has 5000 or 10,000 members.
Create
a welcoming place for both junior and senior women. Their needs and desires are
quite different, yet the community's core values are only served when both
groups are present and are contributing actively
Not exclude any Systers
due to assumptions about the availability of particular hardware or software,
bandwidth levels, and ease of Web access (a particular issue for women in third
world countries).
These goals, combined
with the data from our survey research and with formal and informal interviews
of Systers members, lead to the following design requirements:
Provide
Systers with ways to customize the information they receive, so that they
receive a higher proportion of information that is of interest and value to
them
Provide
both email and Web access to Systers.
Ensure
that the core communications of Systers are accessible to members with low
bandwidth and limited Web access. We assume that everyone has access to email
(possibly text only) and to a Web browser, although the Web access might be
limited.
Be very easy to use.
Provide
new functionality that enhances the sense of community.
Email and Web Access
An interesting
dichotomy among Systers that seems to be correlated with age is whether a
Syster prefers to get postings via email (a push technology) or via the Web (a
pull technology). Younger Systers prefer the Web, whereas older Systers
overwhelmingly prefer email. (A common quote is "if
it didn't arrive in my mailbox, I know I would never make the time to check
Systers out"). This difference may
very well change, if more people find the Web to be the center of their on-line
life. However, in this period of possible transition, we risk losing large
numbers of our senior Systers if we do not provide email access. Furthermore,
email is a more universally available technology than Web access for people in
countries where connect time is charged by the minute. Email also is more
amenable to solutions that span a wide range of bandwidth levels. On the other
hand, some activities are either impossible or extremely difficult to do over
email, yet are quite easy to provide on the Web. We can greatly improve the
services Systers provides if we allow for Web access.
Ease of Use
Since all Systers'
members consider themselves computer professionals, one might hypothesize that
ease of use would be less important than in other populations. On the contrary,
since Systers is an optional activity, users of the system have very little
tolerance for capabilities that are not "obvious.” Furthermore,
since the technical backgrounds of the Systers community are extremely diverse,
ranging from COBOL programmers on mainframes to students at universities using
futuristic technologies in their daily lives, there is almost no common set of
conventions that we can expect everyone to know. Experience with the current
Systers infrastructure has shown that any activity that requires learning or
active intervention to get the desired result (e.g., the default setting is not
the desired setting) is an impediment to use.
Volume and
Customization
Rejected Alternatives
A common general solution to heavy volume on
mailing lists is moderation, which can be performed top-down by a
small set of administrators or bottom-up through collaborative filtering. While the most obvious cost of moderation is the human overhead
required, a more fundamental problem is that moderation schemes limit diversity through the “tyranny
of the majority”. A topic of little interest to the majority may be
of great interest and value to a minority. A better scheme (which we describe
below) would
allow users to customize which messages they see.
A conventional alternative to moderation would be creating static subgroups of Systers interested in particular topics. There might be technical groups focused on Web design, groups discussing how best to deal with maternity leave, or support groups for women in the final stages of their dissertations. However, while an individual Syster may not be interested enough in maternity leave to join a group that discusses it exclusively, part of what members find valuable is hearing about other women's experiences in