1999-UIST-Alternative interfaces for chat

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, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
UIST ’99. Asheville, NC
© 1999 ACM 1-58113-075-9/99/11... $5.00
CHI Letters vol 1, 1 19
Alternative Interfaces for Chat
David Vronay, Marc Smith, and Steven Drucker
Virtual Worlds Group, Microsoft Research
One Microsoft Way, Redmond, WA, 98052
{davevr, masmith, sdrucker} @microsoft.com
ABSTRACT
We describe some common problems experienced by
users of computer-based text chat, and show how many of
these problems relate to the loss of timing-specific
information. We suggest that thinking of chat as a real-
time streaming media data type, with status and channel
indicators, might solve some of these problems. We then
present a number of alternative chat interfaces along with
results from user studies comparing and contrasting them
both with each other and with the standard chat interface.
These studies show some potential, but indicate that more
work needs to be done.
KEYWORDS
Chat, computer-mediated communication, streaming
media, visualization, time-based media
INTRODUCTION
Since the earliest days of computing, people have used
computer systems to enable communication. Simple
command-line programs such as write (and later
finger) have existed on the earliest time-sharing
systems, while computer bulletin board systems (BBS)
were flourishing shortly after the introduction of personal
computers in the mid-1970s.
The advent of large multi-user systems (such as AOL,
CompuServe) and the Internet itself enabled a further
style of communication known generally as “Chat”. Chat
consists of two to twenty or more people who appear
together on a common channel of communication (often
known as a “channel”, a “chat room”, or simply a
“room”). Any of the participants can type a line of text (a
“turn”). By pressing RETURN, the turn can be sent to all
of the other participants. Chat has been incredibly
popular since its inception, and continues to grow with the
expansion of the Internet. Currently there are hundreds of
chat servers, each with dozens of active rooms.
Despite (or perhaps due to) its popularity, the user
interface presented for chat has remained largely
unchanged since its first inception. A typical chat
interface consists of a list of participants, a text history
window, and a single line for text entry (figure 1).
Figure 1:MS-Chat, a standard chat client
As a communication medium, chat is not without its
problems. The conversations in chat tend to be hard to
follow. People have trouble identifying speakers and
remembering who said what. Within a particular room or
channel, there are often multiple conversations or
“threads” going on at the same time, and a single user
often participates in multiple threads at the same time. As
a result, chat conversations tend to be confusing and
disjointed, with a lot of repetition and corrections.
Studies have suggested that in a typical chat conversation,
as many as 40% of the turns are repetitions or corrections
to misunderstandings of previous turns [9].
PREVIOUS WORK
There has been a great deal of work on computer-
mediated communications (CMC) in the past twenty
years. CMC generally considers chat along side of other
computer-based communications, such as E-mail.
Despite the number of years of research, there is almost
no literature evaluating the fundamental user interface of
social chat software, and very few examples with actual
quantitative data on chat conversations [9].
Much work has been focused on the sociological or
linguistic properties of CMC [2][10][12][13][14]. This
research typically deals with the ways in which users
attempt to overcome the limitations of chat software, and
the degree to which these attempts are successful.
CMC is also studied as a component of Computer
Supported Collaborative Work (CSCW) [1] [3] [5] [6] [8]
[15]. In this context, we see more work on the effect of
specific user interface notions or affordances that support
groups of people working together to accomplish some
task. While some of these notions might be applicable to
the more unstructured communication of a social chat,

they mostly involve notions of floor control or turn-taking
that are not appropriate in the free-for-all of a chat room.
CMC is also frequently discussed in the educational
domain [7][9][11]. The interest here seems to be largely
based on the increasing number of classes that are being
held “on-line”, in which CMC serves as the only
communication medium available. The bulk of this
writing is geared around the particular problems that both
instructors and students encounter due to the limitations
of CMC.
FACTORS INFLUENCING CHAT EFFICIENCY
There are many factors that impact the efficiency of chat.
Here we list several of them, and show how they
interrelate to each other.
Lack of recognition – users find it difficult to associate a
nickname with what a user has said in the past if they do
not already know the speaker. Some clients allow users
to set the color and/or style of their text to help out with
this problem.(mIRC). Other clients allow the user to
choose a picture or cartoon to represent them in a virtual
space (MS-Chat, vChat).
Lack of intention indicators – it is difficult or impossible
to tell if a particular comment was intended to be directed
towards a particular user or turn. In transcript 1, Red’s
answer is ambiguous and could apply to either Black or
Pink’s question.
Black: What is your favorite musical?
Pink: Where are you from?
Red: Chicago
Transcript 1: intention confusion
Typing inefficiency – most people type much slower than
they talk, which results in a much slower rate of
communication. The average active chat user might take
30 seconds to enter a turn. This delay is experienced by
the receiver as a silence in the conversation [2] –
something that is established as being socially awkward
and easily misunderstood [17]. To prevent or mask this
delay, people tend not to wait for a response before
starting their next turn. This tends to diminish the
effectiveness of the slower typists and can lead to
confusion. In transcript 2, we see Black and Pink (both
fast typists) go ahead with another thread without waiting
for Red to catch up. As a result, Red’s comments are
ambiguous.
Black: Did you see that new Mel Gibson movie – I
think it is called “Payback”?
Pink: I saw the academy awards last night. Did
you watch it?
Pink: yep.
Pink: It was very violent, but funny.
Black: You saw it? You liked it?
Black: How did it end up – who won?
Red: I heard it was good.
Pink: It was OK. At least Titanic didn’t win
everything.
Black: I guess you can only be king of the world
once.
Transcript 2: thread confusion
Lack of status information – there is often no way to tell if
a user is actively participating in a conversation or has
wandered off from the computer all together. This often
results in out of order turns, where someone asks a
question again in a different way, not realizing that
another user has been composing an answer for the first
question. (See transcript 3).
In this transcript, 10 seconds had passed after Black asked
his first question. Black assumed that this inactivity
meant that Red was not interested, so he asked another
question. However, Red, who is a slower typist, was
actually typing his response the entire time. Now it is
unclear to Black what question Red’s answer is in
response to.
Black: Did you want to go skiing tomorrow?
Black: Or maybe we should just go to dinner?
Red: That sounds good – count me in. I think
that Blue will want to go as well.
Transcript 3: inactivity-related confusion
Lack of context – there is no way to tell what has gone on
before in a conversation or what people are talking about
now. This is true in a normal spoken conversation as
well, but is a particular problem in chat because of the
number of people constantly entering the conversations
and the expense of using additional turns to bring new
users up to speed.
High signal-to-noise ratio – since most of the participants
in chat do not know each other, there is a great deal of
introductory socialization phrases in any chat
conversation. In addition, due to some of the previous
factors mentioned, a large percentage of chat turns are
corrections, clarifications, or repeats of previous turns.
General uselessness of the chat history – although most
chat interfaces save the entire history of a particular
session, it is rarely used in the chat context. For example,
once an answer to a question has scrolled off the screen,
users are far more likely to ask the question again rather
than scroll back into the history. Part of the problem is
the real-time nature of the chat – when you scroll back,
you miss out on new things that are said. But even when
the history has been saved for off-line review, most

people still find it extremely ineffective to use. The out-
of-order turns and high signal-to-noise ratio make it
almost impossible to read.
With all of these difficulties, and with chat having
essentially the same user interface it had from its earliest
console days, we felt the time was ripe to experiment with
some alternative chat interfaces.
VIRTUAL WORLDS TECHNOLOGY
In prototyping different chat UIs, we were aided
tremendously by the Virtual Worlds platform, created
within our group at Microsoft Research. The Virtual
Worlds platform [16] is COM-based technology that
allows programmers to quickly and easily create multi-
user applications.
Virtual Worlds provides a persistent, distributed dynamic
object system. Objects can be created and modified on
the fly, including adding and removing properties and
methods. The system takes care of managing client
connections and replicating data between the server and
the clients.
Clients can be created using any COM-compliant
programming language, including DHTML, Visual Basic,
and C++. The combination of DHTML and the Virtual
Worlds platform makes an excellent rapid prototyping
environment for multi-user applications.
More information on the platform, including downloading
directions, can be found on the World Wide Web at
http://research.microsoft.com/vwg/
THE “CONFERENCE” USER STUDY
We began our project by getting some data on the current
chat interface, in order to have a baseline on which to
compare alternative representations. We used a standard
user study we call the “conference.” It has been used in
the past to compare such things as teleconferencing, video
conferencing, and face-to-face meetings.
In this study, a group of four to six people are asked to
participate in a pair of on-line conferences. In the first
conference, the groups would have to agree upon 5 videos
to rent for a weekend retreat. In the second, they would
decide which 5 albums to bring along on a long road trip.
Typically, the two conferences are done using different
technology (such as teleconferencing vs. face-to-face) in
order to compare the differences.
Each conference lasts approximately 30 minutes, after
which time the users are asked to complete a short
questionnaire. The questionnaire elicits factual questions
about the conference (such as, what types of music the
user believe each of the other participants enjoyed), as
well as qualitative questions (such as how they feel the
communication flowed.) In addition to the surveys, the
users may be videotaped for later analysis.
ATTACKING THE PROBLEM: STATUS
Most of the user interface improvements in existing
commercial chat software are related towards improving
the handling of multiple threads. This includes such
things as giving users the ability to carry on multiple
private conversations [MS-CHAT], or allowing them to
indicate precisely which person or which turn they are
responding to [V-CHAT, USENET].
While we agree that it is worthwhile to explore the
problem of thread management, we have for this study
focused our energies on trying to understand and
eliminate the initial conditions that lead to the multiple
overlapping threads in the first place.
In spoken conversation, people do not engage in more
than one line of conversation at the same time. We have
social conventions against interrupting, and people are
wait for other people to respond before speaking again.
Silence in conversation is something we are conditioned
to avoid, and spoken conversation has far fewer patches
of “dead air” than exact same conversation typed.[17]
This led us to believe that we might eliminate many of the
problems with threading by giving the users something to
do after they complete a turn other than starting another
turn, that did not feel like the equivalent of silence. We
also suspected that if users where aware of what other
participants were doing, they would be less inclined to
enter turns that were simple status requests. These two
suspicions led to the design of the “status client.”
STATUS CLIENT
We designed a prototype of an interface that showed the
status of each user, as determined by keyboard activity.
Figure 2: Status Client
This status client keeps the same basic three-pane layout
of a standard text chat client but augments it with several
type of continuous status information. These are:

1) Keyboard activity fader – a user’s name in the user
list would highlight whenever a key was pressed,
then fade back over 5 seconds to the original level
2) Last line indicator – a user’s last turn would appear
next to their name in the user list
3) Intermediate text indicator – each time the user
pressed a key while composing their turn, the new
text would appear next to their name in the user list.
This text would be in a different color than the
regular last line color to indicate its preliminary
nature.
4) Last line fader – along with 2 or 3, the last turn text
would fade out over 10 seconds.
Our prototype was able to activate each of these features
individually, with the exception of 4), which required
either 2 or 3 to be enabled. The prototype was coded
using DHTML pages with VBScript on top of the virtual
worlds system. For our experiment, we ran with options
2, 3, and 4 enabled. In this configuration, users would see
every character every user typed as they were typing it,
next to that user’s name. When the user pressed
RETURN, their text would be added as a new turn to the
scrolling history, and would gradually fade from the user
list.
STATUS CLIENT TESTING AND RESULT
For our first test, we ran two five-person groups (A and
B) through the conference study. Group A used the client
in standard mode with no status features for their first
conference, and with status enabled for their second
conference. This sequence was reversed for Group B.
Rather than videotape the participants, we used client-side
logging of the conversation. Each keystroke, mouse
action, and incoming event was time-stamped and logged
to a text file. Most of our analysis centered on these files.
The participants were all experienced chatters, and as a
group could type quite fast, averaging 250 CPM.
From a qualitative point of view, the users liked the status
client, and all but one preferred it to the standard
interface. Most people stated that they liked to watch
other people type, but also admitted feeling embarrassed
by their own typing skill being on display. Most users
said they focused all of their attention on the user list with
its live status, and paid little attention to the text history.
We were looking at several factors in the log files. The
first was a quantitative analysis of the number of
correctional turns of a percentage of total turns. We
tracked this information by user as well as by conference.
Our suspicion was that in the status prototype, there
would be fewer corrective turns, as people would have
more context over what other people were responding to.
We were not, however, able to ascertain a difference in
our study, as there were almost no turns dedicated to
correction in either UI. (we will discuss this further in the
conclusions).
The status client did show fewer out-of-order turns in the
history. User would start to type an answer a question
before the asker pressed RETURN, but they would wait to
press RETURN themselves until the question was in the
history. This resulted in questions being directly followed
by their answers in the history, and made the history
subjectively more readable.
We also wanted to see the affect on the total number of
turns. Our suspicion was that in the status-enabled client,
there would be fewer actual turns, as people would see
that other users were typing and would wait for them to
finish before starting another turn. However, the data
showed that there was no significant impact on the
number of turns. Regardless of interface, both groups had
fewer turns on their second round than their first.
We were also looking for other artifacts from the real-
time nature of the status information, such as would
people be more or less likely to participate. Here we saw
something quite interesting that we did not anticipate:
when their keystrokes were being sent over the line as
entered, users’ increased both the speed and accuracy of
their typing. We tracked the typing speed in uncorrected
characters per minute, and the accuracy of the typists as a
ratio of the number of keystrokes needed to generate a
turn of a given length. With the standard client, our users
averaged 250CPM at 75% efficiency. With the status
client, speed increased to 280CPM at 85% efficiency.
Our suspicion is that this was due to the feeling of being
“on display”, and might well go away with time as the
novelty of the system goes away.
In general, the increase in typing speed and accuracy did
not affect our users’ ability to participate. However, one
user (Red) had only marginal typing skills. With the
standard client, Red typed 180 CPM with 64% accuracy –
much worse than the other participants. With the status
client, Red increased to be on par with the other users
(224 CPM with 85% accuracy). But this accuracy came
at a cost: he contributed less than 5% of the turns in the
conversation (versus 20% for the average user), and his
turns were only 15 characters long (vs. 30 for the average
user). A follow-up interview confirmed that Red was
concentrating so intently on his typing that he limited
himself to only the occasional short turn.
With both interfaces, we noticed that people did not use
the history much, if at all. They would not scroll back in
the history to find a piece of information that was located
off the screen. In fact, 10% of the turns were repeats and
restatements of previous turns that were off the screen. In
all four rounds, not a single person used the scroll bar
during the tests. Many people suggested that it would be
helpful to have a whiteboard or other persistent
affordance during the conversation, similar to what one
might find in NetMeeting or other CSCW software.

Our suspicion was that people would have an easier time
answering the questionnaire with status client, due to the
history being more readable. However, this turned out
not to be the case. There was no significant difference
between clients in the accuracy of the survey. Most users
complained that it was too hard to find things a particular
user said and suggested it would be nice to be able to
filter the history by user.
DESIGN AND REDESIGN OF THE FLOW CLIENT
One of the problems immediately apparent in the status
client was that a great deal of the time sequencing
information was lost in the history display. For instance,
if you were watching the screen, you could tell that
someone else had started typing after you did, but if they
hit return first, their turn would still come before your
own in the history. This lead us to suspect that perhaps
the display of turns in a linear fashion based on
transmission time might not be the best representation.
Based upon this observation as well as the qualitative and
quantitative feedback of the status client, we began to
design a new client. Our goal was to make something that
would have the real-time benefits of the status client but
that would preserve the temporal ordering of turns and
provide easier matching to the speaker.
Our fundamental realization was that chat is closer to a
streaming media type than anything else, and that it might
be efficient to view it with a more traditional streaming
media interface – the multi-channel timeline. These
timelines are seen in many streaming media editing
products, such as Macromedia’s Director, Adobe Premier,
and MIDI sequencing programs. In these interfaces, time
flows from right to left, and the screen is broken up into a
number of horizontal channels for different event sources,
such as different MIDI channels, sprites, or film files.
Our first take on this can be seen in figure 3 – the Flow
Client. The Flow client is an MFC application written in
C++, using the virtual worlds technology for its
client/server architecture.
Figure 3: Flow client version 1
In this first version, we can see the user list on the left.
Each user has a track on the screen. Time is constantly
flowing, and the timeline display is constantly scrolling
from the right edge of the screen (which is NOW) towards
the left, at approximately 1 inch every five seconds. As a
user begins typing, a colored box begins to stretch out in
their timeline. The left edge of the box is at the time
where they started typing. The right edge is set at the
time they press return. In the meantime, the right edge
stays aligned to the right side of the screen, and the text in
the box is continuously updated with what the user has
typed as they type it. If the user becomes inactive, the
yellow box begins to fade until they continue typing. At
any time, a user can mousedown and drag the timelines to
peek back a bit in time. When they released the button,
time would snap back to the present. We termed this
“snap-back scrolling.” In addition, the user could click a
button to stop the time flow and scroll at their leisure.
Preliminary user tests revealed that people liked the client
in general, but were bothered by the names being on the
left while the new text was entered on the right. They
also were confused by having to enter text down below
while it showed up up above. Users also reiterated their
request for some sort of persistent information in the UI.
Figure 4: Flow client version 2
These comments lead to the design of Flow Client 2
(figure 4). In this version, the screen is divided
horizontally into three sections. On the rightmost is status
information that is out of the timeline. This is basically a
single line of text that each user can set. In the middle is
the timeline view, and at the right we see the user list. As
a user types, the yellow box stretches out into the timeline
as before, but the text is placed next to their name in the
user list instead of in the yellow box. The text appears in
blue to indicate that it has not been committed. If you
paused while typing, your text would slowly start to fade
to a lighter blue. Your own typing is done directly into
the user list, and the text entry frame on the bottom has
been eliminated. When the user commits the turn by
pressing RETURN, the text is placed into the timeline
rectangle and the text next to their name changes to black.
This is similar to the status client.
We tested in front of users again. They liked the new way
the user list worked, but did not like the persistent
information area on the far left, saying it was “hard to
read” and “confusing”. Users were also bothered by the
scrolling of the timeline ruler at the bottom of the screen.
They felt that it was distracting and did not tell them what

they really wanted to know, which was how long ago a
particular turn took place.
This input lead to the third and final redesign:
Figure 5: Final Flow Client
In the final version (Figure 5), we eliminated the left-hand
pane, and we changed the timeline to be a relative time
indicator from “now” back to –30 or so seconds
(depending upon window size).
FLOW CLIENT TESTING
Once we had a design that was deemed acceptable based
on preliminary qualitative feedback, we conducted formal
user studies.
In testing the previous interfaces, many of the users gave
feedback that the conference room study was not
representative of a “typical” chat experience. This was
because it had a set topic to discuss, rather than the free-
form topic-a-minute nature of normal chat. Indeed, in
analyzing the data, there were very few off-topic turns in
comparison to normal chat – only 3%.
Therefore, in the flow client tests, we attempted to more
closely simulate a typical active chat room. We first did
some less formal tests with two groups of five users from
the status client tests. We then later tested five groups of
four subjects in a more controlled, videotaped
environment. Each group participated in two back-to-
back chat sessions, once with the flow client and once
with the standard client. The chatters were instructed to
enter the chat session and get to know a few things about
the other participants. Suggestions were given for
conversation topics, but no particular topic or goals were
mandated. In addition to the study participants, each chat
session also had two “confederate” chatters who were
instructed to act as study participants and keep the
conversation flowing. The confederates were sometimes
asked to bring up a particular subject (such as a knock-
knock joke) that could be referred to later to examine the
history usage.
RESULTS
Quantitatively, we expected the flow client to share many
of the characteristics of the status client with regard to
typing speed and accuracy. This was borne out by the
data from users who had been tested with all three clients.
In general, people typed more accurately and no more
slowly when their typing was on display. Total number
of turns, % of conversation per user, and average turn
length were all similar between the clients.
We suspected that users would find the history easier to
use and would consult it more readily than in other
clients. The history was indeed consulted more, as every
participant used the snap-back scrolling feature many
times. The most-often stated reason for using it was to
catch snippets of conversation that had left the screen.
The number of turns dedicated to repeating information
dropped from 5% in the standard client to less than 1%.
Despite this, many disliked the presentation of the history
in the Flow Client. People felt that the UI was unfamiliar
and preferred a vertical scrolling format. They felt the
conversation left the screen too quickly and that all of the
live typing areas were too hard to monitor. People were
less accurate in completing a task that involved browsing
the history, and felt more anxious with the Flow client in
general. A similar complaint was the feeling that the flow
client, with its large gaps between subsequent turns, made
inefficient use of screen space compared to normal chat.
This perception is interesting because an analysis of the
conversation shows that in both cases, the flow client is
actually slightly more efficient.
With four or fewer people in the room, a normal chat
interface can potentially provide an efficient use of space.
That is, each user can have four or more turns on the
screen at once. We have seen that in chats with fewer
participants, turns tend to be longer, which also increases
the efficiency of the standard display of text history.
Figure 6a shows a graphic representation of the same chat
in a standard client and the flow client. Note that the flow
client shows three to four turns for each user, while the
standard client can display six of more.
figure 6a: standard UI (left) and flow UI (right)
displaying 3 people in chat
As more members join the chat, the flow client continues
to display three or four turns per user, while the standard
client becomes less and less efficient (Figure 6b).

Figure 6b: standard UI (left) and flow UI(right)
displaying 12 people in chat
With six or more speakers, turns in the standard client
start to scroll off the screen very quickly, and the number
of turns between adjacency pairs expands. This requires
that more turns be dedicated to clarification and
correction. In the flow client, the scroll rate remains
constant regardless of the number of speakers.
CONCLUSIONS
Our usability results revealed some design problems in
the application:
1) Continuous scrolling – a substantial problem users
found with the client was with the large continuously
scrolling area. Several users complained of a vertigo-
inducing effect from having 75% of their field of
vision scrolling slowly to the left. In the future, we
hope to experiment with implementations that do not
use smooth scrolling.
2) Too Much Activity - Users were confused and
distracted by the amount of activity on the screen.
3) Novelty – as with any new interface, we faced
resistance from users familiar with the established
interface.
4) Flicker Problems – On certain machines, the text
boxes would flicker as the users entered text. The
flicker was extremely distracting to the users,
especially when a half-dozen people were typing at
the same time. (Affected users suggested displaying
an epilepsy warning when the application starts.)
This is a technical problem that we are hoping to
correct in a future version, as several users mentioned
it as the primary reason for disliking the Flow client.
FUTURE WORK
Despite the problems with our initial design, we feel that
streaming media representations of chat are worth further
investigation.
We are hoping to set up a chat site where we can conduct
longitudinal studies of different chat clients. We feel
many of the concepts introduced require some time for
users to adapt to them before we can get valid statistical
results. This seems to be particularly the case with
transmitting each character as it is typed. We hope to
have such a site up and running during the Fall of 1999.
One of the problems of busy public chat rooms has been
the lack of context for new people entering the chat. We
have tried experimental interfaces that preserve that last
hundred or so turns of chat for a user when they connect,
but the usefulness of this has been hampered by the
overarching uselessness of chat histories in general. We
suspect that with a variation of the flow interface, such a
history would provide added value and context, and we
hope to test this soon.
We are particularly interested in user interfaces for
indicating the status of the user (typing, reading, away
from the computer, etc.) We are investigating ways of
doing this that do not require sending every character
across the line.
Finally, we want to explore the ability to show chat
overviews. We suspect that the flow client’s display
readily lends itself to a symbolic overview that can
provide information (such as, which people talking and
how often). We hope to add some overview capability
into the next version of the client.
ACKNOWLEDGEMENTS
We would like to thank Shelly Farnham for her assistance
in running the Flow Client user studies, and the Microsoft
Usability Lab.
REFERENCES
1. Ackerman, M., Starr, B., “Social Activity Indicators
for Groupware”, Computer, June 1996, 37-42
2. Altun, A., “Interaction Management Strategies on
IRC and Virtual Chat Rooms”, SITE 98 proceedings,
March 10-14, 1998, 1223-1227
3. Beaudouin-Lafon, M., Karsenty, A., “Transparency
and Awareness in Real-Time Groupware Systems”,
UIST proceedings, 1992, 171-180
4. Bowers, J., Pycock, J., O’Brien, J., “Talk and
Embodiment in Collaborative Virtual Environments”,
CHI 1996 proceedings, 55-65
5. Dourish, P., Bellotti, V., “Awareness and
Coordination in Shared Workspaces”, CSCW
proceedings, 1992, 107-114
6. Dourish, P., Bly, S., “Portholes: Supporting
Awareness in a Distributed Work Group”, CHI
proceedings, 1992, 541-547
7. Dringus, L., Adams, P. (chair), “A Study of Delayed-
Time and Real-Time Text-Based Computer-Mediated
Communication Systems on Group Decision-Making
Performance.”, Dissertation, Nova University, 1991,
217 pp.
8. Ellis, C., Gibbs, S., and Rein, G., “Groupware: Some
Issues and Experiences”, Communications of the
ACM, Vol. 34, No. 1, 1991, 39-58
9. Garcia, A., Jacobs, J., “The Interactional
Organization of Computer Mediated Communication
in the College Classroom”, Qualitative Sociology,
Vol. 21, No. 3, 1998, 299-317

10. Murphy, K., Collins, M., “Communication
Conventions in Instructional Electronic Chats”, First
Monday, Vol. 2, No. 11, 1997
11. Powers, S., Mitchell, J., “Student Perceptions and
Performance in a Virtual Classroom Environment”,
American Educational Research Association Annual
Meeting, March 1997
12. Reid, E., “Electronic Chat: Social Issues on Internet
Relay Chat”, Media Information Australia, 67, 1993,
62-79
13. Rintel, S., Pittam, J., “Strangers in a Strange Land:
Interaction Management on Internet Relay Chat”,
Human Communication Research, Vol. 23, No. 4,
1997, 507-534
14. Sproul, L., Keisler, S., “Reducing Social Context
Cues: Electronic Mail in Organizational
Communication”, Management Science, Vol. 32, No.
11, 1986, 1492-1512
15. Tang, J., Issacs, E., and Rua, M., “Supporting
Distributed Groups with a Montage of Lightweight
Interactions”, CSCW proceedings, 1994, 23-34
16. Vellon, M., K. Marple, D. Mitchell, and S. Drucker.
“The Architecture of a Distributed Virtual Worlds
System.” Proc. of the 4th Conference on Object-
Oriented Technologies and Systems (COOTS). April,
1998.
17. Wiemann, J., Knapp, M., “Turn-Taking in
Conversations”, Journal of Communication, 25,
1975, 75-92
mIRC. By Khaled Mardam-Bey. Shareware software
available at http://mirc.stealth.net
MS-Chat: by Microsoft, Inc. Free download available
from http://www.microsoft.com
v-Chat: by Microsoft, Inc. Free download available from
http://www.microsoft.com

1999-UIST-Alternative interfaces for chat

More Related Content

What's hot (20)

Similar to 1999-UIST-Alternative interfaces for chat (20)

More from Marc Smith (20)

Recently uploaded (20)

1999-UIST-Alternative interfaces for chat