Tracy Narine, Andrea Leganchuk, Marilyn Mantei & William Buxton
Communication often drops dramatically between group members who move a significant distance from their work group despite the provision of mediated communication support. Although the poor quality of the communication exchange via mediated communication is certainly responsible for a significant portion of this drop, lowered suggestion, availability and comfortableness of the communication contact may also cause a drop. With the intent of re-creating these background contact characteristics found in a co-located group, low resolution video pictures of widely dispersed co-workers were exchanged and displayed on each others' workstations via a system called Postcards.
Data on system usage was logged and also collected by questionnaire administration. The study found that all users (close and far) tended to use the system extensively for coordinating telephone calls and visits. Users also reported knowing more about what is happening at work and knowing more about their co-workers time pressures. They saw no change in the amount of telephone traffic, E-mail or informal meetings that took place. Thus, Postcards provided useful cues for contact coordination but did not affect the frequency of communication. In addition, the computer logs showed that the foreground activities of locating and seeking contact information were weakly used in contrast to the report of high usage of the picture display. Overall, the results suggest that the continued awareness of both the presence of one's co-workers and a knowledge of their contactability state are a useful part of maintaining distant contacts.
Postcards, Portholes, Awareness, Video Desktop Conferencing, Media Spaces, Coordination, Collaboration
When co-workers are not co-located, project deadlines slip, miscommunication is more common and optional projects never get carried out . In the modern world, the extensive amount of electronic communication possibilities suggest that distant work collaborations should not be a problem. This is especially true with the widespread use of the cellular phone, electronic mail and video conferencing. Researchers argue that this is because existing media are too limited. Media spaces were developed with the belief that a properly managed video exchange would encourage collaboration . It has been difficult, however, to find much difference in the work performed when telephone is compared to video in experimental settings [12, 14].
Hollan and Stornetta  argue that communication systems differentiate co-located workers from distant workers because they fail to add the richness of face-to-face communication. They further suggest that communication media can add value to communication only if they move away from attempting to be a poor second to face-to-face encounters. We agree with this approach and argue that one mechanism that could enhance both mediated and face-to-face communication is collaboration awareness.
We define collaboration awareness as one case of background awareness [1,2] in which a co-worker is peripherally made aware of the potential for collaboration with a colleague. We use a variation of the Portholes system developed at Rank Xerox EuroPARC . We call this system Postcards. Postcards, like Portholes, takes frequent low resolution video snapshots of co-workers and broadcasts them via computer network for display on co-workers' workstations. It is these updated pictures of co-workers which we believe make co-workers collaboratively aware.
In attempting to demonstrate what value collaboration awareness might have for co-workers, we need to explore why communication drops off so dramatically when colleagues are not co-located. Co-located workgroups engage in multiple forms of communication creating a social network , much like that shown in Figure 1. Each node in the network represents an individual and each link represents a level of communication traffic, e.g., the number of times per day, week or month that the two individuals communicate. The network can represent different types of communication (e.g., social, work exchange, etc.) via different media (e.g., telephone, electronic mail, etc.).
Consider a co-located workgroup that uses telephone, voice mail and electronic mail extensively. If one or more group members moves to a new location that is non-proximal the communication links between these individuals and the rest of the network soon drop below a typical threshold. This is strange especially in light of Haythornthwaite et al's study  which describes a co-located workgroup whose members had weakly overlapping work schedules but yet maintained strong contact through E-mail communications. These individuals were co-located in space but certainly not in time. One argument given for the drop in communication is a failure of the communication media to provide sufficient richness to the contact making contact with co-located colleagues preferable. We do not discount the studies that show that mediated communication is not as rich as face-to-face, but we suggest that the drop in communication may also be due to weak support for the intent to communicate.
We see this intent thwarted by three different mechanisms at play. They are the lack of:
Intent stimulus is any event which triggers an intent to communicate. In the design rationale for the Cruiser system, this was a colleague walking down the hall and seeing someone available in their office for conversation . This form of stimulus is self-reinforcing when workers are co-located, but dwindles when workers are distant because of the need to maintain the exchange of faxes, E-mail or phone calls to continue the stimulus. Once one of the parties involved travels or is extremely busy, the chain is broken and the contact dwindles. Even in the workgroup with widely diverse work times, scheduled meetings where attendance from all workshifts was required, tended to restart the communication exchange.
Intent opportunity is the probability of making a communication contact once the intent to communicate is formulated. In a group with electronic mail support, this opportunity is always there, but its use also depends on the information to be communicated. As an asynchronous medium, it lacks the spontaneity and immediacy of synchronous forms. A common example of thwarted intent opportunity is the frustration of playing telephone tag. In co-located environments, intent opportunity will be much higher because so many more opportunities for communication will exist. Even in the diverse time workgroup, individuals modified their schedules to increase opportunities for contact when needed.
Intent comfort is the sense of the social acceptability of making the communication contact. When people are co-located, they cannot help but have contacts in the hallway or lunchroom, all of which make the contact more natural and less intrusive. When people are not co-located, these easily enacted contacts go away and there is more of a social barrier for making the contact. In addition, there is less shared information to use as conversational starters.
We argue that despite its lack of richness and depth, mediated communication might be used more extensively at a distance if a form of background awareness enhanced the three communication intents. We study Postcards usage to examine this possibility. We have not looked at a widely separate group for our study but at one that is slightly separated. Face-to-face communication in the organisation studied requires walking either up or down or opening locked doors. Thus, we are eliminating any mediated characteristics of communication and looking at how Postcards affects the intent to communicate when the face-to-face communication has a large startup cost.
We expect the presence of the video snapshots to cause co-workers to think about others and thus, generate a communication exchange. If this is so, we should find an increase in telephone calls and visits with the installation of the system. We also expect the video snapshots to indicate the availability of the other person. Thus, once the intent to communicate is established, we should see the snapshots being used to ascertain if the communication can take place. We should also see the use of the sensing agent that is built into Postcards. This agent can be asked to let a person know when the person with whom communication is desired returns to their office.
Finally, people can use Postcards to observe their co-workers and learn more about them. If a group is already co-located, this would not seem to be necessary. Thus, if the group is using Postcards in this fashion, it suggests that it is a useful tool for increasing the comfort intent by providing an understanding of one's fellow workers and their work characteristics. There is a possible fourth form of collaboration awareness that we will call intent aggregation which we do not cover in this paper. Monge et al  were able to predict optimal meeting times after studying four days of co-workers' organizational proximity. We view this intent as a correlation to intent opportunity.
In Postcards, we extend the concept of Porthole's awareness by characterizing three forms of awareness mechanisms that might be at play. Postcards has been built with additional tools to support these mechanisms. Having a good image of one's co-workers availability at all times provides intent stimulus. Providing a mechanism that maintains the intent to communicate despite a co-worker's temporary absence also supports intent stimulus. Providing contact tools for instantly initiating a contact and images to indicate the availability of the contact support intent opportunity. Intent opportunity is also supported by listing contact information through other media, e.g., email address and fax and telephone numbers. Providing regularly updated images of co-workers supports intent comfort. Portholes provides both the co-worker images and the contact capabilities. We have added additional features to Postcards that are based on a more detailed analysis of the background awareness Portholes supports. These are described in the following paragraphs.
The main window of the Postcards user interface is shown in Figure 2. Each image is a grayscale picture 104 pixels wide and 80 pixels high. We have intentionally kept the pictures small and unobtrusive to avoid invasions of privacy. Smaller pictures also save bandwidth. Similar to the original Portholes system, we display the title and the time the snapshot was taken.
A user selects from a set of pictures under the Pictures menu (Figure 3) by navigating through the list of Postcards servers available. Any picture marked with an asterisk is public and therefore available to anyone using the system. Pictures that a user is currently receiving are marked with a check mark.
The Tools menu shown in Figure 4 has eight options. If Postcards resides in a media space, video or voice contact are supported via Contact and Hangup.
If the Sense Option is selected for a given image, the user is notified when activity is next "sensed" in the selected person's office. The Sense option acts like a secretarial agent who lets the user know that a given person has returned from lunch, a meeting, etc. When the sensing alert is sounded, a dialog box appears on the user's screen. It will timeout and disappear after 20 seconds in case the user is not around. Sensing continues to function until canceled by the user.
Sensing works as follows. A simple image differencer compares the amount of pixels that have changed between the previous image and the most recently captured image. If the change is above a set threshold, a sensing alert is sent.
Two other functions provided on the menu are Information and Comment. Information displays a list of useful contact details such as the telephone number and E-mail address of the person associated with the image.
For this study, Postcards was deployed in a private company. The Postcards developers were at the University of Toronto and not readily available for comments, questions and user complaints. Comment allows Postcards users to write and automatically send comments about Postcards to its developers.
Users can also turn their own picture taking on or off with the Picture taking option. The final tool of the menu, Restart background software, allows users to restart software system components that may have failed in some way. This becomes a very useful selection when the Postcard system administrators are some distance away.
Postcards consists of two UNIX processes and an application. The components are the Postcards Server (pc), the Postcards Remote Frame-Grabber (grab) and Postcards Client (pcc). Figure 5 illustrates the components and their information flows. The lines between components represent the two way communication that supports the image and data transfer.
A subscription mechanism governs the low level transfer of data such as a list of possible images or the actual image bitmaps. Upon startup, a client subscribes to data and images from the local server. This opens network channels which passively listen for subscription data and images to arrive. The subscription service also keeps the data in the client current. When a server process is restarted, it sends a "server restarted" message to all of its clients. The clients, upon receiving this message, remove all information from their database particular to that server and re-subscribes, to receive current information.
The Postcards Server is at the center of the system and knows what clients and image grabbers are available. Control information such as frame grab frequency and information about the image are also kept within this process. The Postcards Image Grabber (grab) captures an image of the user's office every N minutes. In this installation, we captured images every four minutes. A sound is played to alert the user that their picture is about to be captured and distributed.
Postcards supports two types of image grabbers. The first type uses a media space infrastructure to perform the image capture. The second type uses the video capture hardware of each user's workstation. Since our study site did not have a media space, we used the second type of frame grabber.
The site chosen for this study was a high-tech company.
Eight individuals participated in the study. We have divided the study group into team and peripheral members. The team members form a close knit group of six individuals. The peripheral members are two individuals in the company who do not carry out the same work as the team but either performed support work for them or served as a liaison between them and the rest of the company.
The team is made up of a senior manager, a project manager and four workers. This team is dispersed within the company having offices on three separate floors. Figure 6 gives an approximate measure of this dispersal. Distances between offices are shown in number of footsteps, doors, locked doors and stairs.
As Figure 6 illustrates, there is a mix of open and closed offices within the study group. Only the three managers have offices with doors. To travel between floors, security doors and doors into stairwells must be opened. Only two participants (1 team member, 1 peripheral) are within a line of sight of each other. Given the distributed nature of the group, the cost to initiate any face-to-face communication is high, not unlike that of distant group members.
The team's senior manager is extremely busy and travels frequently. A queue of people often form at his door when he is in town. When the senior manager is not available, the team's project manager guides the team's work. Three team members report directly to the project manager. The other team member works independently and reports to the senior manager. The team has one regularly scheduled meeting per week. Peripheral members of the study group do not meet regularly with the other members but do have informal spontaneous meetings with one or more of the team's members.
Postcards usage data was collected for a total of three weeks. We logged user actions, distributed a questionnaire, conducted informal interviews with subjects and gathered E-mail comments.
The Postcards client logged all user actions and the duration of any user requests. We captured the amount of time that the person kept the Postcards client open. This told us whether the group actually had used Postcards. We logged the number of pictures that the users displayed. This told us the subgroups of communication that might be forming. We also measured how many times the participants made use of the Sense, Comment and Information functions. The use of Sense and Information would support our belief that Postcards provided the initial stimulus for a communication and also appropriate support for enacting that stimulus.
The questionnaire we distributed to the eight participants in the study focused on their perceptions and memory of what they did when they initiated a communication with a co-worker. It also asked what strategies they used to find a co-worker and what sparked a communication to take place. Answers to these questions would examine the validity of our collaboration awareness claims. On a more practical note, we also wanted to know how users felt about the operation and features of the system and what items they thought would be useful to add to Postcards.
Informal interviews with each of the study's participants provided us with usage anecdotes plus a suggestion list for problems to fix. The E-mail comments also gave us insights into problems users encountered with the system.
The responses to our questionnaire were uniformly similar and provided evidence that all three forms of collaboration awareness were being used to facilitate communication. Participants in the study used Postcards heavily to support finding the window of opportunity for their intent to communicate. They used it less as a reminder to communicate (intent stimulus) but reported that Postcards contributed to increasing their familiarity with co-workers (intent comfort). Anecdotes of more personal expressions via Postcards suggest that users feel quite comfortable with each other, but we have no evidence that Postcards in any way fostered this familiarity.
We asked a series of questions on how group members used Postcards when they wished to contact another group member. The form of the question was:
Before I (type of communication contact, e.g., telephone) someone, I check Postcards to see if they are in their office.
Users had three responses, (1) always, (2) sometimes, and (3) never. Table 1 illustrates that users used Postcards for verifying the availability of a co-worker whenever they telephoned or visited a colleague especially if the person resides on another floor. Six of the eight users always checked Postcards before traveling to another floor and exactly half of the users always check Postcards before phoning others. The two users who never check Postcards when visiting someone on the same floor are close enough to hear and see the people they wish to visit.
Users usually did not wait for a picture to be updated before they telephoned or visited suggesting that the time resolution of the pictures is enough for them to assume that contact opportunity has remained the same since the last image capture. Our questionnaire also asked if the images prompted a communication to take place. Their responses indicated that this happened occasionally for work related issues but not for social exchanges. Although this suggests that Postcards is not used for intent stimulus, it is unlikely that users can accurately report this phenomena as Nesbitt and Wilson  demonstrate in their classic paper. These answers indicate that users do not perceive that Postcards gives them any intent stimulus.
Intent opportunity was also supported by a question which asked what types of strategies people used when they wanted to make contact with a co-worker. The question took the form:
Please indicate the strategies you use to make contact if you need to talk with your co-workers and they are not in their office.
Table 2 lists the responses to this question. The results indicate that all participants used Postcards daily for checking the contactability of others. Since Postcards was not an available strategy until its installation, its 100 percent adoption by all group members as a strategy for determining contact availability indicates the usefulness of this intent opportunity support that Postcards provides. Users did not report using the sensing function for setting up a future contact possibility. This is corroborated by the logged data and by the individual interviews. We believed that use of the sensing function would be related to Postcards provision of intent stimulus, but the lack of immediacy of a contact is already supported by the voice and E-mail media which are used more than sensing.
We asked users to give their perception on changes that Postcards made to their work habits with the question:
Please indicate if Postcards has change your work habits in any of the following listed ways.
On average, the results show increased intent opportunity. In particular, users reported less travel between floors because of Postcards and less frustration in finding their co-workers. They saw no increase in their informal meetings, telephone conversations, or E-mail exchanges. Users also indicated that they have a better sense about what is happening with their co-workers and what their co-workers' work pressures are. This was also reported anecdotally when one co-worker indicated that he saw that another colleague was very busy with meetings in his office. These results are shown in Table 3.
We also asked how Postcards was used with the question:
In the following list please mark all the ways in which you use Postcards daily.
The results from this query are shown in Table 4. Users readily used the system to find lunch and dinner companions and also to check for an intent opportunity. They also simply used the system to view their co-workers. We interpret this viewing as a form of intent comfort, that is, a building of familiarity with one's co-workers.
We also asked users if Postcards had increased their awareness of working in a group. Users responded to a five point Likert scale with 1=strongly agree, 2=agree, 3=neutral, 4=disagree and 5=strongly disagree. The average of the responses was 1.75.
Most of our results are based on user's perceptions of how they use Postcards. The reliability of their responses is corroborated by log data on their use of the sensing and information functions which are quite low, as reported. The other results, however, are suspect until correlated with a behavior measure. The anecdotes about usage and the active user maintenance of the system suggest that Postcards' collaboration awareness is useful to the group. The user's responses to our queries of how it is useful fits our model of providing useful information for the three proposed types of pre-communication planning.
The questionnaire also asked users to evaluate various features of Postcards. About half of the study group found the sound played before picture capture to be disruptive and annoying. Most of the members who did not like the sound recorded their own. One of the participants situated in an open-office environment complained that the sound annoyed surrounding workers. This problem ceased when the user recorded a quieter sound.
Some users were uncomfortable with the picture being taken immediately after the warning sound. Consequently, a 2 second delay was added.
An important attribute of the system is the interval between subsequent snapshots. Within the study group, 4 out of 7 members were content with a 3 to 5 minute interval. The other 3 members wanted updates to be every 1 to 2 minutes. No-one requested a longer interval between snapshots.
The Postcards system attempts to keep all images current. In this study, we had one Postcards server and eight remote Postcards Frame Grabbers running. Network communication or software failure causes images to be out-of-date. This happened in the first week of the study and was not acceptable to the study group. As a result, the Postcards software was changed to "time-out" images that are older than their update period and to send out blank images to the clients. This was more acceptable since anyone in the work group could recognize a system failure. In this case, users usually called each other to get software restarted so everyone could again be part of the Postcards community.
The module that maintains current images within the Postcards server causes an interesting side effect. When an image "times-out" within the server, a blank is sent to all clients that have subscribed to this picture. This time-out usually happens before the new picture is taken. As a result, the client shows a blank and then the new picture appears. Although it's a software bug, several users have commented on it being an important visual cue to know when the picture is about to change.
There were several features that the users mentioned that they would like to see in a later release of the software. This included the ability to send out a standard set of messages to other users such as busy, not busy, lunch, in a meeting or on vacation. They were also interested in being able to send electronic Post-its to other users or to attach Post-its to their Postcard display. This would support a simple reminder system.
Another feature that was discussed several times throughout the duration of the experiment is the ability to immediately update any Postcards picture. We felt that supporting this functionality would infringe on the rights of users of the system. Especially in the case when any individual attempts to continuously update another's picture. In the current implementation, there is no method for a user to control accessibility of their picture. This type of functionality would be necessary before supporting instant updates of any individual's picture.
The questionnaire asked users to list any anecdotes about how they or a colleague used the systems in ways that were unique or different than what was expected. We elicited this information because it often suggests design changes and because it provide some insight into how the system is being incorporated into the workplace and worklife.
Three anecdotes are of interest for these purposes. In one case, a team member put a stuffed animal in front of the Postcards camera whenever they was away from their office or desirous of privacy. The use of the animal for both purposes conveyed a confusing message to co-workers. However, its very ambiguity provided a privacy protection that Postcards had taken away. Colleagues could not know if the person was actually absent from their office or intensely working on a project. The stuffed animal message did not thwart any of the intents to contact because it clearly said that contact was not possible but was removed at other times when contact was possible. This suggests incorporating similar ambiguous messages in Postcards.
Two users were also observed to be making faces (presumably at each other) but humorously shared with the entire work group. This suggests a high degree of comfortableness with other members of the work group since "faces" remain displayed for four minutes.
If members of the study group forgot to tell visitors about Postcards, visitors were often "spooked" by the sound announcing the picture capture. The sound is supposed to sound like a camera shutter clicking but if heard out of context, it is difficult to guess what the sound is. This suggests another problem with the design of Postcards. Visitors have not agreed to have their images broadcast to others, yet by visiting a Postcards office, they are suddenly visible to the entire work group. The work group is also visible to them creating the effect of visiting a virtual open office . Fortunately, the choice to limit the quality of the images also limits the invasiveness of visitor image capture.
Overall, we have found that Postcards was being used by the co-located group, but not completely in the ways we anticipated. Its primary usage was that of managing the intent opportunity - seeing if an individual was available for contact. Its secondary usage was that of co-worker observation which can be weakly interpreted as supporting intent comfort. We saw little evidence of intent stimulus. However, the group, although dispersed in the company, was not separated sufficiently so that other forms of intent stimulus were not already available. Indeed, it can be argued that because the group already had a communication network in place with tight links between all participants, that intent stimulus was not necessary and therefore not measurable. A different form of study would need to enacted, one in which communication links had fallen below a given threshold and were rejuvenated by the use of Postcards.
As a collaboration awareness tool, the Postcards usage demonstrates its simple effectivity in a co-located environment for coordinating communication. As such, it meets Hollan's criteria that a CSCW tool, to be used by both distant and local groups, must provide the same added value to both groups. Postcards appears to do this.
Many thanks to the individuals who participated in this study, Canada's National Science and Engineering Research Council, Alias|Wavefront Inc. and British Telecomm which helped fund this research. Also, we thank Xerox for providing us with the Portholes code. We thank the developers that have been associated with this project, in particular Tom Milligan, Chris Passier, Dominic Richens and David Audrain.
1. Buxton, W. Integrating the Periphery and Context: A New Taxonomy of Telematics, in Proc. GI'95 Graphics Interface Conference, (Québec City, Québec, May 18-21, 1995), ACM Press, pp. 239-246.
2. Buxton, W. Metaphors that Keep us on the Periphery, Human Computer Interaction 9, 1 (1994), 50-52.
3. Dourish, P. and Bly, S. Portholes: Supporting Awareness in a Distributed Work Group, in Proc. CHI'92 Conf. on Human Factors in Computing Systems (Monterey, California, May 3-7, 1992), ACM Press, pp. 541-547.
4 Haythornthwaite, C., Wellman, B., and Mantei, M. Putting Computerized Communications in Perspective: Media Use and Work Relationships in a Research Group. Group Decision and Negotiation, 4, 3 (1995).
5 Hollan, J. and Stornetta, S. Beyond Being There, in Proc. CHI'92 Conf. on Human Factors in Computing Systems, (Monterey, California, May 3-7, 1992), ACM Press, pp. 119-125.
6. Kraut, R. and Egido, C. Patterns of Contact and Communication in Scientific Research Collaboration, in Proc. CSCW'88 Conf. on Computer-Supported Collaborative Work, (Portland, OR, Sept. 26-28, 1988), ACM Press, pp. 1-11.
7. Li, J. and Mantei, M. Working Together Virtually, in Proc. GI'92 Graphics Interface Conf. (Vancouver, British Columbia, May 11-15, 1995), pp. 115-122.
8. Monge, P. R., Rothman, L. W., Eisenberg, E.M., Miller, K. L. and Kirste, K. K. The Dynamics of Organizational Proximity. Management Science 31 (1985), pp. 1129-1141.
9 Nisbett, R.E. and Wilson, T. D. Telling More than We can Know: Verbal Reports on Mental Processes. Psychological Review 84, (1977), pp. 231-259.
10. Root, R. Design of a Multi-Media Vehicle for Social Browsing, in Proc. CSCW'88 Conf. on Computer-Supported Cooperative Work, (Portland, Oregon, Sept. 26-28), ACM Press, pp. 25-38.
11. Scott, J. Social Network Analysis. London, Sage, 1991.
12. Short, J., Williams, E., and Christie, B. The Social Psychology of Telecommunications. London: John Wiley and Sons, 1976.
13. Tang, J. C., Isaacs, E. A. and Rua, M. Supporting Distributed Groups with a Montage of Lightweight Interactions, in Proc. CSCW'94 Conference on Computer-Supported Cooperative Work (Chapel Hill, NC, Oct. 22-26), ACM Press, pp.23-34.
14. Willliams, E. Experimental Comparisons of Face-to-Face and Mediated Communication: A Review. Psychological Bulletin, 84 (1977), 963-976.
University of Toronto
Dept. of Computer Science
10 Kings College Road
CANADA M5S 1A4
Alias | Wavefront Inc.
210 King Street East
Tel: +1-416-362-8558 extn 8435