CHI'94 format description

ABSTRACT

We argue that standard contact lists do not accurately represent the nature of people. s social networks. In the present paper we discuss the linear structure in existing contact representations, the historical basis for that structure, and some of the problems created by that structure. We explored more intuitive visualizations of social contacts by conducting a study in which users drew their social networks on paper. We then applied the results of that user study to rapidly prototype several new representations of contacts. A usability test of the prototype showed that a more intuitive representation of contacts facilitated the completion of contact list tasks.

KEYWORDS: contacts, address book, user study, visualization, rapid prototyping, personal information management, personal networks

INTRODUCTION & BACKGROUND

Computers have been used to store names of people. s contacts for many years. Early mainframe systems supported lists of recently-used email names, and address books were some of the classic uses for personal computers touted in the early days of the Apple ][ [2] and TRS-80 [13]. Most of these early electronic address books were modeled after pen and paper address books, where people organize their contacts by listing them in alphabetical order.

In the past decade, with the advent of the Internet, the computer has been transformed into a tool for social interaction [7]. In response, contact management software has become increasingly complex. Personal information management systems must afford a range of user activities, including a) contact information storage and retrieval, b) the coordination of contact information across multiple computer systems, c) the association contact information with decision-making tasks, and d) the integration of contact information with online communication [11]. Modern contact management software packages have added the ability to send emails [9], make phone calls, see if people were online [6, 1], and even organize contacts into groups [9]. However, despite the greater complexity of modern personal information management systems, the contact representations themselves have maintained a user interface that reflects their simple, alphanumeric, list-based roots [11].

As the nature of people. s online interactions become more complex, how personal contacts are represented in email, chat, online meetings, Internet telephony, and instant messaging becomes increasingly important. Millions of people have to deal with contact management as a primary computing task in the course of their online interactions. The old list-style representation breaks down under these increasing expectations. In email systems alone the sheer volume of communications and the complexity of the responses required prove too much for the simple personal archiving systems currently in place [16].

We propose that one of the primary problems with the standard contact list is that it does not allow the user to accurately represent the nature of their interactions with their social contacts. People. s interactions are not arranged in terms alphabetical lists, but rather in terms of dynamic networks [14]. A person. s network involves all the connections that link him or her to other people, and through them, connections to yet other people. Interactive networks involve all those with whom interactions typically occur (usually 16-26 people) [8]. Connections between people vary in importance, frequency, and context, and occur at both the individual and group level. Our hypothesis is that if people. s electronic contact software better represented the nature of their social networks, they would be better able to manage their interactions.

Our ultimate goal was to develop a user interface that enhanced people. s ability to easily communicate with individuals and groups in their social networks. We expected that the best means for doing so would be through a graphical representation of social networks that allows users to visualize their contact information.

Information visualization is a powerful tool for amplifying cognition, allowing people to grasp complex relationships otherwise obscured [3]. Card et al. [3] suggest developing information visualizations by starting with the raw data, then developing data tables, and finally visual structures. Shaw and Gaines [10] use such a method to represent social networks in a special interest group by starting with similarity data of people. s views on topics.

We briefly explored the possibility of developing a similar representation of people. s interactive networks by using data from the frequency of people. s online interactions. Eick and Wills [5] developed just such a network for a small department using emails. However, we thought it was important to maximize the effectiveness of the graphical interface. s ability to represent the users. social networks, and believed the most direct route towards doing so was by asking the users themselves how they would do so.

In the present paper, we take a fresh look at the representation of contacts through a series of steps. First, we examine some of the specific problems found in current online contact lists in the context of online communication. Second, we explore possible solutions through a study of how people naturally represent their day-to-day contacts. Third, we develop an electronic address book prototype that provides an alternative to the standard list-style. Finally, we compare users. experiences of storing and retrieving contact information in the prototype and in Outlook Express [9].

THE CONTACT LIST

As mentioned earlier, the typical contact management user interface displays a list of names in alphabetical order. Additional information, such as email address and phone number, is listed in conjunction with the contacts name. Figure 1 shows a typical contact list, from Microsoft. s Outlook email and scheduling program [1].

Figure 1. The standard user interface of a contact list [MS Outlook]

In addition to arranging names alphabetically, some programs support arranging names by groups, either as properties of the contacts, or as entities in themselves. Some programs allow users to organize their contacts into a hierarchy of groups.

ASSESSING PROBLEMS WITH THE CONTACT LIST

The contact list format does not integrate well with online communication systems. People generally communicate with others online through the use of email or instant messaging. Email systems rely heavily on the users to provide contact information every time they send an email, and do not readily represent people in terms of groups or social importance. During the summer of 1999, we informally studied a half-dozen different representations of contacts in Instant Messaging software. Some of the problems that we observed with Instant Messaging contact lists were as follows:

· It was difficult to utilize a contact representation for multiple purposes (i.e., a representation that was good for finding emails was bad for showing when the user was online).

· The contact representation did not accurately indicate the relative importance or significance of contacts (i.e., my best friend I call everyday looks the same as my dentist).

· It was difficult to accurately represent complex relationships between individuals and groups (i.e., the software forced each contact into a single group).

In sum, the alphabetical list format did not effectively represent people. s interactive networks. In a previous study [4], we found that some of the problems with the chat user interface were a direct fallout of the modern chat UI being a direct copy of the original chat UI, which had been designed for an ASCII terminal. The alphabetical list UI for contacts had the same ancestry, and we thought it was appropriate to look at contacts from a completely clean slate.

THE MALL STUDY

We wanted to develop a visual representation of people. s social contacts, but did not want to do so without better understanding people. s actual social contacts, and how they thought about them.

To this end, we performed a user study at a local shopping mall. With the mall. s permission, we set up a table in a high-traffic area near an entrance. Study participants were offered a $5 gift certificate good at any store in the mall in exchange for their participation. We had a total of 106 participants, with widely different ages, education levels, and computer experience.

In the study, half of the participants completed a questionnaire designed to assess their interactive social networks. The questionnaire asked participants to list all of the people with whom they had interacted in the past week. We then asked them to indicate the nature of their relationship with each person, and how many hours they spend interacting with that person either in person, over the phone, through email, through chat, or through some other means.

The other half of the study participants completed a questionnaire designed to assess how they mentally represented their social networks. Participants were given a 81/2 x 11. piece of paper with a large rectangle labeled . People Space. . In that space, they were asked to . Draw a picture or map that shows how you are connected with all of the people with whom you have significant interactions.. Participants were not given additional instruction or hints of what structure they were to use.

THE MALL STUDY RESULTS

Interactive social networks online. We wanted to get a sense of the size and nature of people. s online social networks. We found that in general, people had on average 18 people in their entire interactive social networks. They interacted with most of those people in person. They interacted with about five people a week through email or through chat, for approximately five hours a week [Table 1]. 42 out of 56 people used email at least once over the course of the week, and 10 used chat at least once. Over half of the people interacting with others online were students or skilled professionals.

Table 1: People. s interactions with others over the course of a week.

People who had some form of online interaction over the course of the week had on average 6.2 online contacts. The range of online contacts was from 1 to 25. Note, however, that the majority of people had less than 12 online contacts. See Figure 2.

Figure 2. The frequency distribution of the number of people contacted online over the course of a week.

We asked people to indicate the nature of their relationships with their contacts. We coded the contacts as being either a significant other, an immediate family member, a distant family member, a friend, a coworker/fellow student, or other. We found that in general, most of people. s contacts fell into either the friend or fellow worker/student category (80.5%). See Table 2.

Table 2. The nature of the relationship for online contacts varies. Most online contacts are friends or fellow workers/students.

In sum, people who maintained online relationships spent on average 6 hours a week in communication with 6 other people, who were primarily friends and fellow workers/students.

Participant representations of social networks.

We were interested in how people mentally represented their social networks. An examination of over 50 drawings of people. s social space revealed the following:

. Me. is important - The most striking result was that almost every participant (96%) put a large . me. at the center of the page, despite being provided with a blank page and not being instructed to do so. A couple of people put . me. off to one side and made it the root of a tree.

Small # of People - Most participants had between 10 and 20 people represented, not including themselves. Interestingly, 10% of the participants included deceased people and/or God among their contacts. Note that we found there to be more people (18) in participants. interactive networks, when they were asked with whom they had . interacted. as opposed to with whom they had . significant interactions. .

Decorations . Most people decorated their maps with little doodles, sketches, or symbols. In particular, people would place hearts next to people they liked.

Groups Matter . While we were unable to find any common elements in terms of how participants divided their contacts into groups, most (88%) made groups of some sort. Often we found that groups were given as much attention as the contacts themselves. Groups held multiple people, and people were in multiple groups . often in fuzzy ways. For example, one subject drew a . friends. balloon and put people in it, and then drew a blob around some of them and connected it via a line to his wife, labeling the line . wife. s. . This was then extended to include some members of the family group.

Diverse Styles . Perhaps not surprisingly, no one organized their names into a simple list. Instead, we saw a considerable range of organizational schemes, which we placed into four loose categories: . balloon-style. , . web-style. , . list-style. , and . art-style. .

Balloon-style 44% of participants drew large balloons, circles, pyramids, squares, etc. on the page representing relationships, and then placed individual people in these areas. For instance, a participant would have a circle for family, with all of the family members inside. The actual groups were quite diverse . school family, hobbies, teams, work, church, etc. Everyone broke it up differently. For example, whereas most people had different types of friends (i.e., work friends, school friends, spouse. s friends), others would just have . friends.. Balloon-style was by far the most common.

Figure 3: Balloon-style, where people represented groups by circles, squares, etc., and then placed people in the groups.

Web-style 30% of participants drew all of their contacts and then drew connection lines between then. Sometimes the lines represented relationships (i.e., father), and sometimes the lines were an associative link (e.g., a spouse. s friends were linked to a spouse). Most often, the lines were ambiguous, and did not have a particular meaning. Sometimes, the web connections were tangled, but for most part, each person was linked to one . root. person and some small number of . branches.. Some people combined balloons and webs . drawing a large web, then encircling different nodes with balloons.

Figure 4: Web-style, where people represented individual contacts and drew links in between them.

List-style 10% of participants drew out one or more lists or shallow hierarchies. We originally suspected that people with a more technical background or people with many contacts would tend to prefer a list style, but that was not the case.

Figure 5: List-style, where people represented their contacts in lists.

Art-style 16% of participants drew little cartoons, pictures, stick figures, etc., and had basically no abstract organization whatsoever. They would, for example, draw their friends waving to them, or draw themselves on the phone with their parents.

Figure 6: Art-style, where people drew their contacts but did not arrange them in any abstract order.

CONTACT LIST REQUIREMENTS

Based on the results of our study, we decided that a new design was in order for the contact list user interface. Our goals were as follows:

· Make it easy to leverage a wide-range of activities off of the contact representation: email, online status, etc.

· Make it easy to represent both groups and individual contacts, preferably at the same level.

· Give the user a lot of control over the appearance and arrangement of contacts, with no set notion of a . correct way.. For instance, don. t require an email address, since dead people don. t have them.

· Keep the representation a bit soft, so users do not feel . locked in. to a particular representation or categorization.

· Support list-style, web-style and balloon-style representations.

· Aim for the . sweet-spot. of 10-20 contacts.

DUST Rapid Prototyping System

We did our prototyping using a tool called DUST[4]. DUST was developed in the Virtual Worlds Group at Microsoft Research. It is a rapid prototyping tool that combines an object-oriented graphics system with a dynamic object system. DUST has the following features:

Container-based graphic & event system . any object can contain any number of other objects. Objects are clipped to their container. Events such as mouseEnter, mouseLeave, and click are passed up the containment hierarchy.

Transforms . every object has an inner and outer transformation matrix, so it can control its own scale and rotation, as well as the scale, translation and rotation of objects inside of it. This allows DUST to be used to create user interfaces that pan, zoom, and/or distort.

COM-based scripting . DUST is implemented in COM, and can be programmed using any COM-compliant language, from C++ to VBScript.

Advanced Graphics Capabilities . DUST takes advantage of new graphical capabilities that are provided the latest version of the Windows operating systems, including gradient fills, translucency and anti-aliasing.

Dynamic Objects & Methods . DUST is a fully dynamic object system. It is prototype-based, instead of class-based, and so any object can inherit from any other object. Objects can be created at runtime; methods and properties can be added or removed, and even the inheritance hierarchy can be modified.

We used DUST because we felt it gave the widest range of novel interface capabilities. Other tools we evaluated either had a predetermined user interface (like Visual Basic), or were too hard to integrate with existing compiled code (such as Director or Flash).

We are investigating the possibility of releasing the DUST source code as a sample application for the Virtual World platform software. For the latest information on our source code releases, please check the following URL:

BLOB UI PROTOTYPE

Based on our results and observations, we developed a user interface called the . blob UI. . The first version of this UI is shown below in figure 7.

Figure 7: The Blob UI allows people to represent groups and individuals with blobs, and relationships among them through distribution in the work area.

The user interface consists of a tray of buttons on the left side, and a large blank work area on the right. The user adds groups and individuals by dragging and dropping them from the tool tray into the work area.

In the work area, the large blue blobs represent groups, and the smaller red blobs represent individuals. The blobs are opaque in the center and become increasingly transparent near their edges.

Blobs can be moved around the work area and positioned in whatever way the user desires. Users can adjust the size of a group by pulling on any edge. People can be added and removed to groups by dragging and dropping them either into or out of the group boundary. The user can also control the scale of the work area, zooming in and out.

Normally, the blobs do not have names drawn on them. When the mouse is moved into a group, the group draws its name, becomes 50% more opaque, and also draws the names of its members. (figure 7)

Figure8: The . Friends. group contains Bob, Sally, and Joe.

When the mouse is moved over an individual, the individual draws its name, becomes 50% more opaque, and draws the names of its groups (figure 8).

Figure 9: Joe is a member of both Friends and Co-Coworkers.

The Blob UI satisfied most of our original goals:

· The ability to zoom in and out allowed different tasks in the same user interface. The display could be minimized into a tiny corner to show which contacts are online, or could be blown up to full screen for more active management tasks, such as entering address information.

· Groups and individuals are given the same prominence, despite having an immediately apparent visual difference (color and shape).

· The interface imposes no set structure on the user. They have freedom to place contacts close to or far from other groups and contacts. No meaning is ascribed to the arrangement, other than the user. s own meaning.

· The blended edges and translucency reinforce the feeling of the representation being . soft. . A user can, for example, move a casual friend closer to the fuzzy edge of a . friends. group.

· Our representation is primarily balloon-like, but because we do not require any particular arrangement, users can have representations that are more list-like or web-like.

· Interface is optimized for 15 to 25 people, but can scale larger or smaller.

USER STUDY

As soon as the prototype. s graphical interface was working, we performed a user study. Since we were working with a radically different user interface, we wanted to get feedback from real users as early in the process as possible, before we went too far down the wrong path. We had study participants use both the Blob UI and Outlook Express so that we could compare both their subjective experiences of the two programs, and assess their objective performance. We expected that in storing contact information people would find the Blob UI intuitively easy to use and understand. We also expected that in retrieving contact information people would be able to retrieve information from Blob. s graphical representation of contacts more easily than from the Outlook Express list-style representation of contacts.

7 people participated in our study - 3 males and 4 females. All participants had some previous computer experience, but they ranged in their use of address book software from not at all to many times a day. The participants were videotaped during the course of the study, and all computer actions were recorded.

Participants began by filling out a simple questionnaire that asked about demographic information and current address book usage. Next, the participants were instructed to write the names of 15 contacts on small pieces of index card. Once finished, they were instructed to arrange the cards onto a large piece of paper as if they were . trying to show someone how you categorize those people in your mind.. Participants were told they could draw anything on the paper that would help them communicate their categorizations. We had them arrange their contacts prior to using any software so that they could establish their own view of their contacts without influence from the representation in a particular program.

The participants were then instructed to recreate their contacts in both the Blob UI prototype and a standard email client . in this case, Microsoft. s Outlook Express . as accurately as possible. After each program, they answered a questionnaire about their experience using 7-point Likert scales.

After entering their own contact information, participants were shown the same two programs (Blob and Outlook Express), this time with pre-existing contact information sets loaded into them. They were instructed to answer questions about the organization of contacts therein.

STUDY RESULTS

Overall, people preferred Outlook express by a small margin . 53% vs. 47%. This result was affected by both sex and current address book usage.

On a 7-point Likert scale, males preferred Outlook Express, (5.3 vs. 6.4) whereas females preferred the Blob UI (5.0 to 4.5). When asked if they had a preference between a graphical representation versus a text-list, males indicated no preference (4.0) whereas females preferred the graphics (4.5). Given a choice between the two programs as is, males would choose the Blob UI 43% of the time, whereas females would chose it 60% of the time.

People. s preferences were largely affected by their pre-existing experience with address books. Those who already used a computer program for an address book preferred Outlook (5.9 vs. 4.6), whereas those who only used a paper address book preferred the Blob UI (5.9 vs. 4.4). Given a choice between the two programs, those with only paper address books chose the Blob UI 66% of the time.

When asked to explain their preference, those who preferred Outlook Express cited its familiarity and that it had more options. Those who preferred the Blob UI felt it was easier to use, and that it provided a more accurate representation of their contacts.

In addition to examining participant preferences, we wanted to assess whether people found it easier to retrieve information from the Blob UI than from Outlook Express. We expected that the visual representation of contacts provided by the Blob UI would help people understand relationships between individuals and groups. We had people answer a series of questions about the number of groups in a list, the number of individuals in particular groups, and so forth. For this test we generated our own contact information so that the information would be equivalent across the two programs. We found, as expected, that people were able to retrieve information more easily from the Blob UI than from Outlook Express. All participants completed the tasks faster with the Blob UI than with Outlook Express (2:53 minutes vs. 4:53 minutes). Participants were also more accurate using the Blob UI than in using the Outlook UI (82.5% vs. 75%).

STUDY CONCLUSIONS AND DISCUSSION

Although there was an overall preference for Outlook Express, females and new users appeared to prefer the Blob UI. In addition, people were able to retrieve contact information more easily using the Blob UI. Although the differences were not extreme and the test was small, we were encouraged by these moderate results. We feel that this alternative representation of contacts is worth pursuing further.

We have noticed in our previous tests of novel user interfaces is that it is difficult for users to acclimate to novelty in a short period of time. People tend to prefer programs that work the way they have always worked. Similar to biological evolution, a new user interface must offer immediate significant advantage before something that works ok is abandoned. We found a similar effect in the present study, with experienced Outlook users preferring Outlook, and novice users preferring the Blob UI.

Another problem that we noticed was that participants had a negative reaction to the roughness of our prototype. The lack of standard keyboard shortcuts, window management controls, and support for Undo gave the program an unfavorable perception in the participants. minds. This is just a fact of life for those of us working with prototypes. It is hard to know where the optimal point is between getting something polished and getting something in front of test subjects.

FUTURE UI DIRECTIONS

Our first concern is to make the Blob UI prototype into more of a . real. application, with standard window management commands and features such as . undo. .

We would like to extend our Blob UI to allow users to draw fuzzy lines in between the blobs. This would enable users to have better list and map views. We would also like to enable users to add clip art and other images to people, groups, and lines.

We want to integrate standard contact functionality into the user interface. For example, we want to enable users to send and view email directly from a person. s blob, and we want the blobs to change color to show on-line status.

Once these changes have been made, we would like to run some more longitudinal tests, to see how subjects like the interface over a longer period of use.

SUMMARY

We explored some of the problems with existing representations of contacts, which tend to be based on an alphanumeric arrangement of contact names. Through a study of the nature of people. s social networks, we were able to construct a new user interface for contact management. A user study of the new interface showed that a visual representation of contacts based on how people think about their social networks might hold some promise. Finally, we discussed some requirements for contact representation software in the future.

We feel that in general, visual representations of contacts have been an under-explored in the field. More research is needed examining the impact of contact visualization on a) people. s ability to store and retrieve contact information, and b) the use contact information in the course of people. s online interactions.

ACKNOWLEDGMENTS

Thanks to Debbie McGhee for her help in running the user studies and analyzing the data.

REFERENCES

3. Card, S. , MacKinlay, J., and Schneiderman, B. Readings in Information Visualization: Using Vision to Think. Morgan Kaufmann Publishers, CA. 1999.

4. DUST, computer program, Virtual Worlds Group, Microsoft Research, Microsoft Corp. Jan 2000

5. Eick, S., and Wills, G. Navigating Large Networks with Hierarchies. In Card, S. , MacKinlay, J., and Schneiderman, B. (eds.) Readings in Information Visualization: Using Vision to Think. Morgan Kaufmann Publishers, CA. 1999.

7. Kiesler, S. Culture of the Internet. Lawrence Erlbaum Associates, New Jersey, 1997.

8. Milardo, R. M. (1992). Comparative methods fore delineating social networks. Journal of Social and Personality Relationships, 9, 447-461.

10. Shaw, M., and Gaines, B. Supporting Personal Networking through Computer Networking. Proceedings for Conference on Human Factors and Computing Systems, New Orleans, USA, 1991.

11. Thomas, P.J., and Meech, J. F. Person information management: applying HCI techniques to develop usable personal technology. Proceedings of HCI. 94, Glasow, 1994.

12. Thomas, P. J., Meech, J. F., and Macredie, R. D. A Framework for the Development of Information Appliances. ; Proceedings of the 1995 ACM Symposium on Applied Computing , Nashville, 1995.

14. Van der Poel, M. Personal Networks. Swets & Zeitlinger, PA. 1993.

15. Vronay, D., M. Smith., and S. Drucker, . Alternative Interfaces for Chat. , User Interface Software and Technology (UIST), Nov 1999, 19-26

16. Whittaker, S., and Sidner, C. Email overload: Exploring Personal Information Management of Email. Conference proceedings on Human Factors in Computing Systems, Vancouver, 1996.