Using Rhythms of Relationships to Understand Email Archives

Adam Perer,1,2 Ben Shneiderman,1,2 and Douglas W. Oard1,3

Human-Computer Interaction Lab, Institute for Advanced Computer Studies(1), Department of Computer Science(2) or College of Information Studies(3), University of Maryland, College Park, MD 20742 adamp@cs.umd.edu, ben@cs.umd.edu, oard@umd.edu

Abstract. Due to email’s ubiquitous nature, millions of users are intimate with the technology. However, most users are only familiar with managing their own email, which is an inherently different task than exploring an email archive. Historians and social scientists believe that email archives are important artifacts for understanding the individuals and communities they represent. In order to understand the conversations evidenced in an archive, context is needed. In this paper, we present a new way to gain this necessary context: analyzing the temporal rhythms of social relationships. We provide methods for constructing meaningful rhythms from the email headers by identifying relationships and interpreting their attributes. With these visualization techniques, email archive explorers can uncover insights that may have been otherwise hidden in the archive. We apply our methods to an individual’s fifteen-year email archive, which consists of about 45,000 messages and over 4,000 relationships.

Introduction

Since 1971, email has grown rapidly in popularity and has become a central part of many users’ personal and professional lives. Despite its impressive role in society, there are still few tools available to explore archives of email. The need for such tools will grow as valuable email archives increase in availability. The U.S. National Archives preserves emails as government records (Baron, 1999), a recently released collection of Enron emails has attracted significant public attention (Grieve, 2003), and some individuals have now accumulated email collections that span decades. Historians and social scientists will undoubtedly find these archives to be a valuable basis for understanding the individuals and organizations that created them. However, it is currently far from clear how these explorers will gain the context they need to understand the archive’s numerous conversations.

Figure 1 illustrates one way in which the universe of tools for interacting with online conversations can be subdivided. Email is created by individuals, and often in some organizational or social context. There has been a great deal of work on individual and organizational email productivity tools (regions A and B), and on the management and analysis of conversations in public email venues such as mailing lists and Usenet News (regions C and F). Our work in this paper focuses on region D, as we present new techniques for exploring the archived email of an individual.

Individual

Organizational

Social

Current

Region A:

Managing an individual user’s current inbox

Region B:

Managing current email within an organization

Region C:

Managing current conversations in a public space

Archived

Region D:

Exploring an archive of an individual’s messages

Region E:

Exploring an archive of an organization’s messages

Region F:

Exploring an archive of a public space.

Figure 1. Types of interactions with email collections.

Although the principal content of email is free text, when attempting to browse archives, the shortcomings of a text-only display become clear. Email archive explorers have previously tackled the archives by keyword searching, but this approach will often result in losing a conversation’s context (Donath, 2004). Visualizations are one way to provide this missing context. In this paper, we show that valuable information can be uncovered by visualizing the temporal rhythms of social relationships that are evidenced in email archives. Each relationship that is evidenced in an email archive has a rhythm that can be characterized by the intensity of the correspondence over time. Relationships that are brief but intense have rhythms with sharp growth and steep decline. Relationships that are durable and strong have consistent and continuing rhythms. This paper presents insights achieved by analyzing the rhythms, which help archive explorers question why certain relationships start and stop, why certain relationships share similar activity patterns, and the nature of the relationships that yield different interaction patterns.

Detecting long-term rhythms, our focus in this paper, requires a collection spanning many years. Ben Shneiderman, a co-author of this paper and a pioneer in the fields of Human-Computer Interaction (HCI) and Information Visualization, has archived the emails he produced and received since 1984. The archive portrays over 4,000 of Shneiderman’s relationships, totaling around 45,000 messages. That archive spans a longer period than any other collection that was available to us when we started this work, offering us a unique opportunity to study the long-term rhythms of relationships present in a real email collection. In the next section, we review related work on interacting with online conversations. Next, we define what we mean by “relationships” and the “rhythms” that they produce. We then present our analysis methods and illustrate the use of those methods on the Shneiderman archive. Finally, we conclude with some suggestions for future work.

Previous Work

In this section we briefly review prior work on email management, organizing the discussion using the task decomposition shown in Figure 1. Interaction with the user’s own current email (Region A) is by far the most actively studied email management task in the research literature. An early ethnographic study by Mackay in 1988 provided compelling evidence that different people deal with large quantities of their personal email in many different ways (MacKay, 1988). Whittaker and Sidner’s later study resulted in the same conclusion, while also describing tasks that individuals use email for beyond the asynchronous communication for which it was designed (Whittaker and Sidner, 1996). Recent attempts to integrate visualizations into email clients seek to help users better manage their email. For example, enabling users to see the thread structure provides them with a better understanding of the how conversations evolve over time (Kerr, 2003; Venolia and Neustaedter, 2003). Another example is the Remail project, which provides a “correspondents’ map” that allows users to quickly see who they haven’t replied to recently, as well as a “message map” to see messages with similar attributes (Rohall et al., 2003).

Some recent projects have investigated exploration of personal email archives to uncover trends and patterns (Region D). PostHistory explored email archives that extend as far back as five years, seeking to support the development of insights that would be socially relevant to the owner of the email (Viegas et al., 2004). PostHistory featured an interface that animates over time to allow users to get a sense for their steady and intense relationships, and to illustrate fast-paced rhythms (e.g., resulting from project deadlines) and slower-paced rhythms (e.g., during vacations). Social Network Fragments, by contrast, focused on revealing groups of correspondents that emerge through email exchanges (Viegas et al., 2004). This interface also used time as a dimension to see how connections among correspondents appear and dissolve, thereby providing a way for the user to visualize the evolution of their own social network. In small studies, users were able to see meaningful patterns with both PostHistory and Social Network Fragments, sometimes using the visualization as instigation for telling stories.

The ubiquity and persistence of email has important consequences for the management of information within organizations (Region B). Ducheneaut and Bellotti studied the use of email in three organizations, and discovered that patterns of email use vary with individual roles within those organizations (Ducheneaut and Bellotti, 2001). They also noted that characteristics of each organization influenced the ways in which people used and organized their email collections. Tyler and Tang added to the understanding of email use within organizations, observing that responsiveness patterns vary in ways that reflect the dynamics of interpersonal relationships within an organization (Tyler and Tang, 2003). That observation led them to suggest that tools for estimating expected response latency could help users detect communication breakdowns. Another example of an organization tool is the “Email Mining Toolkit,” developed by Li et al. to support anomaly detection by creating behavior models. They then used these models to detect aberrant behavior of individuals or groups that may indicate abuse or policy violations (Li et al., 2004).

Exploration of archived collections of organizational email has also been studied (Region E). Tyler et al. used the social network analysis concept of “betweenness centrality” to identify communities in a large collection of email from a single organization, discovering that evidence of the management hierarchy for that organization could be found in the structure of the resulting graph (Tyler et al., 2003). Leuski’s “eArchivarius” system combined clustering based on content or co-addressing with activity timelines and biographies to explore activities in the U.S. National Security Council during the Reagan era using a small collection of declassified email messages (Leuski, 2003).

Usenet News, a distributed management system for a large collection of public mailing lists, has been archived since 1981. Mailing list usage differs somewhat from the use of personal email, both because privacy expectations are reduced and because the group-oriented communication structure alters interaction patterns. Smith used the “NetScan” system to study social accounting metrics for Usenet participation (Region F) and reported statistics on authorship and on activity over time (Smith, 1999; Smith, 2002). Usenet News is immediately available to both participants and nonparticipants (“lurkers”), which makes the distinction between management and exploration somewhat less defined than it is in the case of individual and organizational email. Users of the NetScan system can, for example, use it to find intense discussions and related “newsgroups” (Region C). Sack’s “Conversation Map” also explored Region C, focusing on the structure of long-term conversations by using social network diagrams, lists of discussion themes, and semantic network representations to support visualization of conversational structure and content (Sack, 2000).

The work described in this section is, of course, only a small sample of the extensive research on email utilization that has been reported since the first email was sent over the ARPANET in 1971. Looking broadly at that body of work, however, two trends emerge. First, the vast majority of the reported research has focused on managing current activities rather than on understanding what happened in the past. There has been much less work done in Regions D and E. That makes sense, since only recently has email’s ubiquity become clear and archives of email are accruing. Second, the retrospective analyses on individual email (as opposed to mailing lists or Usenet News) that have been done have had limited scope; we are aware of only one study that has looked at even five years of email. In this paper, we take a longer view, looking back at a fifteen-year period that spans 1984-1998, as Internet email moved from adolescence to adulthood.

Relationships in Email Archives

In this section, we describe the email collection that we worked with and the analytical framework that we applied to explore the long-term rhythm of relationships in that collection.

The Shneiderman Archive

This archive begins in 1984; one year after Ben Shneiderman received tenure as an Associate Professor and founded the Human-Computer Interaction Lab at the University of Maryland. We chose to limit our study to the first fifteen years, culminating in 1998, because Shneiderman changed his email file structure significantly in 1999. The resulting set includes 44,971 messages. That is certainly not every email received or sent by Shneiderman during that period. Rather, it includes those that Shneiderman purposefully stored. Although analysis of the results of intentional retention will not provide a complete picture of an individual’s email traffic, it does serve to filter out spam and other less significant messages. The saved email gives historians a picture of what Shneiderman felt at the time were the significant conversations in his professional life. However, our analysis will miss some subtle and friendly exchanges, which could also serve as sources of interesting rhythms (e.g., as described by (Tyler and Tang, 2003)).

Relationships

Email provides a medium in which users may foster relationships with individuals, organizations, and a global community. Relationships are fundamental to any form of human interaction, so we have chosen to aggregate this collection by relationship rather than the more commonly studied granularities of “threads” (i.e., reply chains) or individual messages. Aggregation into relationships facilitates exploration by masking some sources of variation (e.g., multiple email addresses for a single individual or individuals that participate in multiple relationships) that might otherwise conceal the broad themes that we wish to uncover. By “relationships” we mean a set of conversations over time that reflects a type of interaction that was meaningful to the person that created the email archive. Examples could include conversations with a specific colleague, discussion of a particular topic (e.g., academic governance) involving several members of an organization, or a group of messages regarding the planning of an event (e.g., a professional conference).

The process of discovering unique identities in an email archive is not trivial, especially when dealing with an archive than spans fifteen years. People move to various organizations and universities, obtain new email addresses, change their surnames, and evolve their academic interests. For this reason, individuals are not classified simply based on their email header information. Instead, each relationship is identified with help from Shneiderman’s filing metadata, as he typically stored relationships in separate folders. Conversations with individuals are usually stored in a folder labeled with their name. If conversations occurred with many participants on a particular topic, such as organizing a conference, these are usually stored in a folder labeled with a description of the topic.

We were interested in applying our techniques to learn about Shneiderman’s professional life, and not his personal life. In the archive, there were several relationships present that did not include any content related to his professional career. These relationships include his family, and friends from outside his professional circle. Only about 20 of the 4,051 relationships in his archive fell under this category, resulting in a small number of deletions. Those relationships were manually tagged and deleted before any analysis was performed.

In order to take advantage of the manually tagged relationships, there was a significant amount of work necessary to ensure the data’s representation was valid. Occasional misspellings were present, surname ambiguities occurred over time (e.g., folders named ‘norman’ in early years versus folders named ‘normandon’ and ‘normankent’ in later years), and an occasional misstep from naming conventions (storing a message from Catherine Plaisant in a folder named ‘catherine’ instead of ‘plaisant’). These findings are consistent with Ducheaneut and Bellotti, who remark about users’ confusion as to whether store a message from a corporate colleague in a folder named after the company or the person (Ducheaneut and Bellotti, 2001). These inconsistencies were corrected by fixing typographical errors and standardizing the naming convention for relationships that contained conversations with similar email addresses.

Before our analysis, Shneiderman categorized each relationship into one of three distinct groups. A relationship could be tagged as a person, which meant the messages in that folder all revolved around the relationship of a single person. A relationship could also be tagged as an organization, which meant the messages contained within that folder revolved around a variety of individuals all communicating about or within the same organization. Finally, the relationship could be tagged as a topic, which meant a variety of people from one or more organizations all communicating about a similar topic. Of the 4,051 relationships, almost 95% were tagged as people (3,836), compared to only 197 organization relationships and 18 topics.

We should note that our human-assisted categorization methods are not a strict requirement for exploring archives. For example, relationships could be postulated automatically based on email addresses and/or message content. However, the availability of Shneiderman’s personal categorization scheme gave us comfort that we would be analyzing an accurate representation of the corpus, reducing the noise present in our rhythms.

Rhythms of Relationships

Figure 2. Examples of rhythms of relationships.

By the “rhythm of a relationship” we mean the pattern of activity for a relationship over the duration of an email archive. For example, in Figure 2, two relationship rhythms are shown. The left rhythm depicts a relationship that was inactive during the early years, becomes active in the middle, and then grew to be an intense relationship in the later years. Conversely, the rhythm on the right shows a relationship that starts out intensely and then eventually dies down into sporadic contact. These types of rhythms can be extracted from information that is present in email headers alone, thereby minimizing the need for access to text in the bodies of the email that would naturally be more problematic from a privacy perspective. Due to our interest in understanding long-term patterns, we construct rhythms that have a granularity of a year.

Profiles of Shneiderman’s Most Active Relationships

Clearly not all relationships are made equal; certain relationships are very intense whereas others are quiet and infrequent. In fact, about a third (31%) of relationships in the Shneiderman archive have less than two messages and 55% have less than four messages. Only 11% of the relationships present in the email archive ever reach 20 or more messages.

Examining the key relationships in an email archive provides an understanding of the nature of the owner’s work. Since the Shneiderman archive consists of only 3,836 individual relationships, it is likely that the contents are tied to only the most valued relationships. To gain an understanding of the most frequent correspondents, we extracted the relationships with 100 or more saved messages, leaving only 76 professional relationships.

These 76 professional relationships were only 2% of the 3,836 professional relationships, but they produced 12,771 saved messages (31%) out of the 41,420 saved messages. The power distribution of relationships is seen in Figure 3. We expect this distribution to be common in email archives of individuals, with a bulk of the messages tied to a small number of key relationships.

Having contact with the archive’s owner is not a luxury we expect most historians and social scientists to have. However, we exploit our contact with Shneiderman to attain accounts of who these 76 most active relationships were. This knowledge is useful, as we can judge our techniques against these verifiable truths. The information provided by Shneiderman is described below, as it provides insight into the types of intense relationships that emerge in a fifteen-year email archive.

The top ten most active professional relationships had between 240 and 634 total messages. These relationships included four key colleagues at the University of Maryland (Plaisant, Marchionini, Norman, Chimera), conference organizing partners (Light, Soloway, Rotenberg), and collaborators on other projects (Simons, Ahlberg, Grudin). These reflect Shneiderman’s major projects; some with a small number of intense years of activity with over 140 saved messages (Ahlberg, Simons, Light, Rotenberg), while the rest show a more steady pace of exchanges.

Figure 3. Power distribution of relationships.

Number of Messages

Relationships

These 76 most active relationships were relatively easy for Shneiderman to assign to categories. On a large table, he created a small card for each relationship and sorted them into clusters. About a dozen of the names had more than one role, such as when a University of Maryland colleague moved to another university, a former student became a corporate partner, or a book editorial worker was also a colleague at another university. Assignment was by major role, as determined by the majority of saved messages rather than duration.

As expected, many of the most active professional relationships are from the University of Maryland, with 11 being close colleagues, 9 being students, and 11 others being superiors (chairs, deans) and staff (secretaries, administrators). Colleagues at other universities accounted for 17 of the most active professional relationships, while conference organizing partners and related efforts covered 10 relationships. Corporate partners including financial supporters, consultancies, and book or lecture collaborators covered 9 relationships.

Other important relationships included 4 colleagues tied to the USACM Public Policy group, in which Shneiderman was a member of the Executive Committee. Development of Shneiderman’s book, Designing the User Interface: Strategies for Effective Human-Computer Interaction (Addison-Wesley Publishers), showed strong activity for 3 people in the years when the first edition (1986), second edition (1991), and third edition (1997) were in production. Finally, close collaboration with 2 government partners at the National Library of Medicine and the Library of Congress generated high levels of activity for several years.

Most Active Professional Relationships

more than 100 saved messages (n = 76)

Number

Avg. Years Active

Avg. Total Messages

UMD- Close colleagues

11

9.2

209.7

UMD- Superiors and staff

11

9.6

123.0

UMD- Students

9

9.0

183.8

Colleagues at other universities

17

11.3

152.4

Conference partners

10

8.3

172.7

Corporate partners

9

9.1

137.6

USACM Public Policy

4

5.5

252.3

Book editorial workers

3

8.7

183.0

Government partners

2

9.5

171.5

Figure 4. Shneiderman’s most active relationships, categorized by role.

Methods for Understanding Email Archives

In this section, we identify certain tasks that lead to insights by analyzing the rhythm of relationships in email archives. For each task, we describe the visualization methods that lead to the insights and the set of features on which that visualization is based. We illustrate the utility of these analysis methods with examples from the Shneiderman archive.

Evolution of Relationships

With a corpus that spans 15 years, it is to be expected that the nature of some relationships will change over that period. By examining relationships individually, it is possible to witness certain relationships blossom, while other relationships conclude. However, when looking at all the relationships together, one might wonder what sorts of collective patterns emerge: Did the frequency of archived emails change as email became more ubiquitous? Are there specific periods in time when the social circle changed more rapidly than others? Questions of this type can be answered with the following approach.

Figure 5. Growth rates for messages and relationships.

One of the simplest analyses that can be done is to count the number of messages over time. Figure 5 illustrates the rapid growth in the number of archived messages over time, increasing from 98 emails in 1984 to 8,499 in 1998. Figure 5 also shows the number of active relationships, counted for each year over the same period. The growth in the number of active relationships is well fit by linear interpolation, while the growth in the total number of messages is well fit by a quadratic function. This archive spans a period in which the number of ARPANET/Internet users grew exponentially, and in that context, the more sedate linear growth in the number of relationships is interesting.

By counting the number of messages and active relationships over time, explorers can get a sense of how an email archive evolves. Interesting characteristics can be determined, such as if the individual fosters more relationships over time and if the growth is consistent with the growth of the Internet. The limitations to this approach are that these averages mask considerable individual variation, witnessed in Figure 6, which provides a superimposed image of over 4,000 relationship rhythms from the archive. Figure 6 also illustrates a somewhat surprising (and presently unexplained) absence of brief-but-very-intense relationships during the middle years of the archive.

Figure 6. Over 4,000 relationship rhythms superimposed.

Relationship Rhythm Patterns

Useful insights about relationships can be discovered based on the pattern of its rhythm. For example, if a historian was looking for evidence of relationships that were strongly related to a temporal event, a search tool that could find relationships that peaked around the time of the event might be useful. One way to support this is by allowing the user to sketch a graph to query the time-series, a technique introduced in (Wattenberg, 2001).

Figure 7 illustrates an example of this type of search on the Shneiderman Archive using the “Hierarchical Clustering Explorer” (HCE) (Seo and Shneiderman, 2002). Suppose the searcher postulated that Shneiderman’s activities related to policy issues grew markedly in the mid-1990’s. If they had an interest in exploring relationships that were unique to that period, they might then construct a query (represented in Figure 7 by a bold line), seeking relationships that sharply grew in 1994, peaked in 1995, and declined in 1996. Rhythms that match this query are shown as thinner lines. The gray background provides a contour based on most active relationships in the corpus for each year. This technique allows explorers to quickly find relationships that follow expected patterns. Of course, there are also situations in which a searcher may not have a specific question in mind when they begin exploring an archive. In this case, providing the searcher with clusters of similar rhythms might offer a point of departure for further investigation.

Figure 7. Searching an email archive with a rhythm query.

K-means Clustering

Clustering based on similarity can be a useful way of revealing characteristic rhythms. Figure 8 shows the result of clustering the 76 most active relationships (i.e., those with the largest total number of messages) in the Shneiderman Archive into 9 clusters. We applied k-means clustering (MacQueen, 1967) to the 15-year rhythms of these active relationships. The number of clusters, k, is a parameter of the algorithm. The k-means algorithm then divides the 76 rhythms into k clusters until the total distance between the rhythms and their cluster’s centroid is minimized.

Choosing an appropriate k is a difficult choice, especially for an searcher unfamiliar with the overall structure of the rhythms or archive. In our initial run, we asked the archive’s owner, Shneiderman, to group every relationship with more than 100 messages into distinct groups. By printing out the names on cards, and sorting the 76 relationships manually, he came up with the 9 distinct groups listed earlier in Figure 4. It is important to note that these categories were not chosen based on rhythm patterns. Rather, groups were chosen based on the roles of the people (e.g. academic colleague, corporate collaborator or graduate student). There was no evidence that each of these roles should constitute their own rhythm clusters, but it provided an interesting value of k to start with.

Figure 8. Nine groups found using k-means time series clustering on the 76 most active relationships.

The k-means clustering algorithm provides meaningful results, as it successfully displays similar patterns, such as those that accelerate in the later years (Cluster 2), relationships that start strong and then die down (Cluster 3), and relationships that peak in similar years (Cluster 4). However, this algorithm classifies most of the relationships into the first cluster, providing little useful information on that set. Selection of a different number of clusters might yield more insight in those cases, but in general users often find a priori selection of the number of desired clusters to be problematic. Also, the clusters found had no noticeable correlation with the clusters identified by Shneiderman in Figure 4.

Hierarchical Clustering

Hierarchical clustering is another algorithm that can group similar rhythms, but does not require a predetermined number of clusters. Hierarchical clustering works by finding the pair of relationships with the most similar rhythms. It then iteratively builds a hierarchy by pairing these relationships with each other, or with a existing cluster of similar relationships. Figure 9 shows results of hierarchical clustering using HCE on all 4,051 relationships. The hierarchy that HCE builds is shown using a dendrogram, displayed in the top panel of the figure. Each subtree of the dendrogram, alternating in gray and black, represents the cluster of relationships that were most intense in each of the 15 years. These subtrees are not arranged in chronological order, but instead retain their order from the constructed dendrogram. These subtrees lead down to the leaves, where each relationship is represented as a column of tiles. Each tile in the column is shaded to correspond to that relationship’s intensity in a given year. In this figure, gray shading means a strong intensity.

The subtree surrounded by a black box at the top, labeled ‘1988’ and in the middle of the dendrogram, represents those relationships that were most intense in 1988. Notice how the tiles below this subtree have an obvious gray line in the fifth row of the columns (we annotate this row with a white arrow for clarity). That row represents 1988 and the shading conveys the large number of messages. The rhythm profiles that correspond to the selected subtree are shown in the bottom panel, where the intense activity in 1988 among these relationships is confirmed.

Hierarchical clustering also detects groups of relationships that are similar beyond one year. Subtrees of the dendrogram isolate relationships that have peaks in multiple years. For example, the algorithm constructs a subtree for those relationships that have modest intensity in 1996, grow a great deal in 1997 and then grow a little more in 1998. Looking at this cluster’s list of relationships, the

Figure 9. Hierarchical clustering results on all 4,051 relationships.

four most intense relationships involving Ben’s interest in policy are found (Gelman, Brownstein, Ellis, and Simons). This provides evidence that clusters can convey meaning, as the four relationships, remarkably, can be identified when using HCE to zoom in on the subtree (as shown in Figure 10, a view which shows only 2% of the entire tree structure).

However, a weakness of this approach is that not all of these clusters have meaning. For example, the algorithm finds three relationships that have peaks in the disparate years of 1988 and 1994. After exploring deeper into the email content, it appears that is about all these relationships have in common.

Figure 10. A zoomed-in view of the dendrogram. The four relationships related to Shneiderman’s interest in policy are denoted with triangles at the bottom of the graphic. One of these relationships (Ellis) is highlighted.

Aggregating Related Rhythms

In addition to looking at the pattern of individual relationships, it is also a useful exercise to visualize rhythms of related aggregate relationships to see trends based on other attributes, such as organization and location. For this corpus, we generate the aggregates from information contained within the email headers. For each relationship, the most frequent email address will represent that relationship’s attributes. Of course, when dealing with an individual’s email archive, all of the addresses used by the owner should be disregarded. For each relationship, we extract organization names (IBM from user@ibm.com), organization type (educational from user@umd.edu versus commercial from user@spotfire.com) and country codes if present (Israel from user@technion.ac.il). With this extracted information, we illustrate some of the types of analysis that can be performed.

Although the number of active relationships increases over time, it became clear that many of Shneiderman’s emails were still dedicated to relationships within his organization. Over the fifteen-year period, 24% of all of his emails were in communication with relationships at his own university, the University of Maryland. This percentage is comparable to the total fraction of messages in relationships with colleagues at other academics institutions (25%) and all corporations (23%), and double the number of messages beyond the U.S. borders (12%). Figure 11 shows a plot of the number of messages with each type of organization over the fifteen year time period.

Figure 11 also shows how the contact base of international contacts grew over the fifteen year time period. As Shneiderman’s total number of messages grew, so did his correspondence with international contacts. Segmenting the data by country allows us to easily find the most popular international relationships. The top five countries are the United Kingdom (84 relationships), Canada (63), Germany (39), Israel (35) and Japan (31).

Grouping relationships by country allows explorers to notice trends present in Shneiderman’s international rhythms. Countries such as Germany, Canada, Japan and the United Kingdom have stable rhythms throughout most of the archive. However, there are countries like Australia, France and Italy that only grow towards the end of the archive. Other distinct profiles, like that of Austria and Finland, peak in intensity towards the middle of the archive and then fade as time goes on.

This approach allows explorers to find patterns and trends based on relationships sharing similar attributes. However, the email address might not be an accurate representation of the relationship, thereby skewing the rhythms. Furthermore, individuals may change their organization and location over time, but our method will only assign the relationship its most frequent attributes over the duration of the archive.

Figure 11. Aggregate Rhythms generated from Domain Names.

Collaboration Rhythms

One important feature of email is its ease of distributing messages to more than one person simultaneously. This is a typical activity when collaborating with colleagues and these collaborations are evidenced by email headers addressed to multiple people. To gain insights, we construct collaboration rhythms: rhythms characterized by the intensity of correspondence between two individuals, besides the archive owner, over time. Collaboration rhythms can be constructed by calculating the number of times two unique people are a part of the same conversation over the duration of the archive. These rhythms can be generated with an O(N2) algorithm which iterates through every email address in the corpus that doesn’t belong to the archive owner, and counts the number of times it is a part of an email (e.g., listed on the to/from/cc lines of the email header) with every other email address in the corpus.

When plotting the collaboration rhythms of Shneiderman’s archive, some interesting trends become evident. Most collaborations seemed to last less than a year, and it was rare for a collaboration to last more than two years. The collaboration rhythms with the most interesting patterns generally turned out to be mailing lists (e.g. a common poster to a particular list), as mailing lists have unique email addresses too. However, even with these shortcomings, it was easy to discern the top collaborators by glancing at the sharp peaks after superimposing all collaboration rhythms into one plot. These collaborations reinforce the notion that Shneiderman’s intense email relationships focus on coordination of distinct projects over time. Without collaboration rhythms, it would be hard to get a sense of the nature of collaborations between individuals in the archive.

A limitation of this approach is that if users change their email addresses over time, the rhythms will be incomplete. However, folder metadata and the referencing user’s full name from the email header could help reduce the noise by creating more robust identities of users.

Future Work

Rhythms of relationships offer a class of information that is hard to discern from keyword searching or reading the body of the emails. However, our rhythms will only answer a subset of questions that searchers may have. Our research interests are to build on the knowledge gained in this paper, and devise additional ways that searchers can learn more about the archive.

One weakness of our use of the clustering algorithms is that they do not cluster independent of time. For instance, if two relationships have identical curves over a time segment, but occur in disparate years (e.g. one rhythm segment centers around 1989 versus a second rhythm’s center of 1996), our algorithms do not consider them similar. Interesting results can emerge by finding similar peaks and growths, such as determining if there is a typical rhythm associated with classes of people over time (e.g. a typical graduate student curve) or if a certain initial pattern of activity predicts a durable or intense relationship.

The rhythms discussed in this paper use a granularity of a year, which was motivated by our interest in understanding long-term rhythms. However, we suspect different evidence will emerge if the analysis were repeated with a granularity of months, weeks or days. In the case of Shneiderman, we predict distinct trends of rhythms surrounding academic semesters, conferences and weekends.

Although we believe our techniques are universal, so far they have only been tested on the Shneiderman email archive. In the future, we plan to test these methods on other archives to see if similar success is achievable on archives of various durations and sizes.

Conclusion

Historians and social scientists believe that email archives are important artifacts for understanding the individuals and communities they represent. However, there are currently few methods or tools to effectively explore these archives. This paper presents a novel approach by analyzing the temporal rhythms of relationships in an email archive. By visualizing these rhythms, important relationships become evident, searchers can find patterns of interest, and aggregate trends can be identified. We apply these techniques to the Shneiderman archive, and discover insights that may have been otherwise hidden.

Rhythms of relationships are an innovative way to understand email archives. However, the novel approach also comes without rigorous testing. More evaluation is necessary, but the insights observed from the Shneiderman archive offer promising expectations. We feel the techniques we introduce help provide context that is necessary for historians and social scientists to make effective use of the archives. The number and size of email archives will undoubtedly grow in future years and searching them will become a more customary task. By presenting new ways to approach the exploration of email archives, not only do we provide a new step for effective exploration, but also raise awareness for the difficult task of understanding email archives.

Acknowledgments

We would like to thank Susan Davis, Danyel Fisher, Mara Hemminger, Dave Levin and Anthony Ramirez for their thoughtful comments on prior versions of the paper. We would also like to thank Jinwook Seo for providing assistance with Hierarchical Clustering Explorer. This work has been supported by DARPA cooperative agreements N66001002810.

References

Baron, J. R. (1999): ‘Email Metadata in a Post-Armstrong World’, 3rd IEEE Metadata Conference, http://www.computer.org/proceedings/meta/1999/papers/83/jbaron.html

Donath, J. (2004): ‘Visualizing Email Archives (Draft)’, http://smg.media.mit.edu/papers/Donath/EmailArchives.draft.pdf

Ducheneaut, N. and Bellotti, V. (2001): ‘Email as habitat: an exploration of embedded personal information management’, Interactions, vol. 8, no. 5, Sept/Oct 2001, pp. 30-38.

Grieve, Tim (2003): ‘The decline and fall of the Enron empire’, Salon.com, October 14, 2003.

Kerr, Bernard (2003): ‘Thread Arcs: An Email Thread Visualization’, 2003 IEEE Symposium on Information Visualization, pp. 27.

Leuski, A., Oard, D. W., and Bhagat, R. (2003): ‘eArchivarius: Accessing Collections of Electronic Mail’, Proceedings of the 26th annual ACM SIGIR Conference, pp. 468.

Li, W., Hershkop, S. and Stolfo, S. J., (2004): ‘Email Archive Analysis Through Graphical Visualization’, Proceedings of the 2004 ACM Workshop on Visualization and Data Mining for Computer Security, pp. 128-132.

MacKay, W. (1988): ‘More than Just a Communication System: Diversity in the Use of Electronic Mail’, Proceedings of the 1998 ACM conference on Computer-Supported Cooperative Work, pp. 344-353.

MacQueen, J.B. (1967): ‘Some Methods for classification and Analysis of Multivariate Observations’, Proceedings of 5Berkeley Symposium on Mathematical Statistics and Probability"th , pp. 281-297.

Rohall, S. L., Gruen, D., Moody, P., Wattenberg, M., Stern, M., Kerr, B., Stachel, B., Dave, K., Armes, R. and Wilcox, E. (2003): ‘ReMail: A Reinvented Email Prototype’, Proceedings of CHI 2004, pp 791-792.

Sack, W. (2000): ‘Discourse Diagrams: Interface Design for Very Large Scale Conversations’, Proceedings of the 33rd Hawaii International Conference on System Sciences, January 2000, p. 3034.

Seo, J. and Shneiderman, B. (2002): ‘Interactively Exploring Hierarchical Clustering Results’, IEEE Computer, vol. 35, no. 7, July 2002, pp. 80-86

Smith, M. (1999): ‘ Invisible Crowds in Cyberspace: Measuring and Mapping the Social Structure of USENET ’, in Smith, M. and Kollock, P. (eds.): Communities in Cyberspace , Routledge Press, London, 1999.

Smith, M. (2002): ‘Tools for Navigating Large Social Cyberspaces’, Communications of the ACM, vol. 45, no. 4., April 2002, pp. 51-55.

Tyler, J. R. and Tang, J. C. (2003): ‘When Can I Expect an Email Response? A Study of Rhythms in Email Usage’, Proceedings of ECSCW 2003.

Tyler, J. R., Wilkinson, D. M., Huberman, B. A. (2003): ‘Email as Spectroscopy: Automated Discovery of Community Structure within Organizations’, Communities and Technologies, pp. 81-96.

Venolia, G. and Neustaedter, C. (2003): ‘Understanding Sequence and Reply Relationships within Email Conversations: A Mixed-Model Visualization’, Proceedings of CHI 2003, pp. 361-368

Viegas, F., boyd, d., Nguyen, D., Potter, J., Donath, J. (2004): ‘Digital Artifacts for Remembering and Storytelling: PostHistory and Social Network Fragments’, Proceedings of the 37th Hawaii International Conference on System Sciences, January 2004, pp. 40109a.

Wattenberg, M. (2001): ‘Sketching a graph to query a timeseries database’, Proceedings of CHI 2001, pp. 379-380.

Whittaker, S. and Sidner, C. (1996): ‘Email overload: exploring personal information management of email’, Proceedings of CHI 1996, pp. 276-283.