Reflections on the Transformation of Email*

Thomas Erickson

snowfall@acm.org

My computer screen looks weirdly 'wrong,' as though some catastrophic system or network failure has just occurred. Everything is really just fine, but my reaction to what is really just an unusual state of affairs causes me to pause, and reflect on the cause of my unease: My email in-box is entirely empty.

The disconcertingly empty in-box hanging before me belongs to the Eudora email program. Its emptiness is unusual because, like many people, I use my in-box as a holding bin for messages that I want to be sure to read over later, or respond to, or otherwise remind me of something. I think it's been a couple of years since it's been entirely empty. (This is not to say I'm out of things to do -- I have other to do lists and reminder systems that are quite well stocked, thank you.)

Figure 1. A relatively rare sight: an empty in-box.

The causes for the emptiness, per se , are relatively straightforward, if unusual. Rather, what is interesting grist for the reflective mill is the way in which my use of email has shifted over the last several years. This shift is the result of multiple factors, ranging from the software I use, changes in my own work practices, and changes in social practices, and I think it's worth exploring some different ways of thinking about the role of email -- and more broadly, communications -- in future computing systems.

I began with Eudora, appropriately enough, as a consequence of an email message: a friend wrote suggesting I try it, saying that it would change my life. As my friend is a well known researcher and writer on the social effects of computing technologies, and particularly chary of hyped claims about high tech stuff, I was intrigued by his seemingly hyped claim. So I tried it.

Eudora did, indeed, change my life.

Of course, at the time I was using the UNIX "mail" program to read my mail (and I was accessing UNIX through a terminal emulation program), and so it was not hard to do better. The ability of Eudora to automatically download email every so often -- whether it be every five minutes or once a day -- only returned me to a state of grace that had been familiar when I lived in UNIX in my pre-Apple days. The other new feature that Eudora had was more novel: Eudora included an easy way to create multiple mailboxes, and made it straightforward for me to a particular message into a particular mailbox. So, as I used Eudora, I begin developing a fairly well ordered archive of email. Taken together, these two features greatly improved the experience of receiving, reading, managing, and storing email; however, I can't really say that they changed the way in which I used email. That came with the next major release of Eudora.

Sometime in the spring of '94 I began using version 2 of Eudora, which featured, among other enhancements, automatic filters. Automatic filters allowed me to define rules for automatically filing mail, so that Eudora could put messages that fit certain characteristics (e.g. to or from a particular address) into particular mailboxes. At the time, filters just seemed like a convenience, but as a consequence I gradually began to use email in new ways. The proximate change was that I began to subscribe to more and more mailing lists, mailing lists of sufficient volume and insufficient relevance that I was neither able nor inclined to read them on a daily basis. I could now create automatic filters that would allow the email from various lists to slip unseen into various folders, where I could browse or read it at my leisure.

Now let's jump ahead in time, and look at two ways in which I'm now using Eudora that are a consequence of this automatic filtering. I subscribe to a mailing list called Red Rock Eater (RRE). RRE is a moderated, send-only list. It consists of articles that its moderator, Phil Agre of UC San Diego, finds of interest, primarily articles about the social impacts of computing, particularly those that are relevant to his progressive/humanist point of view. The list is high volume, not so much in number of messages as in the length of each message. Even though I find about one out of every fifty or a hundred articles immensely valuable, without Eudora's automatic filtering and filing I wouldn't subscribe, because I couldn't keep up with the list's volume.

Another example is my subscription to the FAQ-Maintainers list. Last fall I became interested in FAQs (Frequently-Asked-Questions lists), because they represent one of the relatively few examples on the internet of cases where knowledge is being constructed in a concentrated, well-structured, accessible form (as opposed to being distributed through long, divergent conversations). Wanting to understand more about the creation and management of FAQs, I subscribed to the FAQ-maintainers list so I could get a glimpse of the various issues that come up as the practices and policies of FAQ custodians are described and debated. Now a dozen messages a day pour into that mailbox--and this has the useful side effect of keeping FAQs in the periphery of my awareness. Occasionally I skim a few to get a sense of the conversation, but by and large I'm simply letting them accumulate, building a corpus of information that I can examine at my leisure.

So here are two new ways in which I'm using mailing lists, now that I'm freed of the chore of managing the message flow on a daily basis. In the first case, messages are filtered by a person whose point of view I value, and then I 'mine' the result for things of interest to me. In the second case, messages from a community of people involved in a common activity (FAQ creation and maintenance) are automatically being compiled into a research corpus.

Now, in some ways, these uses of mailing lists doesn't seem terribly different from getting on the internet and reading newsgroups. However, there are two key differences. The first has to do with awareness: the messages come to me, rather than I going to them. This is important because these are not mission-critical things, but rather topics I'm mildly curious about. While curiosity might provoke me into going out and browsing once or twice, it would rarely keep me doing this on a daily basis unless the topic in question became central to my job. But, with mailing lists, I can turn on a stream of messages on a particular topic, so that they keep tickling me with their subject lines if not with their content, until I either grow more interested, or grow bored and turn the stream off. And while they stream by in the periphery of my attention, at least some of the messages have a chance to engage, to resonate with, to play off of, or otherwise interact with my more central interests. This makes my intellectual environment richer and more stimulating.

A second difference between my use of Eudora and going out and browsing the net is that the messages are present on my own disk, on my own PowerBook. This is important because the times in which I usually browse the collected messages are "spare moments," when I don't have other pressing things to do: perhaps on a plane trip; perhaps in the audience for a less-interesting-than-anticipated talk; perhaps while waiting for a late meeting to finally get started. Significantly, these spare moments generally occur while I'm away from my normal infrastructure, including network access; it is this partial isolation that creates the spare moments: if I were in one of my regular workplaces, there would likely be more pressing tasks at hand. So, for me at least, having the information available locally, rather than over the net is crucial.

These ways of using Eudora don't feel at all like "doing email;" I'm not dealing with personal messages directed specifically to me. To me, it feels more like the way I use newspapers. Consider the following parallels:

• there is more than I generally want to read in any one edition
• the default way of interacting with the content is browsing -- I only read when I find something I particularly care about
• the main purpose of browsing most of the content is simply to make me aware of what is going around me: in my group, division, company, profession, and industry.

I think this parallel between my use of Eudora and my use of newspapers is quite interesting because six years ago I participated in the design of an electronic newspaper... and what I designed then lacks a number of the features that I now value in Eudora. The nature of the features that are absent -- and the reasons for their absence -- is telling.

The electronic newspaper was originally named "Hearst" by a high level manager, but the team, not appreciating all the connotations of that moniker, renamed the project Rosebud, drawing from the movie Citizen Kane. The basic idea of Rosebud was that it allowed its users to keep up with information that was scattered across a network of distributed, heterogeneous, and ever-changing databases. Rosebud had three components: Reporters, the Newspaper, and Notebooks. Reporters were search agents that could be given topics (i.e. queries), and pointed at one or more databases, and programmed to look for new information every day (or however often you wished). (Rosebud's Reporters --and only the Reporters -- eventually evolved into the product AppleSearch.) Each Reporter owned a column in the Newspaper (figure 2 ), and so when it returned with its findings it would publish them there. Typically the Reporter would just publish an excerpt from each thing it had found so that the Newspaper was more browsable; if the user was interested in something, the Reporter could fetch the rest of it. Finally, Notebooks were provided as a place to store and annotate interesting information. (A more complete description of Rosebud and the studies which informed its design can be found in Erickson and Salomon, 1991). So, the basic use scenario we envisaged was that a user would create a Reporter for each type of interest he or she had, point it at appropriate sources, and articles would begin showing up in a new column in the daily newspaper. The user would reach the news for each day, and save interesting articles in the Notebook. As the user developed new interests, new reporters would be created; when a topic became uninteresting, its Reporter would be decommissioned or re-purposed.

Figure 2. The Rosebud Newspaper. Each Reporter (listed in "Contents")
would gather new information on its topic and publish it in its column.

Now let's look at how Eudora, and particularly the features that support my 'newspaper-ish' use of it, differ from those of Rosebud, which embody what I envisioned as an electronic newspaper five years ago. I see three important differences. The first difference is that unlike the newspaper, Eudora does not have daily (or even separate) editions--it's cumulative: the new messages all pour into the same (multi-part) container. No management is required. In contrast, the newspaper model assumed -- implicitly -- that the newspaper would be read every day; as designed, you'd have to open seventy two newspapers to browse through the last six weeks of content from a particular source. It's interesting to note that there was no strong principled or technical reason for this decision: it was simply inherited from the choice of metaphor.

A second difference is that Eudora provides a framework within which conversation can occur. Because it is easy to respond or forward any piece of content, Eudora is biased towards discussion, whereas Rosebud was designed as a one-way retrieval and presentation system. Let me give an example of why the "conversational nature" of Eudora is important. On August 12th I was browsing through a few of the 200 unread messages in my RRE folder. I came across an interesting message from early June, and casually forwarded it to my group and one other person, with whom I'd been discussing similar issues. I had no expectation that anything concrete would result; I was just sharing an interesting article. However, the "other" person forwarded the message to his group, and some of the replied, and, over the next three days more than a dozen messages were sent, a brainstorming meeting was organized, and so on. The point is that Eudora made it very easy to shift from acquiring information to sharing and discussing it with others. In contrast, the Rosebud Newspaper was designed principally for reading, and for annotation by its user (e.g. allowing highlighting, clipping of articles, etc.). While we were quite conscious of the fact that newspaper articles were shared -- indeed, we observed that some of the users we studied made a regular practice of clipping newspaper articles, marking them up, and circulating them to colleagues -- we envisioned such sharing being done via a separate email application into which the user cut and pasted articles. However, at least in this case, had there been the extra overhead of launching a separate application, I might not have gone to the trouble since I had no reason to expect any particular payoff from sending the message. Why didn't we simply integrate email into Rosebud? After all this time it's hard to say, but again I find myself wondering about how much of a role the newspaper metaphor played.

A third important difference between Eudora and Rosebud is that Eudora uses people as agents rather than having electronic ones. That is, in Eudora, mailing lists moderated by people are playing the role that I originally envisioned for Reporters. And that's a good thing. For me, at least, AppleSearch never worked because the Reporters I defined brought back nothing but junk. Perhaps the retrieval technology was not sufficient, or perhaps the Reporters were pointed at sources where there was nothing I wanted, but nevertheless they always brought back junk. Sure, some mailing lists are full of stuff I don't care about, but not all. Mailing lists have the advantage of having human intelligence behind them, and mailing lists that are moderated by one person, or that have a very specialized topic and a small group of contributors, are very worthwhile reading... or at least browsing.

What these last two differences point to is the importance of allowing room for human, social action. Rosebud was conceived as a single user system: it's electronic agents gathered info, and brought it back to one person who could mark it up and save it. In contrast, all the content within Eudora has links back to people, and Eudora makes it trivially easy to 'follow' those links by replying, to to forge new links by sending and forwarding. Eudora takes advantage of a lot of things we take for granted: people are smart; people are engaged by topics that interest them; people gain recognition and status by finding and sharing information; people like to transmit and reinforce their points of view; people like to talk things over. Eudora works significantly better than, I believe, Rosebud newspapers would have (though the newspapers were never actually implemented), because Eudora provides a framework which leaves room within it for humans to do what they're good at.

Reference

Erickson, Thomas D. and Salomon, Gitta. "Designing a Desktop Information System: Observations and Issues." Human Factors in Computing Systems: CHI '91 Proceedings (ACM: 1991).