Fragmentation, windows, tasks: myths, realities, novelties

posted by Eleanor Wynn on February 14, 2007

I remember the pathbreaking article Larry Tesler wrote in Byte Magazine in about 1980, discussing the absurdity of having to change “modes” in order to change applications. So you would start from a command line and choose your app and then continue with your content. The windows desktop concept originated at Xerox PARC, that is where Steve Jobs saw it; and Larry eventually went to work for Apple. One early version of the concept at PARC was called “OfficeTalk” (in the same family as “Smalltalk”), designed by William Newman. I made a chance remark to him about how people find things visually, and he designed an interface around that idea of physical objects. OfficeTalk had a file cabinet, sheets of paper and other simulations of objects used in the old fashioned “office”.

Rabbity desktop windows

Now those objects, which we see incarnated as desktop windows, have proliferated like rabbits. I counted 12 open windows on my display just randomly, with unknown numbers of minimized windows in the tray below. I admit my physical desktop is also problematic. I don’t want to put things out of sight lest I forget them, but they soon get out of sight anyway, and unsightly. I don’t really have a hierarchical mindset, though I do use hierarchical file structures, especially when I know what I am looking for and where something “belongs.” All of this mess, regardless of personality types, is compounded by the way we work these days.

Multi-teaming, multi-tasking, multi-apping

In our trending surveys of Intel’s workforce, we found that 2/3 of the people work on 3 to 5 teams, or more. Strangely, satisfaction with teams was lowest at only 1 team, moved up at 2 peaked at 4 teams, and after a dip at 5, gained a steady state 5+n . Our guess is that at 5 teams people are actually doing too much work; after that they may just be “participating”. Our further guess is that people like to be on multiple teams, up to a point, because it gives them a broader social network, better shots at having positive colleague feedback and a feeling of knowing what is going on in the organization.

Learning from TaskTracer

There is a cost to multi-teaming, but I don’t believe it is in the realm of infomania per se. It is related to how the desktop is implicitly structured. Of course I am not talking about all desktops and all operating systems, but I am talking about the 2D desktop, overlapping windows, and application discreteness. We have been working with Oregon State University Department of Electrical Engineering and Computer Science for over a year to test a machine learning program called TaskTracer. My initial reason for wanting to try TaskTracer was simply to get a sense of the effect of multi-teaming and multi-tasking on user activity. I knew this was not something you can sit and observe because it is highly granular and quickly changing. I thought a machine learning program would be the best way to really understand multi-tasking, which, by the way, is partly a result of multi-teaming.

What the heck is a task?

Something I did not anticipate was that there is no clear definition of “task”; that is something that our collaborators at OSU had given a lot of thought to. It turns out that there is a kind of exponential curve: multi-teaming, multi-tasking and then: multi-application tasks. That’s right. This notion that you sit down with one application and labor a way at a “task” is an oversimplification. In the preliminary data that TaskTracer produced, it turns out that a “task”—something the user labels as a particular activity—often requires access to documents, spreadsheets, e-mails, web searches, file folders, etc. The desktop doesn’t treat these applications as related at all. So every time you want to do that task, you have to go look for all the parts of it.

TaskTracer is trying to learn the configurations of a given user’s tasks and then link them together for the user. But in doing that it has given us some astonishing information about patterns of activity. For instance, very high (33-44) search repeat rates, both on Internet search engines and through hierarchical file folders. It seems as if users actually spend the bulk of their time navigating! I am not citing hard data because of the small numbers we had for preliminary tests; but even with a small sample it is possible to see holistic information such as you would gather in an ethnography—of how everything ties together. I have used TaskTracer myself and was amazed at the amount of subliminal mult-tasking I am doing. With an active TaskTracer, that actually predicted activity, you would have related pieces of a task pulled up together. Once you remove the assumption that this—switching constantly between windows—is “how the world is”, it becomes amazing how many weeds we have to whack through on a daily basis. That is, one set of application interconnections is likely to be replicated across many users in different permutations because it is structurally dictated.

There is another way

Well, this leads to another story, but that will be my next post: about the Miramar 3D user interface developed at Intel Architecture Labs in 2000 and currently experiencing, a rebirth or re-use in IT Research. This was one of those lucky inventions rescued from the trash bin of history, just because I sat near the inventors and had to come up with a visual environment for a demo of a concept—and hey, why not Miramar? We “recycled” it and it is now, after a lot of development work, getting the attention it deserves.

Watch this space for compelling breakthroughs in how we navigate those screens we are all staring at right now. So, do you think this would be a good use for Quad Core with all the graphics, scaling, snap-to movement, linking, cascading permissions, and so forth?

Comments (1)

tagged: 3D, applications, clutter, Miramar, multi-tasking, multi-teaming, task, windows