|
Visual Who |
|
Professor Judith Donath Dana Spiegel, danah boyd, Jonathan Goler |
|
In the real world we continuously, though often peripherally, observe the social patterns around us. At a conference, for instance, we notice groups of people who always appear together: students whose common bond we recognize because of their shared sartorial eccentricities, groups of suited salesmen amidst a sea of t-shirt clad researchers. Such observations help us to make sense of the complex social world we live in. Similar patterns exist in the virtual world, whether it be communities of people in an online service, the employees of global corporation, or the electronically registered participants in a conference. Yet in the virtual world, such patterns are usually invisible, hidden within vast databases; they can be seen only if explicitly visualized. Visual Who depicts these social patterns and recreates – though in a very different way – the fascination we have with people watching in real life.
Visual Who is a tool for visualizing the complex relationships among a large group of items where each item is characterized by a set of attributes drawn from a large pool of possible attributes.Visual Who can be used to investigate a wide range of datasets. Our focus is on its social use -- as a way for members of a community to explore and come to understand the roles, ideas and histories that bring them together. Users choose anchors, each representing a topic relevant to the community, and place them on the screen. As they do so, the names of the community members rearrange themselves, showing who is especially drawn to certain ideas and which members share similar sets of interests.
For example, we have used Visual Who to depict the Media Lab community, using the Lab's mailing lists as the dataset. To explore this social space, the viewer chooses a mailing list to serve as an anchor, and places it on the Visual Who screen - the names of everyone who is drawn to this list then go streaming towards this anchor. The viewer then adds another mailing list as an anchor in another part of the screen - now the names of the community members are stretched between these two anchors. Adding a third anchor spreads the population between the three marked points, etc.. Each person's location shows their relative affinity with the given anchors and people with similar affinities are clustered together. Color shows membership in specific categories, in this case, subscription to a particular mailing list. The viewer can assign a color to any mailing list, thus highlighting various patterns in the community's structure. Depth and brightness indicate how strongly a particular item is being pulled by the given anchors: two items will be near each other if they share a similar proportional affinity to the given anchors, but one may be very strongly affiliated with them (and will thus appear bright and near the top layer) while the other is only distantly related (and will be further back and dimmer).
How it works: An anchor represents a category in the dataset (e.g. membership in a mailing list). When an anchor is placed on the screen, the program calculates a profile of the prototypical member of that category: it looks at everyone who is a member of that category and enumerates the other categories in which they are members. This profile is weighted, taking into account how many members of the chosen category are members of each of the other categories, as well as how exclusive are such memberships. The program then compares each item's category membership (in our example, each person's mailing list memberships) with the weighted prototype and attaches a spring from the person's name to the anchor, the strength of the string being proportional to their similiarity. Thus, when there are three anchors on the screen, each person has three virtual springs pulling them, with varying strengths, towards the three anchor points. For a detailed discussion of the Visual Who algorithm, see "Visual Who: Animating the affinities and activities of an electronic community", ACM Multimedia '95.
