Managing information overflow

-- lecture notes, September 21, 1999 --


The goal of many researchers and designers is to help people cope with large quantities of information. The interest is of course fueled by the increasing adoption of the Internet. The Web as viewed through a regular browser appears to most as a very crowded information sewer. It is fairly obvious that there is room for improvement.

However, useful groundwork has been done for quite some time in the fields of interaction design and scientific visualization. Here, I would like to outline three relevant dimensions to consider along with some examples. There are also a few suggestions for further studies.

Level of abstraction  

An information presentation can be based very closely on the structures and content of the underlying data. This could be called a literal approach.




The early work at Xerox Parc has been characterized by literal visualizations. The examples above show the ConeTree and the WebForager. The ConeTree is a straight representation of a hierarchical data structure. The WebForager builds on the notion of web "pages".
From Robertson et al., Comm. ACM 36(4):57-71 (1993) and Card et al., Proc. CHI'96, 111-117 (1996).



At the other end of the spectrum, we find designs that abstract much more from the data. Typically, this involves condensing and fusing large amounts of data into a few perceptual variables.

The example on the right is a concept for the visualization of several Usenet hosts from a user's personal perspective. The shapes of the spheres, and the color, texture, shape and position of the mold patches, indicate new discussion activity that the user might find interesting.


Clicking a mold patch obviously leads to the corresponding discussion threads.


Graphic design typically relies on static displays. In interaction design, the dimension of time is more explicitly introduced. In other terms, we design interactive information artifacts.

The dynamic query concept is a fine illustration of interactivity in information visualization.




The database scatterplot (left) is nothing in itself. The power comes from the tight coupling of input and output, where filter manipulations (right) are instantly reflected in the presentation. This facilitates serendipitous discovery of data relationships and trends that are invisible in a static display.
From Ahlberg & Shneiderman, Proc. CHI'94, 313-317 (1994).


In the Visual Thesaurus from Plumb Design, the interactive qualities play an important part in changing the user's view of what words and concepts are.

More on the VT as seductive technology.
The Visual Thesaurus (requires Java).

Social aspects  

So far, we have concentrated on a single user wading through seas of information. But in most cases, and certainly on the Internet, the user is above all a member of different communities. Some are explicit -- dedicated interest groups for parakeet owners, for example -- but most potentially relevant communities are not known to the user.

Much experimentation has been devoted to social filtering and social navigation. A typical definition is that social filtering works from data that users generate anyway (navigation patterns, etc.) wheras social navigation involves active interaction between community members.




Phoaks is an example of social filtering. It indexes Usenet newsgroups and presents resource recommendations in frequency order. The idea is that if several posters have recommended a particular source, it might be useful also to people outside the particular newsgroup.

Alexa is also primarily a social filtering tool. It suggests places where many surfers have moved from the site you are currently at.




However, Alexa also has an active recommendation component. If you particularly like or dislike a site, you can indicate it (above). Your rating goes into the database and affects whether the site in question is recommended to other Alexa users.

Of course, there are many examples of more interactive community support systems. In my opinion, the trick is to balance short-term effort and long-term gain. Most people need to anticipate benefit for themselves before undertaking any significant amount of voluntary effort (such as rating a large number of sites, or writing reviews). Alexa might be on the right side of the threshold for collecting even casual surfers' ratings.

In general, I believe that the smaller and tighter the community, the more you can ask and expect in terms of altruistic contribution.

Further studies  

Card S, Mackinlay J, Shneiderman B (1999). Readings in information visualization: Using vision to think. San Francisco: Morgan Kaufmann. An exhaustive reader in the field of information visualization. Most of the classical papers are there along with some useful new material.

Tufte E: The visual display of quantitative information (1983), Envisioning information (1990), Visual explanations (1997). Cheshire, Conn.: Graphics Press. Necessary reading for anyone concerned with efficient and useful presentation of information. Also beautiful examples in themselves.

An Atlas of Cyberspaces. A collection by Martin Dodge on various means of mapping information spaces. Includes topological, conceptual, social and artistic approaches. And much more.