Eric Fischer recently published a visualization of traffic in the Bay Area based on location and time data associated with photographs. It’s a lovely image that traces the major routes in the Bay Area and highlights points of interest to locals and tourists alike. The image is beautiful, reminiscent of sketches or studies done in preparation to painting.
But what about its effectiveness as a visualization? What does it tell us that we didn’t know?
Jeremy Pickens (wearing his curmudgeon hat) wrote:
Think about your own mental model of San Francisco, and where you think the interesting or appealing places are. I.e. if you were visiting the city and/or out for the day taking pictures, where in the city would you go? You’d probably go to Golden Gate Park, to anywhere along the water, and to the downtown shopping areas (the grid around and north of Market). And that’s pretty much what the map shows, right? So we don’t really learn much from this, do we? It confirms our existing intuitions.
He is partially correct.
Jeremy’s point is that visualizations should help us explore and gain insights into data. To those familiar with the Bay Area geography, this image is a useful, elegant reminder of what we know to be interesting places, but there is little true insight in it. The visualization doesn’t tell us anything that we don’t already intuitively know. Visitors to the area, on the other hand, might use such an image in an interactive way to help them plan their itinerary, to explore, to learn about the place.
There are nice parallels here between exploratory search and recommendation systems: recommendation systems (such as the image in question) show where others have been (here, literally) to help others find their way later. Exploratory search systems, on the other hand, are intended to help people map out previously-unexplored areas. These are uncharted, or at least under-charted, areas in which few have yet ventured.
The Choose Your Own Adventure piece by Christian Swinehart that I described last November is an elegant example of exploratory visualization. He built some interesting diagrams to display various narrative structures in a set of hypertext fiction books. The visualizations are designed to reveal structure that is not necessarily visible otherwise, and the elegant presentation and bold graphical design make interacting with them a pleasure.
The upshot is that visualizations can support both entertainment and exploration, and that the two aren’t always separable. But we should be clear about the intent of the visualization and try to design visualizations with enough depth to be useful not just as pretty pictures, but also as vehicles for generating the appropriate type of insight.
Jeremy Pickens contributed to the writing of this post.