Edward Tufte commented on the use of his sparklines idea in Knowledge Flows. He writes: "Those aren't sparklines, which are small, intense, high-resolution, wordlike graphics. The graphics in the pdf look like a routine simple barchart." Fortunately, I phrased it carefully and can understand why Tufte doesn't like the knowledge flow graphics to be called sparklines. They are not simple barcharts either though.
A significant challenge of visualisation is to think of ways to visualise information that is not so easy to capture as a sequence of numbers over time. Most examples of sparklines Tufte gives in the draft version of the sparklines chapter of Beautiful Evidence have two dimensions. Often these are time and value. Tufte deals with visualisations of data that can be obtained and then abstracted in an unambiguous manner, for example by measuring an observable (e.g. temperature) or by counting (e.g. the US deficit). The average knowledge worker needs visualisations that are not solely based on such "absolute" numeric values, often the abstractions are derived from computations that are founded on statistics, heuristics or a combination of these. Many corners may have been cut between the original data and it's interpretation before the results can be visualised. I think we are entering an era in which there is a need for visualisations for observables to which we cannot assign numbers in some absolute sense.
One visualisation perspective I consider very powerful (especially when precise data is not available and the visualisation itself is complex) is to take advantage of the human capacity to compare visualisations and to let the user derive meaning from mentally interpreting the difference between visualisations arrived by the same method using different sets of data. The graphics in the knowledge flows post are an example. The precise meaning of a single of these graphics cannot be easily established, the value becomes apparent when one compares the graphics. I thought this was also one of the purposes of sparklines: by making them small comparison becomes possible on the same page. Apparently, I'm mistaken given Tufte's comment.
In order to produce a visualisation it is necessary to abstract from the available data. The traditional method is to turn the data into numbers (usually by observation), select the dimensions and then produce the visualisation. Tufte's books cover this process extensively. What to do if there is no obvious (or traditional) way to follow this process?
Visual Settlements is an example. It compresses a single weblog to a visualisation in relation to linkage to a set of related weblogs. All traditional methods to display this kind of information seem to fail here and one has to revert to something beyond the ordinary.
In summary, I think the main visualisation challenges and issues are:
- What to do if data cannot be compressed to numbers in an unambiguous manner. (Think of a visualisation of a Google query as a graphic rather than a textual table; is that possible and would it help?).
- Exploit the human capacity to compare visualisations.
- Visualisation and interactiveness. Not addressed above, but interactiveness certainly introduces a dimension that is potentially very different.
I'll pursue these challenges and issues in the knowledge of making many mistakes during this process. It has taken hundreds of years to establish the visualisations we now appreciate and can interprete, as for example documented in Tufte's books. The honesty of visualising data takes on a new meaning when it is necessary to derive numbers from data that does not naturally carry them.
Join!
Intuitively, I'd support your idea that human eyes can compare and calibrate even faulty visualisations. But I couldn't prove it, and I'll hope you can. Has Tufte lost the idea of the immense computation that powers intuition? Practically, the existence of a connection is a big number compared to the absence of one. He seems to be being excessively purist in his definitions of what a quantity is.
Posted by: Matt Whyndham | April 12, 2005 at 10:15 PM