Edward Tufte: Visualizing Data

I went to a one day course in Boston at the end of October led by Edward Tufte, the author of many books on data and information visualization. I have long admired Tufte’s work, having read The Visual Display of Quantitative Information a number of years ago. He creates beautiful books full of graphs and maps and other visual displays of information. In my General Education Coordinator role, I’ve been thinking a lot about how to present information to students, faculty, and staff regarding the Gen Ed program and other aspects of the Integrated Cluster initiative. I’ve been experimenting with infographics and thought Tufte might be able to help me make them better. In addition, I was intrigued that the one day course would also cover “A new, widely-adopted method for presentations: meetings are smarter, more effective, 20% shorter.”

The workshop started at 10am but the agenda for the day had an hour long “study hall” that started at 9am. I didn’t understand what that was about but left my house that morning in plenty of time to get to the workshop by 9:15ish, depending on traffic. The trip was fine and I was on time until I got about 8 miles from the venue. Because of heavy rain and heavy traffic, it took me an hour to travel the last 8 miles and get the car parked. So I ended up being late and didn’t arrive in the large hall until about 10:10am. So I missed the initial information about the role of the “study hall.” Tufte explained it later in the day, however, since it plays a big role in his “widely-adopted method for presentations.” More on that later.

I learned quite a bit from the course, some of which I’ve already started to implement. I also disagree with Tufte about some points. But my main takeaway is that Tufte’s presentation of this information is far too slow! My mind wandered quite a bit as he focused on details about particular visual displays that seemed superfluous to me. They might have been key to the points that he was trying to make but too often, he didn’t pull out larger points that we could learn from the particular visual displays to be able to apply them to other displays. I’m still glad I went but I think we could have learned a lot more that would be immediately useful if he had someone who could tell him which details are crucial and which are not or who could help him identify why he believed each detail was important enough to expound upon.

So here are the main points about presenting data and information that I learned.

Tufte hates PowerPoint as a presentation tool. On p. 158 of Beautiful Evidence, he writes: “PowerPoint is presenter-oriented, not content-oriented, not audience-oriented. … PowerPoint’s convenience for some presenters is costly to the content and the audience.” (emphasis in the original) He thinks that creating bullet points foreshortens evidence and thought by requiring the evidence and thought be broken down into slogans using imprecise language. Each slide is low resolution, with very little information content on it to the point where many slides are needed to convey information. The slides fragment data and any narratives the presenter might tell using that data. The audience is required to consume the data and information in the hierarchical single-path determined by the presenter. These weaknesses of PowerPoint, Tufte says, turn the presenter in someone trying to sell an argument. So what would Tufte suggest instead?

In Tufte’s view, the purpose of displaying information is not to sell an argument. Instead, these displays should help people reason about the content. One of Tufte’s main assumptions is stated on p. 158 of Beautiful Evidence: “Visual reasoning usually works more effectively when the relevant evidence is adjacent in space.” After doing an analysis of PowerPoint presentations and a number of worldwide publications such as Science and Nature, Tufte concludes that the statistical graphs created as part of PowerPoint templates are very low resolution, presenting an average of 12 numbers per graph compared to more than 1000 numbers per graph for Science and more than 700 numbers per graph for Nature. To display information in ways that help people reason about the content, Tufte starts with the basics of analytical thinking, the tools that people use to analyze information, and develops six design principles related to those basics.

The first principle for the presentation and analysis of data is to show comparisons, contrasts, and differences because Tufte believes the fundamental question asked in statistical reasoning is “Compared with what?” So the visual presentation of information should make the important comparisons as clear as possible.

The second principle is show causality, mechanism, dynamics, process, and systematic structure. As we collect data, we start to ask questions about how the data are related to each other. Does one thing cause another? How do the various pieces of data work together to create a system? By representing the relationships among data, the presentation helps the viewer reason about that data.

The third principle is to show multivariate data. To reason effectively about data, a viewer needs to see multiple variables adjacent in space. The world is multivariate, Tufte argues, so our displays of information about the world must also be multivariate.

The fourth principle is to completely integrate words, numbers, images, and designs. That is, the viewer should not have to look in a variety of places to understand the data presented. For example, tables of data should not be presented in an appendix. Instead, the data should be tightly integrated with the text to make reading easier.

The fifth principle is to provide documentation of the evidence as part of the evidence. “Thoroughly describe the evidence. Provide a detailed title, indicate authors and sponsors, document the data sources, show complete measurement scales, and point out relevant issues. (p. 133, Beautiful Evidence)” This documentation should not be relegated to the back of a book but instead should be part of the visual display of the information. For example, Tufte uses sidenotes in his books, rather than footnotes or endnotes. The sidenotes appear space adjacent to the text that they are documenting.

And finally, the sixth principle for the presentation and analysis of data is “analytical presentations ultimately stand or fall depending on the quality, relevance, and integrity of their content. (p. 136, Beautiful Evidence)” In other words, the most effective way to improve a presentation is to get better content. No amount of gimmickry will make the presentation better if the content is flawed.

What do these principles mean in practice? It means that visual displays of information are packed full of information. Information about different kinds of data are presented adjacent in space in the display to aid the viewer in understanding relationships and drawing conclusions. In fact, many of the displays in Tufte’s books are so jam-packed full of information that I find them challenging to decipher. And I guess this is where I start to disagree with Tufte a bit. While I agree with him that many information displays today are so oversimplified as to be meaningless, I think we can easily go too far in the other direction where we present so much complex information in a single display that it ends up being meaningless as well. So that’s the topic I would have liked to have heard something about–what is the optimal amount of information to put into a visual display. If there is no single optimum, I would have liked to hear some discussion of when we cross the line into trying to do too much with a single display. Tufte seems to think the more the better with no thought of overwhelming the viewer.

A similar issue came up for me when learning about Tufte’s “new, widely-adopted method for presentations.” Here is how it works. The presenter creates a 2-6 page document that contains the agenda and major topics for the meeting. The document uses full sentences and paragraphs as much as possible. The document is handed to participants at the start of the meeting, not before. The meeting then begins with a “study hall” that comprises 20-60% of the total meeting time. During the study hall, meeting attendees read and review silently review the document at their own pace. By doing so, they are able to jump from spot to spot in the document as their interests guide them. After the study hall, the presenter talks about the information in the document, kind of orally annotating that information. The meeting attendees already know the big picture of the meeting and can help direct the presentation with questions they might have about the information presented.

I actually love this idea but have to say that Tufte’s own implementation of it was quite problematic, recreating many of the problems he designed this process to avoid. The one page document that he gave us consisted almost entirely of points (without the bullets) that were references to other work. For example, he pointed us to read his chapter on why PowerPoint is so problematic. The biggest problem I saw, however, is that Tufte spoke at us for the entire workshop without soliciting a single question or really engaging with us at all. In fact, for much of the time, the room was dark with a spotlight on him and some sort of image projected on a screen. These were not PowerPoint slides but they could well have been since he displayed them one after another. This felt problematic to me because a big part of this workshop was focused on how to organize information in space, rather than through time. In other words, rather than creating a slide deck which presents information through time, we were encouraged to develop visual displays where information and data exist adjacent in space. But Tufte’s entire presentation presents information through time, with one point after another. And because he did it so slowly, I had a lot of time to think about how he wasn’t following his own advice.

As I said, I’m glad I went to this workshop because it has given me a lot to think about and I’ve already started to try to use some of what he said in my own presentations. It’s always good to question our own practices and try to get better at them.