If you want to visualize something in its fullness you have to look at every perspective and every angle, and that means - sometimes uncomfortably - looking at things you don't want to look at.

The classic one for me is one of my favorite images - left-versus-right political-spectrum image. I was trying to visualize the concepts on the political spectrum. I'm left-leaning, and I discovered as I was doing it that I had an impulse to make the left-hand side appear better than the right-hand side. That was manifesting in the way I was choosing certain words, framing certain ideas. I shared it with a few people, and they all said, "Oh my God, this is really biased." I hadn't seen it at all.

There's a tendency in graphics to allow the trimming of certain parts. But I think that if you're open about your process, your methodology, such as introducing thresholds, introducing filters, techniques people use in research and data management, it's legitimate. It's legitimate to say, "We're only going to show data above this level, or between levels."

I think everyone is struggling somewhat with presentation. The Internet is generally well designed, if you look at the most popular websites, so we expect our visuals to be at that level of quality. When you sit in a presentation and you're looking at nonsensical pie charts and the like, your audience does disengage. People across a range of industries, not just science, are struggling with their communication because their output doesn't compete with what people see on a day-to-day basis.

For years I thought I was just a writer, but when I sat down to design and started playing around with it, I realized that, really, it's pretty easy. Obviously it's more than just a set of rules, but the basics of design are actually pretty simple and quite mathematical. The link between data and design works at quite a fundamental level.

I'm doing a lot of cognitive processing. I'm gathering research. I'm processing it. I'm arranging the data. I'm sorting out the narrative. I'm designing. It's almost as if I do all the cognitive work that you then don't have to do. I digest it, process it, and then offer something that's very easy for you to digest.

What I tend to do is blend quantitative with the qualitative to allow me to plot the qualitative data in some way. It's a question of what quantitative data are most applicable. So I'm playing with that, merging the two.