The Problem with the “Deception” and “Lie” Framing in Visualization
Lies and deception are just the tip of the iceberg. The challenge is way bigger and should be framed more positively.
In 2015, with some great colleagues, I published one of the papers I am most proud of, “How Deceptive are Deceptive Visualizations?: An Empirical Analysis of Common Distortion Techniques.” In the paper, we devised an experiment to verify to what extent a set of standard distortion techniques affect the reader’s interpretation of charts (it turns out the techniques affect interpretation substantially). I like this work because I had great fun working on it with my colleagues and my former student Anshul and also because when we came up with the idea, we were like, “Somebody has to run that experiment!” And we did. However, after many years, I regret using the term “deceptive” in the title. Let me explain why.
In visualization, there is a long history of framing specific problems as people “lying” or “deceiving” with visualization. For example, Tufte introduced the idea of the “Lie Factor.” For many years, the IEEE VIS conference (the premier conference in visualization research) had a session called “Vis Lies.” And new papers appear periodically using the lie and deception framing.
And, of course, in general, Statistics has always been targeted as a way to produce sophisticated types of lies, Huff’s classic “How to Lie with Statistics” being a prominent example.
While I am sure some people lie and deceive intentionally with visualization, I have two main problems with this framing. First, the negative and pessimistic stance. Second, the narrowness of the framing. Let me explain each in turn.
The lying and deceiving frame implies an explicit negative intent on the visualization designer’s part, but many more people are driven by motivated reasoning than a desire to fool others. Interpreting data is hard. Creating good data visualizations is hard. Heck! Even scientists are pretty terrible at designing their charts when writing papers, which should be their bread and butter. I worry that the deception framing induces people to think that a large majority of charts are designed to deceive when they are primarily designed to persuade.
Most charts are made by people who have a preconceived notion of something and want to show you the evidence. Most of these charts may be biased, but I don’t think they are made in bad faith. In fact, a better framing for charts is “argumentation.” People just provide evidence for the kind of argument they want to put forward. It’s the reader’s responsibility to consume this information critically, as we would with written text.
The second issue is that when we label the problem as “deception,” we implicitly assume that if we could somehow prevent the bad actors from operating, we would have made some progress. But, as I mentioned above, data analysis is hard. Data communication is hard. Data interpretation is hard. So, even when people are in good faith, we can have suboptimal or “sincerely biased” charts on the one hand and poor reading and interpretation skills on the other. In other words, lies and deception are just the tip of the iceberg. The challenge is way bigger.
One question is how to improve the situation. I honestly don’t know. I have been dabbling with data long enough to be humbled by the challenge. More education and higher statistical literacy seem to be potentially good. I remember someone arguing some time ago that giving people more data-thinking skills will not make people more reasonable and objective but more skilled in defending their arguments with data. This is possible. But maybe if we look at the aggregate effect of having a multitude of people with high data-thinking skills, then at least their arguments will be of higher quality.
That said, it’s evident that better skills won’t suffice. What is needed is engaging with people in good faith and with respect. But in the last few years, we have gone backward, and something needs to be done. In fact, as I write this, I realize that teaching how to have productive dialogues when we disagree is maybe the master skill we have to cultivate. Heck! Even very high-profile people I respect can’t resist making ad hominem arguments! This is maybe the playground where we need to make the most of the progress. But how are we supposed to do that?
Hi Enrico — strongly agree with this on the flawed binary framing emphasizing intent to deceive. Intead, in most cases it is making choices based on persuasion of peers or the broader readership. I think there is fertile ground to think about this in relation to how data visualization is used at major news outlets like the New York Times or Washington Post on a broad range of topics — but especially on climate change. Consider for example this 2019 widely circulated and influential ProPublica/NY Times series on "Climate Change and Migration." Notice that the default view for readers in the visualized climate impact projections are RP8.5 which is a "worse case" scenario that even by 2019 most scientists knew was an impossible scenario that could be ruled out [it assumes continued, widespread deployment of coal plants for power generation worldwide.] Yet it is that scenario rather than the likely RP4.5 scenario that is the visually prominent projection and what is emphasized in the text of the article. The same is for the related studies that the article draws on. RP8.5 is visualized in the main body of the studies and RP4.5 in the supplementary materials. Something to discuss together when we have a chance. Below are links to background: https://www.nytimes.com/interactive/2020/07/23/magazine/climate-migration.html; https://www.propublica.org/article/climate-change-will-force-a-new-american-migration; https://www.science.org/content/article/use-too-hot-climate-models-exaggerates-impacts-global-warming; https://www.nature.com/articles/d41586-022-01192-2; https://issues.org/climate-change-scenarios-lost-touch-reality-pielke-ritchie/;