All great points and nicely organized and clearly explained. Two expressions that captures concerns about data's provence, interpretation, and selective focus, is that: "data is never raw" and "data are made, not found".
My reflection is that each point is essentially about applying the scepticism you mention in the first rule. At each stage you're testing out whether the chart could, in fact, say something different.
Also love the idea of using questions in titles to be more equivocal whilst still being engaging.
I agree with all your 10 points! And I really love how many of your points focus on the data collection, processing and initial analysis. I think we talk about the final aesthetic choices far too much, and often forget the huge iceberg of choices that went before anyone even tried to plot the data.
I also like the list format. I can imagine turning this into a checklist you could run through for each viz you make.
Great points Enrico. What's your viewpoint on including an item on 'conforming to visual perception standards and customs'? This is at least in part covered by Scaling Visuals Mindfully (e.g. truncated axes), but my point goes beyond cut-off bar charts. It also entails selecting appropriate colors, and designing the graph in a natural order (e.g. I've seen graphs with time axes running from right to left). Of course, sometimes one wants to introduce a novel visualization style/type, so it again is a balancing act.
This is a fantastic observation! You might have noticed that I have kept the guidelines on purpose generic enough to include many particular cases. Your comment made me think about how to introduce another kind of problem that pertains to the visual aspect, which is about using appropriate affordances. It can also be described as avoiding mismatches between data semantics and visual expression. I have several examples where the problem stems from using inappropriate visual mappings. One is using unidirectional color scales for diverging quantities. Another is using ambiguous axis directions. There are probably many more. What do you think? Would that capture your ideas here? (Thanks for commenting! It's really useful to hear what my readers think! 🙏)
Yes exactly this. For example, I know a relatively innocent graph which presents Ireland's position in the Olympics Medal table.. Using red and a downward sloping line chart (which relates to a lower number and hence, a 'better' position in the medal table) that would make an Irish reader probably first think things are deteriorating ("ireland olympic medal line" visual). Multiple affordances indicating something different from the key message. There are more pressing examples where time axes are inverted (e.g., "violent crime is a big problem" visual), or y-axes are inverted (e.g., "Florida stand your ground" visual) to show an 'apparent' drop.
But again this is a case of "How to spot a misleading one", instead of how to be truthful. Like you mention: for truthfulness, it really boils down to using appropriate affordances.
Thanks so much, Frank. These are very useful. One question is how other visual properties have similar effects. Most issues of this type in charts stem from improper axis use. Of course, the visualization literature is full of examples of ways in which certain design choices create ineffective charts, but most of these choices tend to lead to inefficiency or confusion, rather than misleading messages.
All great points and nicely organized and clearly explained. Two expressions that captures concerns about data's provence, interpretation, and selective focus, is that: "data is never raw" and "data are made, not found".
True!
Excellent list!
My reflection is that each point is essentially about applying the scepticism you mention in the first rule. At each stage you're testing out whether the chart could, in fact, say something different.
Also love the idea of using questions in titles to be more equivocal whilst still being engaging.
I agree with all your 10 points! And I really love how many of your points focus on the data collection, processing and initial analysis. I think we talk about the final aesthetic choices far too much, and often forget the huge iceberg of choices that went before anyone even tried to plot the data.
I also like the list format. I can imagine turning this into a checklist you could run through for each viz you make.
I like the checklist idea! I have been thinking about it too!
Excellent post; interesting, many constructive ideas.
Thanks Stephen!
Great points Enrico. What's your viewpoint on including an item on 'conforming to visual perception standards and customs'? This is at least in part covered by Scaling Visuals Mindfully (e.g. truncated axes), but my point goes beyond cut-off bar charts. It also entails selecting appropriate colors, and designing the graph in a natural order (e.g. I've seen graphs with time axes running from right to left). Of course, sometimes one wants to introduce a novel visualization style/type, so it again is a balancing act.
This is a fantastic observation! You might have noticed that I have kept the guidelines on purpose generic enough to include many particular cases. Your comment made me think about how to introduce another kind of problem that pertains to the visual aspect, which is about using appropriate affordances. It can also be described as avoiding mismatches between data semantics and visual expression. I have several examples where the problem stems from using inappropriate visual mappings. One is using unidirectional color scales for diverging quantities. Another is using ambiguous axis directions. There are probably many more. What do you think? Would that capture your ideas here? (Thanks for commenting! It's really useful to hear what my readers think! 🙏)
Yes exactly this. For example, I know a relatively innocent graph which presents Ireland's position in the Olympics Medal table.. Using red and a downward sloping line chart (which relates to a lower number and hence, a 'better' position in the medal table) that would make an Irish reader probably first think things are deteriorating ("ireland olympic medal line" visual). Multiple affordances indicating something different from the key message. There are more pressing examples where time axes are inverted (e.g., "violent crime is a big problem" visual), or y-axes are inverted (e.g., "Florida stand your ground" visual) to show an 'apparent' drop.
But again this is a case of "How to spot a misleading one", instead of how to be truthful. Like you mention: for truthfulness, it really boils down to using appropriate affordances.
I'd love to see these examples if you can find them! Thanks for suggesting this idea. It's very valuable.
Ireland medals: https://www.datawrapper.de/blog/irish-times-chart-redesigned-olympics
Violent crime is a big problem: https://www.reddit.com/r/assholedesign/comments/o6r4ek/this_cnn_graph_on_polling_about_violent_crime/
Florida stand your ground: https://www.livescience.com/45083-misleading-gun-death-chart.html
Thanks so much, Frank. These are very useful. One question is how other visual properties have similar effects. Most issues of this type in charts stem from improper axis use. Of course, the visualization literature is full of examples of ways in which certain design choices create ineffective charts, but most of these choices tend to lead to inefficiency or confusion, rather than misleading messages.