Last february, 26th we discovered that Donald Rumsfeld, USA secretary of Defense had definitively closed the Office for Strategic Influence, apparently after an issue of the New York Times that revealed the disinformation plans that the already mentioned office planned to put into practice, including the delivery of false information to periodical publications outside the USA. This news made me think about the role of lies or, if you prefer to say it the subtle way, disinformation in connection with visualisation.

Edward Tufte devotes a whole chapter of his excellent book “The Visual Display of Quantitative Information” to “Graphical Integrity”, where he disassembles some of the tricks that allow you to lie subtly with graphics. For example he defines a “lie factor” as the ratio between the size of the visual effect that the graphic depicts and the size the effect has in reality. 

Representation of the diminishing purchase power of the dollar Apeared in the Washigton Post, issue corresponding to October, 25, 1978. Colour reconstruction, adapted by the author. (Click on the image to enlarge it.)

As a classical example, the chart showing the shrinking purchasing power of the dollar between 1958 and 1978 appeared in the October, 25, 1978 issue of The Washington Post. The chart shows 5 one dollar bills each of them with a length proportional to the purchasing power at its time ( see the adaptation of the chart) . 

In this case, while the variation of the purchasing power varies in just one dimension, the visual effect is the one associated to the area of the bill, that is two dimensional. This way the smallest bill, one 1978 dollar, is worth 44 cents of a 1958 dollar. Its length is, correctly, 44% of the biggest one that represents a 1958 dollar. 

Nevertheless its area is just 19% of that of the biggest one. This gives the impression of a dramatic reduction of the purchasing power. In order for the area to reliably represent the value, the size of the 1978 bill should be about twice as big as that shown.

Howard Wainer, in his book “Visual Revelations” develops 12 rules in order to represent data badly, that are based on the negation of the 3 principles that define a good graphic. Since a good graphic shows a lot of data in little space, in a precise and clear way, if you want to disinform:

• Don’t show the data (in any case show only some of it).
  
  
• Should you need to show some data, do it the inaccurate way. 
  
  
• Obfuscate the interpretation of the data.
  

The twelve rules that follow from these principles are (with some literary license…):

1. Minimise the data density (show little data in a lot of space). 
   
   
2. Minimise the data/ink ratio (little ink for data and a lot for accesories, axis, grid, decoration...).
   
   
3. Ignore the visual metaphor (the magnitudes don’t need to correspond with their representation).
   
   
4. Maximise the lie factor (no comment).
   
   
5. Graph data out of context (why show the part of the graphic when the business ran better?).
   
   
6. Change the scale in mid-axis. Nobody will notice it and you can change the behaviour of your data.
   
7. Emphasise the trivial and distract the attention from what’s important.
   
   
8. Make comparison difficult, in the end what graphics show are differences, hence don’t help others to notice them.
   
   
9. Order alphabetically, for sure it will be far more difficult to detect patterns than when you order them according to increasing values of some of the variables.
   
   
10. Label as badly as you can: illegibly, incompletely, incorrectly and ambiguously, it’s very helpful in confusing the audience. 
    
    
11. Add useless information: more decimal places and more dimensions than needed.
    
    
12. If that chart worked in the past, think of a new way to do it, almost certainly you can find a worse graphic.

Obviously, turning the twelve rules upside down helps, although doesn’t guarantee, to make more accurate and clearer graphics that reveal the data and its differences. On the other hand it helps to stand the impact of the image and to build a critical turn of mind that could allow us to detect the tendentious and confusing graphics that populate many presentations that too many times we accept quite happily.

Visualisations are not neutral, they depend on our selection and on what we want to show the others and to ourselves. Cultivating this critical turn of mind regarding what we do and what others show to us is vital to avoid confusion.

Three amusing and interesting books that are worth reading:

How to Lie with Maps  by Mark Monmonier

How To Lie With Statistics   by Darrell Huff

How To Lie With Charts   by Gerald Jones

Links of this issue: