Correlation != Causation

One of the things I love most about science is hearing other people call science a “thing”. “Science says the planets are round”. “Science says vaccines reduce the spread of herd diseases”. While these conclusions are true, the way in which they are spoken belies some level of ignorance as to what science is.

We live in a society exquisitely dependent on science and technology, in which hardly anyone knows anything about science and technology.- Carl Sagan

Science is a way of thinking. It is a process that has been developed over generations to support the human endeavour to figure out how things work. The scientific process is not perfect and it certainly has led to some incorrect conclusions from time to time. But a big strength of the scientific process is that it is self-correcting. If you’ve ever spent any time in the open source community, you know how it works. The same organized scepticism and peer review that keep open source projects churning out good code are the same facets of the scientific process that keep good ideas flowing.

This leads me to my point. Because science is a process, it must be taught. Most people do not have any training in the cornerstones of the scientific process which are the ability to apply reason and the desire to think critically about ideas. Therefore, most people will accept virtually anything as evidence. Nowhere is this more apparent than in people’s insatiable appetite for complex information boiled down into simple graphs.

Graphs show the correlation between two or more things: “How much do I weigh as I age?” “How many cars are red over time?” What graphs do not show is causation: “Ageing causes people to weigh more.” “The spinning of the world produces red cars.” Everything that you can think of can be related in some way and reduced to a couple of lines on a graph but that does not mean thing A caused thing B. The common mistake of people who do not take the time to reason things through is that they conclude that the correlation of two things must mean one causes the other.

The most entertaining example of putting two unrelated things together on a graph in an effort to correlate them has got to be Tyler Vigen’s effort to build Spurious Correlations. That is, correlations that have no causal effect on each other, but seem to when plotted together on a graph.

Take a run through. It’s a good laugh. And a few of them actually seem legit.