Friday, September 25, 2020

Mono-causal explanations

This is Joseph

One thing that drives me a little nuts is how people want to have things happen due to a single cause. In most of life, things are never so simple. Mark noted this in a previous post. Many things can contribute to an event occuring, in a combination of necessary and sufficient causes.

The triggering event for this was a tweet on oral anti-diabetic medications suggesting that diet and lifestyle should be tried. I think that everyone prescribing medications would prefer that lifestyle changes would be wonderful. But the real world is complicated. For example, if it is Type 1 diabetes there are no lifestyle changes that will help -- before the development of insulin this was a certain death sentence.  Even for Type 2 diabetes, where the medications are most likely, it will not always be the case that lifestyle is sufficient to cause remission. Instead a combination of lifestyle and medications, together, are likely to be more effective than either alone. 

But this concept applies to all sorts of other conditions. For example, when a person leaves a job there are almost always a series of causes. Some of it could be salary whereas location or family could matter more for others. As a result you can never infer anything about a workplace from a couple of people leaving, but a string of attrition looks like a bad sign. 

As Mark noted, this also comes into play with all sorts of political systems. People like single explanations for why a specific person wins a race. The truth is that there is probably a wide range of motivations for specific votes -- ranging from random chance to fierce interest in a specific policy. 

But more important, we should discipline ourselves not to conflate risk factors with causes. Obesity might increase the risk of a deep vein thrombosis, but many non-obese people have these events. The counter-factual person (same person just not obese) might well have had one regardless of obesity. 

Looking for a simple explanation always suggests you are missing important information. 

1 comment:

  1. My favorite phrase on this is "piss-poor monocausal social science," from Daniel Drezner: