Thursday, May 23, 2013

Simplified is not necessarily better -- the tyranny of the average

There is a strong tendency in our society to try to make things "simple". I don't know whether it is because we are so time rushed or because we have such an imbalanced system of education. Sometimes having it simplified works to a specific person's benefit -- sometimes it is quite unfavorable to a specific person.

Simplification is often linked closely with statistics. As Mark Twain is quoted as saying -- "there are liars, damned liars, and statisticians". Statistics are applied math so they should be accurate -- however, the actual use of the formulas and numbers are decided by people and people make decisions based on what they want to have be true (either consciously or subconsciously).

One example of this are the figures quoted by politicians during election time. "The average income tax went down during my administration". Hmmm. Well, this statement could be "true" if the average income also went down during the time. It could be "true" if the tax rates went down (this is the interpretation the politician probably hopes you will make). It could be "true" if tax rates went down for one segment of the people (usually the high-income group) but went up slightly for other segments (and this has been happening in the U.S. for the past 12 years or so).

Three different realities -- all "supported" by the same "facts".

One area that hits hard for me is the Body Mass Index (BMI) number. The BMI is a fast, easy, "simple" method to indicate whether a person is overweight. It is reasonably accurate (+/- 5% or so) for about eight out of ten people. Who are the other 20% of the population? They are people who are particularly tall (more than 15% above average) or short (20% or more less than average) or people who have large amounts of muscle tissue -- yes, the "fit" are most at risk from inaccuracies on the BMI.

All of this would be academic if it wasn't so easy to fit these simplistic numbers into other formulas -- such as actuarial tables used by insurance companies. So, if you happen to be a body builder then be prepared to pay more from private insurance companies (on life and health) for being "too fat". If you are very short, then you can be rather overweight and still have the insurance count in your favor. If you are very tall, however, then you need to be prepared to pay more once again. Oh -- and I forgot to add -- computerized dating systems tend to like to use BMI in their calculations so expect body builders to be matched up with others of ample dimensions.

The reason the BMI is used is that it is easy and cheap -- take your weight (in kg) and divide it by your height (in meters) squared and you have a handy, dandy, all-purpose number. In order to really find out an accurate number, it is necessary to find real body fat percentage. There are formulas that measure different areas on one's body and then use those in conjunction with weight to get a number that works for 95%+ of the population. But it takes more time and more time means fewer patients seen and that means smaller profits. It is also possible to do a submersion test (where your body is submerged into a tank of water to accurately determine volume) that is the most accurate way to measure density (weight divided by volume equals density).

These are two methods where the average can hurt those who don't fit. There are many others. But do your best to understand the implications of statistical statements -- it may not mean what it seems.

User Interfaces: When and Who should be designing them and why?

     I am striving to move over from blogs to subscription Substack newsletters. If you have interest in my meanderings please feel free to ...