I received a reader comment regarding my last blog which reviewed the book Invisible Women by Caroline Criado Perez.
I am interested on the thinking of those who hold a different opinion on this matter. One concern in academic circles has been the use of identity politics to drive conclusions. The consideration of identity, some might be tempted to argue, drives studies, social policy, politics, and, as of late and possibly, science itself. I like the idea of blind studies that produce data that can be replicated. I’m concerned that our society is moving away from data-driven conclusions towards identity-driven conclusions. I’m not saying such is the case here, but if that approach is accepted, then we may end up harming all demographics in a society.
It took me a while to figure out how to reply. My commenter obviously believes that data-driven decisions should be used instead of the fashionable identity politics. However, my blog was not about identity politics, it was about how bad data created bad decisions.
My friend is concerned because the arguments within Invisible Women align with identity politics, in this case, pro-female politics. My friend isn’t anti-women, but clearly is concerned about ideologies that assume biases are present in society and promote a different set of biases in a purely ideological fashion.
I don’t try to defend Ms. Perez on those grounds. She might be pro-female and anti-male on ideological grounds. However, Ms. Perez impressed me because she points out the skew in the data. She recognizes the very scientific fact that bad data undermines the validity of data driven decisions – often with a high impact on our lives.
She points out that current car test results can only tell us how the average male driver and average teen male passenger will fare in a car crash. Drugs sold to women can only truly be declared safe for the males upon whom those drugs were tested. Our ideal of blind-studies driving data-driven decisions has been undermined because of the assumption that a specific population, in most cases, an average male, represents the population as a whole.
This isn’t new. As outlined in episode 226 of the podcast 99% Invisible, these biases are profound and deadly. When the first statistical averages were obtained from data (Chest measurements of 5,000 Scottish soldiers), the mathematician ideologically proclaimed the average measurement to be the platonic ideal, and anyone with a different sized chest was a deviant. This philosophy drove the decision of the Army in 1926 to design their airplane cockpits for the average 1920’s male.
After the Air Force split off from the Army they found that even improved training could not stop deaths during training. The army contracted researchers to measure the size of the 1950s airmen (no women yet), to see if men had simply grown larger and the cockpit needed to be adjusted.
One researcher, Gilbert Daniels took the research one step further. After measuring thousands of airmen on ten physical dimensions, he then decided to see how many airmen fit the ‘average’ model. Not one. Even when just comparing three dimensions out of the ten, only 5% of the pilots fit the average!
The data driven decision to fit cockpits to the average person created a plane that did not fit a single person in the Air Force and killed pilots. The improved data driven decision recognized that people needed adjustments to fit their own specific case and created an adjustable cockpit that saved lives. We can thank this data for creating adjustable seating in our automobiles as well!
I also believe in data-driven decisions over any ideology. However, we must not fool ourselves into believing that every data set provides the information we need to make the decisions we want to make. If we collect physiological and biochemical response data from males, then we can only legitimately make decisions about males. Women are physiologically and biochemically distinct enough that the results may not apply.
Sadly, while my friend is concerned that society is moving away from data-driven conclusions, we have too much evidence that ideology has skewed the data and driven incorrect decisions in the past. It seems to me, my friend should be grateful that the ideology has finally risen to the surface. Purely ideological decisions, like many within the now-disbanded Soviet Union, will eventually collapse under the weight of reality. The question is how long it will take for reality to smack the idealists in the face…
Perhaps the questions in regards to the text that may be more pressing involve the concept of generalities. Can we make generalities or general statements? Do averages (or medians) help us understand the world around us, or do these numbers only create dangers to our impressions by presenting evidence that actually skews perception. Right now, it seems that society moves away from drawing conclusions from generalities. I often see the presence of the logical fallacy of burden of overhasty generalization. A single case is presented to discount a generalization that may have some validity. Also, in full disclosure, I cannot comment intelligently on the reviewed text (nor am I) because I have not read it, so these comments I am generating here are free flowing and do not apply to the reviewed text. Such is true with earlier comments made. They are more of a reflection on what I have seen around me. I think if we cannot draw conclusions from averages and data sets (which may in fact be the case), then we may find ourselves wrestling with a certain kind of paralysis of judgment. Such a situation also could disadvantage a society. Whatever conclusions we do draw through averages and data sets will, it seems to me, always have outliers and exceptions. If that is the case, should we refrain from drawing any conclusions?
Time to blog!