Back when I was in graduate school, a very long time
ago, I attended a meeting or two of the local university peace group,
whose membership was predominantly liberal faculty members. I only attended one
or two meeting because, while they were all for world peace and getting
everyone to live together in harmony, they were engaged in a bitter battle
among themselves as to what to name their group. I reasoned that if they
couldn't keep the peace among themselves on a trivial subject, they were
unlikely to do anything effective about peace in the wider world.
I thought of this again as I read a piece entitled The
Social Science Monoculture Doubles Down by Keith Stanovich, a professor
emeritus of Applied Psychology at the University of Toronto, who turns out to
have authored a number of good books, the latest of which is The Bias That
Divides Us: The Science and Politics of MySide Thinking, published this
year by MIT Press.
Stanovich's point, discussed in great detail in the
article (this is not an easy, quick read), is that social scientists, almost
exclusively ultra-liberal academics, are producing studies that are
seriously technically flawed, and heavily biased toward what they want to
believe, studies which the media then picks up and trumpets, especially if it
fits the stereotypes they believe in.
Doing good science is hard. Nature is subtle,
experimenters get their egos wound up into their theories, good experimental
design is difficult, and the interpretation of the results is always subject to
the experimenter's expectations and biases (and too often, their poor
understanding of the limits of statistics). Indeed, several recent studies have
found that well over half the studies reported in the literature cannot be
replicated, even in the hard sciences. In the social sciences this figure rises
to the 70-80% level.*
So under the best of circumstances it is always wise
to be skeptical of the result of studies. This skepticism isn't
"science denial". It is, in fact, exactly what good science is
supposed to do. It is part of the process that eventually weeds out poor
studies or fabricated results or misused statistics. Stanovich's article is a
good introduction to how this happens.
----
* See, for example The
Economics of Reproducibility in Preclinical Research by Freedman, Cockburn
and Simcoe, which estimates more than 50% of preclinical studies are not reproduceable.
Or the article Estimating the
reproducibility of psychological science. Science, Vol 349, Issue
6251 (28 August 2015), which shows overall about two-thirds of psychological studies
reported in major psychological journals can’t be replicated.
Posts
