Being notionally correct beats false precision (Part 1)
We live in a complex world, so we can be excused for wanting to lean on data as much as possible when making decisions.
However, there is a real problem with this when 1) our data is unreliable, 2) the conclusions we are drawing from that data are non-sequiturs and 3) we are looking at outcomes one dimensionally.
Unfortunately, unless you are dealing with the sterility and perfect cause and effects of physics, this is much more often the case than most people assume.
Unreliable data: Our data is often unreliable either because our measuring techniques are imprecise or because the person recording the data is actively incentivized to manipulate it. This is much more frequently the case than one might think. It’s why we have a replication crisis in science. Mind you, this is not to say that most scientists are evil manipulators. They just swim in a sea of bad incentives in which certain practices have become normal. After all, a world of “publish or perish” in which some participants cheat, is equivalent to a world of road cycling in which some participants dope — not cheating or not doping simply isn’t an option.
Non-sequiturs: In a complex multivariate world it is very easy to draw the wrong conclusions from certain observations. Confounding variables are frequently overlooked and correlation confused for causation. This seems to be a particular disease in the social sciences.
One dimensionality: Experiments are great when we only care about one particular outcome and whether it occurs or not. They are more problematic, when there are many additional, potentially relevant effects that are not being measured. The best example of this is drug testing. Most drug testing limits itself to testing the efficacy of a drug as well as immediate side effects. It cannot look at potential longer term side effects of a drug. This is because such an experiment would simply require too many participants to deliver statistically significant results and would therefore be too expensive. It is the reason why pharmaceutical companies pay billions and billions in settlements for drugs that were once deemed safe.
The above is a summary of why simply trusting the science can lead to false precision and deliver sub-optimal outcomes at best. In tomorrow’s post, I will discuss an alternative approach.
