PROBLEMS OF PREDICTION and CAUSE
by Phil Bartle, PhD
Prediction is the main way of testing the validity of a theory in science
Much prediction is backwards prediction (in time) for social sciences, because the events may have already occurred, but we look at other variables to see if they appear to have some influence on the variables we examine and want to predict.
When there are many variables which jointly act on the behaviour of something, it is impossible for us to differentiate (don’t you just love that big word?) between them.
This is not only true for the social sciences, but even for physics.
We may know all the relationships of force that can affect a single leaf –– gravity, wind, location –– yet not be able to predict the exact path that a leaf will take when it falls of a tree and moves toward the earth.
We are a little better at statistics and predicting that, when many leaves fall, they will form a bell curve in a ring around the tree –– so long as the wind is calm.
Our concept of "causality" is fraught with problems.
It is a bugbear to all scientists, including social scientists.
We can examine data from two variables and see that they vary in association with each other, but there is no logic or observation that will confirm that one causes the other.
When we say a change in variable "a" causes a change in variable "b," we mean that a change in "a" is sufficient and necessary to change the variable "b."
Easier said than done.
Also see key word: Cause.
If you copy text from this site, please acknowledge the author(s)