In 1976 a statistician, George Box, stated “All models are wrong, but some are useful."

So, what does this mean? It means that a model is only a representation or simplification of reality. If it were reality, it would be reality and not a model. The amount of model “wrongness” is a matter of degree. The real question is, how wrong do they have to be to not be useful. Put another way, how well does the model reflect reality? For some models we may never really know how closely they reflect reality, and for some we have a pretty darn good idea. For instance, remember when the Google map cartographers had it all wrong and the Google map app kept sending people in the wrong direction? Those models were so wrong they weren’t useful.

Generally, in the hard sciences, models are less wrong than models in social sciences because they do and can ignore less (i.e., fewer variables or factors) than we do. As social scientists, we have to ignore more because we often cannot randomly assign participants to treatment groups (i.e., job training programs, educational programs, HR benefits, performance bands, etc.) because it is not only impractical, but it is unethical. So, our models in social science may be more wrong, but they are still practical. Our models still help us to explain, predict, and shed insight into our areas of interest. We have to keep Box’s sentiment alive when interpreting our data and when making data-driven decisions.

For each project, ask yourself and your stakeholders, “What degree of wrongness am I willing to accept?”

A Logical Proof that Santa Exists!
  • No comments found
Add comment