The need for predictions.

http://www.cs.utexas.edu/~kuipers/opinions/electrons-vs-fairies.html

Making good decisions relies upon a mix of luck, and the ability to predict the outcome of those decisions. (preferably more of the later). For knowledge to be helpful, it must improve the ability to make predictions. To determine if that is so, one must have testable predictions made from that knowledge. If the knowledge cannot produce testable predictions, then the usefulness of the knowledge cannot be known.