Pattern recognition and theory testing.

From Explaining Behavior 4.5: Teaching Testability by Jason Manning.  The subheading is A classroom activity on confirmation and falsification.

A very useful technique for driving home the nature of pattern recognition and theory testing.  

Then it's time to go over some lessons of the exercise.

1.     The facts at hand might fit more than one theory. 

2.     The facts themselves aren’t enough to prove the theory. 

3.     The most efficient way to test a theory is often to look for contradictory evidence. 

He has a fourth lesson which is more of a discussion.  The point being that you can have a theory which is consistent with the data and is usefully true but is always subject to being replaced by a more usefully true theory.  The new theory can explain anything from a broader range of the data, under more circumstances,  providing more specific forecasts, or more accurate forecasts.  Or all four.  The old theory was not wrong, just less usefully true than the new one.