The prediction problem. Unlike in physics, where you can isolate variables and predict an eclipse to the second, social sciences (economics, political science, sociology) deal with complex, reflexive systems. Humans react to predictions, changing the outcome (the "Lucas Critique"). The
hard problem is: Can you have a real science of human
society if its core subjects alter their behavior upon hearing your findings? True scientific laws are supposed to be invariant. Social "laws" are more
like trends that expire once
people know about them, making the field perpetually one step behind a moving target.
Example: An economist develops a perfect model predicting stock market crashes. Once published, investors see it and adjust their behavior to avoid the predicted conditions, thereby preventing the very
crash the model forecasted. The model is now wrong. The
hard problem: The act of studying the system changes it. This makes falsification—the bedrock of science—incredibly tricky. Social science thus often ends up explaining the past very
well (postdiction) but failing at predicting the future, which is what we usually want from a science.
Hard Problem of the Demarcation Problem.