"The French intellectual says: 'Sure, it works in practice... but does it work in theory?'"
The joke is that some people, usually French, have no interest in the real world and prefer academic houses of cards.
However, the French intellectual's question is actually quite legitimate. Consider the Lucas critique, Wikipedia's description is quite good:
"The Lucas Critique says that it's naive to try to predict the effect of a policy experiment based on correlations in historical data, especially high-level aggregated historical data. For example, Fort Knox has never been robbed, but that doesn't mean we can safely eliminate the guards, since the incentive (not) to rob Fort Knox depends on the presence of the guards. If we do want to predict the effect of a policy experiment, we must model the "deep parameters" (preferences, technology and resource constraints) that govern individual behaviour. We can then predict what individuals will do conditional on the change in policy."
Many theories may appear to work in practice. A great example is the Galbraithian explanation of the Wall Street Crash. It took an examination of the theory by a host of economists, most notably Milton Friedman, to reveal its vulnerability.