It took me a while to get used to the idea of 'experiments' in behavioral economics. My idea of experiments is as follows:
(i) It should be possible to control parameters in an experiment (ii) The 'unit' under experiment should not know that there is an experiment going on (quantum mechanics seem to suggest that this may not be possible - "the photons know" - but that is another story) (iii) The experiment should be repeatable (iv) Most importantly, the results should not be known before the start of the experiment. (v) And finally, once an experiment is performed and results concluded, every subsequent similar (but not identical) experiment should be predictable. (A result of such experiments that does not behave as predicted becomes the basis of the next hypothesis).
The experiments in Predictably Irrational, seem to be following the above criteria, more or less. So over a period of time you get used to the word experiment being used in not-strictly scientific context.
But athere is one fatal flaw, as I see it. One vital parameter has been tampered with. In many of the experiments, an economic gain is made without any cost to the particpant. What is economics without cost? Dan Ariely goes all out to prove that humans behave irrationally when encountering the word FREE! and then proceeds to perform experiments where the objects of experiment get easy money or goods that are free. Doesn't that make experiments unrealistic? After all you are trying to predict human behaviour in normal circumstances.