Suppose a man were translated to a planet, the sky of which was constantly covered with a thick curtain of clouds, so that he could never see the other stars. On that planet he would live as if it were isolated in space. But he would notice that it revolves…” (Henri Poincaré, Science and Hypotheses)
Poincaré goes on to note that this man would, if he were observant, notice that a free-swinging pendulum — something akin to Foucault’s pendulum — gradually rotates.
Assume, for the moment, that this man looks around him, at the birds flying through the sky, the trees undulating in the breeze, the houses firmly rooted to the ground, and thinks that the planet does not–cannot–rotate. This is a commonsense conclusion to make. After observing this rotation, the thought-experiment man may conjecture that the planet does or does not rotate.
In the latter instance, he would have to revise some of his background knowledge in order to keep his insistence on the non-rotational status of the planet. For instance, he might say that free-swinging pendulums are subject to an unexplained phenomenon that is not detectable with current technology. Free-swinging pendulum-influencing-rays, known as ‘FWPI-rays’ for instance. He makes a presently-untestable addition to his knowledge in order to prevent revision: he makes an ad hoc adjustment to his theory of the non-rotational status of the planet.
An analogous ad hoc adjustment might take place if the man looks far off in the distance and sees that, while the planet is surely not flat, the mountains and valleys and rolling hills would, if averaged out, be a flat surface. When the man adopts the (quite reasonable at the time!) conjecture of a flat planet, he notes that whenever a lunar eclipse takes place, the nearby moon is fully obscured. In fact, whenever the planet’s shadow is cast on the moon during a lunar eclipse, it appears to be the shadow of a globe. In order to save his flat-earth theory, the man may attempt all sorts of ad hoc explanations for this contradicting occurrence. He might end up a member of the Flat Earth Society. He might argue that the shape of the planet is kept secret for some conspiratorial reason. Doctored photos from satellites or the moon, fake space rocks, fake space walks, and so on.
Of course, there has to be a motive for this cover-up:
The motive for ‘The Conspiracy’ is unknown (unlike the existence of ‘The Conspiracy’, no specific motive necessarily follows from the FET) and as such it is open to speculation. However, financial profit is the most commonly assumed motive. (Flat Earth Wiki)
Thus, if the man is willing to salvage his prized ideas from criticism, he is quite capable of all sorts of logical contortions to keep his commonsense belief from revision. He can argue away the results of an experiment by positing an error in the experiment, or a misinterpretation of the data, or a grand conspiracy. In fact, it very well may be the case that there was an experimental mistake, or a misinterpretation of the data, or a grand conspiracy.
In the former instance, he would have to drastically revise his view of his place in nature. This change would involve rejecting what his senses directly perceive. While it may be psychologically painful to reject his pet theories, this man might follow what the experiment, and not his direct perception, tells him. If he rejects the experiment and retains his pet theory, what reason is there for him not to adopt the flat-earth theory?
In summation, adopting ad hoc adjustments to salvage a theory from refutation does not, in principle, allow for adopting cases of refutation for reasons. For instance, if the man with the non-rotational theory should encounter the man with the flat-earth theory, how is there to be any sort of argument between the two of them? There can be no critical discussion. This can be generalized: if everyone should adopt ad hoc adjustments to their pet theories, there can be no critical discussion.
I make the methodological choice to favor a critical discussion over contemplative silence. This is why I do not accept the ad hoc choice to save theories from criticism. This does not mean that I think that all experimental tests are final. Far from it. Experiments may be mistaken. In that case, it is proper to ask whether or not there was a mistake, and perhaps conduct the experiment again with different equipment. Thus, I also follow the methodological rule to pay attention to experiments that can, in principle, be repeatedly conducted, rather than one-time events.