Suppose two people are trying to quit smoking. They both want a cigarette, but they wish they didn’t want a cigarette. So there is a difference between what they want and what they metawant. Now, suppose one of the two people smokes a cigarette. Does that mean that the person who smoked had less willpower?
If somebody does not do something that they have a first order desire to do and a second order desire to abstain from, this could be viewed as an example of willpower. However, we do not know how strong their first order desire was. In the above example, perhaps the person who smoked had a much stronger first order desire to smoke, but they both had an equal second order desire to not smoke. Did the smoker really have less willpower?
Denote by W (for ‘want’) the level of first order desire that a person has to take some action. Denote by MW (for ‘metawant’) the level of second order desire to not take the action. We could roughly think of the decision to take the action or not as determined by whether MW>W.
What we observe is whether the person took the action, i.e., whether MW exceeded W. But we do not observe MW and W.
If the person who successfully quit smoking, or if the skinny person who doesn’t overeat, brags about superior willpower, it seems to me that they are inferring (assuming) something about other people’s MW and W, even though they only observed whether MW exceeded W.
This post’s conclusion contradicts evidence that willpower is a generic (multi-purpose), depletable resource, unless you want to say that any time a metawant triumphs over an object-level want, all other metawants are temporarily decreased relative to their respective object-level wants.