This has been edited in the light of a comment.
It isn't possible to believe Pareto improvements- i.e situations where a guy can report being made better off without anyone else reporting being made worse off- are desirable without also believing all agents are sovereign over at least one choice affecting themselves- i.e. a Paretian Illiberal, of the Amartya Sen type, is impossible. Assuming unrestricted domain, this becomes trivially true by introducing a notion of doxastic self-ownership- for e.g. by valorising the preference to have one's own preferences.
Suppose a Paretian Liberal is impossible- in other words, Amartya Sen isn't a shithead- then it would be the case that a particular sort of rational being would
1) not prefer to have his own preferences and, worse still,
2) voluntarily subscribe to his own logical impossibility.
However, to do so isn't a Pareto improvement because the said Liberal could always chose to believe Sen is a shithead and thus be rescued from the fate of being logically impossible in a manner that makes no one else worse off- provided Sen is either himself a Paretian Liberal (in which case he either disproves his own existence or, and this result holds even if he isn't a Paretian Liberal, that he has hit upon an Acceptation for 'Paretian Liberal' which is wholly meaningless and thus he is talking shit and has shit for brains- which is why he is a shit-head) or else has a meta-preference not to be a shithead (in which case he can mend his ways) or else is too fucking stupid to ever grasp that he's a total shithead, in which case his feelings in this regard can't be hurt.
Truly unrestricted domain with 'nosey preferences' (i.e. which feature impredicativity) means we have an uncountable infinity of pair-wise choices which can't be well founded and so Sen's proof fails immediately. But this is scarcely a surprise. Impredicativity will do that to you.
Suppose I have a choice to either eat or not eat this biscuit. Permit me to have meta-preferences- i.e. introduce impredicativity. I now can prefer to eat but prefer to prefer not to eat this biscuit (i.e. prefer to have ascetic preferences) or prefer to prefer to prefer to eat this biscuit (i.e. prefer to depass my ascetic preferences by reason of its sublation by a more refined notion of atraraxia) and so on and so fort. What is the upshot? Unrestricted domain with impredicativity of the type Sen stupidly imports (being insensitive to Arrow's Tarskian training) gives us something immediately bigger by a Godelian argument than its own acceptation (i.e. Unrestricted Domain can't mean whatever it was supposed to, thus no modal fixed points are available) and so, by Ackermann's Reflection principle, everything just got apophatic and nothing is well founded. But this isn't some wonderful new insight. It is simply childish.
A more serious reason to spurn Sen-tentious shitheadedness arises out of the manner in which 'Just So' representative individual type theories are wholly misleading.
Alan Kirman, explaining why mainstream Econ has fucked up so badly in recent years, explodes 'representative agent' models and asks us to shift our attention to Organisation, as opposed to Efficiency, as central to Social Science.
Kirman- a good guy who was once a Geography teacher- explains what is salient for Significs is that Impredicativity or Reflexivity does arise in Econ- which is why we can meaningfully speak of a Paretian Liberal- but this notion is only captured by interactions of an Organizational, i.e. second order, not Transactional, type. As a matter of fact, not theory, the Paretian Liberal does exist in all countries with the Rule of Law. Sen may say- boo to the Rule of Law, boo to Organization, boo to 'Niti'- but he can't deny that Paretian Liberalism is a better description of what obtains in Rule of Law, Democracies than his own Entitlement theory or Capabilities approach. The fact is, 'territoriality' or the 'bourgeois strategy in the hawk dove game' is evolutionarily stable and eusocial (vide Zahavi on the handicap principle) which is why it is the basis of Paretian Liberalism. True, this is not indefeasibly so but defeasibility is a good thing because, if Darwin not Dueteronomy is right, the future must always be, at least retrospectively as Popper and Dunnent and Huw Price have, albeit inadvertently, argued, not just unknown but radically unknowable. That last being, as Collingwood posthumously proves to his ageing Oakshotteian self, a distinction without a difference.