They don't all have to be played by Hugo Weaving, but agents surround us. In his Theory of Moral Sentiments, Adam Smith took pains to describe the characteristics and the benefits of sympathy, a virtue I find inextricably linked to the business of being human. Our sympathy guides us to craft principal-agent relationships whether they pass a cost-benefit analysis or not.
Planning demands sympathy. It's an easy sympathy, a simple matter of putting yourself in your own shoes some time hence, but still an act of sympathy, similar in kind to the sympathy we feel for others, but perhaps blessed by higher-quality information. Recall from Buchanan in Cost and Choice that all costs are opportunity costs and that these costs are incurred at the moment of choice. That the benefits may be realized by a future self does not imply that the costs are not borne by that person's agent, the present self.
I can agree that it's sort of silly to imagine myself as my agent when I order the spicy shrimp at lunch. Yes, I'm making an informed guess about what I will want to eat in fifteen or twenty minutes, and while that is technically a future version of me, but one for whom the cost-benefit calculus is still likely to land in a pretty tight distribution around spicy shrimp. The story changes a bit when I'm making allocation decisions about my retirement, about continuing my formal education, about whether to have more kids, about any big decision that has a substantial influence over the endowment of my future self. Regular, well-ordered feedback helps me reduce my bias over lunch, the lack of such feedback in big decisions could be a problem. Once bitten, twice shy, yes, but mistakes at a Thai restaurant are heel nips from a schnauzer. Commit a systematic error in choosing a graduate field of study, and you got gnawed on by a deepwater shark.
It's quite natural to want to avoid being a snack for a mako. What a pity it's so often difficult to check the water while we're bobbing around in it.
Given our ample practice at being agents for ourselves, and to those of us over whom we retain moral authority (dependent family, incapacitated friends--I have a suspicion that at least one of my readers has felt responsible for making sure a drunk friend makes it home safely), it doesn't feel like too much a stretch to transform far-mode sympathy for the mental mannequins we craft as stand-ins for the poor, the addicted, the hungry, the cold, the oppressed, the desperate into concrete moral agency. It's not wrong to make decisions on behalf of my future self, and it feels roughly the same to imagine my future self as it does to imagine a future other, so by the transitive property of moral legitimacy, it feels appropriate to make decisions for others on their behalf. If this is underpinned by lay interpretations of rationality, remaining qualms can be soothed fairly quickly.
Agency is pretty easy to justify when there's real skin in the game. If I'm wrong about the regret aversion of either me or someone important to me, I'll be the one to eat that sandwich. If I'm wrong about drafting paternalistic policy, I can pass the buck in the event the median voter cares or just ignore it if she doesn't. There's a spillover risk with regard to paternalism based on regret aversion by proxy: I can get the glory for expressing a widely-held (and usually functional) moral virtue of sympathy, but I don't have to bite the bullet if it lands on its head. This can lead to the overproduction of paternalism. Constitutional limits on the scope of involuntary collective action can keep these spillovers in check contingent on the observation of the prescribed limits. It is an empirical question of to what extent these limits bind, especially given enough time.
Regulated by good institutions, sympathy elevates humanity. Perverted by grand ambitions and hubris, it quickly morphs into suffocating tyranny. Beware false agents.