What should you do? Perhaps you should do what you presently desire the most. Or maybe you should deliberate about what the best course of action would be independent of your desires, and work up the motivation to do that instead. These possibilities reflect the subjectivist and realist conceptions of value, respectively.
Or perhaps you should do something a bit more subtle. Perhaps you should think about what a perfectly rational version of you, free of your biases, would prefer. You could also imagine this perfectly rational version of you being in possession of all non-normative facts relevant to making any given decision, as well. This version of you would be able to reason perfectly and know all relevant facts. While you are not this being, you can do your best to imagine what this being would be like, and approximate the course of action you have reason to believe it would favor.
Such a possibility carves a middle path between acting on our present desires and acting on reasons that have no connection to those desires. This middle path appreciates the connection between normative value and our personal goals and values, while also recognizing that we can be misinformed or biased in ways that we would not on reflection wish to influence our decisions.
You may desire to go for a walk, but you presumably wouldn’t desire to go for a walk on a particular day if you knew you’d be hit by a car if you did. And you may desire to eat at the same restaurant again rather than try a new one, but you presumably wouldn’t if you knew if you’d only try the new restaurant you’d have an incredible experience. Sure, we generally want to act on our desires, but our desires can prompt us to pursue courses of action that are misinformed or warped by our psychological shortcomings.
If only we knew. If only we could reason in a perfectly rational, unbiased way. What would we choose?
That is the rationale behind this suggestion. And it may appear to be a compelling one. Of course this ideal version of you is a fiction. A fantasy. Not someone who could ever really walk the earth. None of us could be functionally omniscient. None of us could fully extirpate even the most minor of biases. But this fictional version of us nevertheless represents, so we may be inclined to suppose, a single, fully optimized version of us. A complete, or perfect version of us. A normative North Star around which our deliberations can gravitate. Sure, we can never reach this ideal, but we can at least do our best to approximate it. And perhaps, with time, we can asymptotically approach our ideal selves.
I don’t think this makes much sense. The whole point of the notion of outsourcing what you should do to this ideal version of yourself is to provide a single, definitive answer about what you should do in a given situation while preserving stance-dependence. But the very notion of a perfectly rational version of you, or a notion of all the relevant facts, as well as the very concept of an ideal version of you already rely on normative concepts. There are an infinite number of possible versions of you that think differently and have access to different information. Which ones count as the perfectly rational ones? Which ones count as the perfectly informed ones? Answers to these questions will inevitably turn on normative considerations. And how do you determine what the proper normative considerations are? You already gave up appealing to your own subjective standards and to stance-independent standards, so your only option would be to appeal to the standards of some ideal version of yourself…but that’s the very person whose standards you’re trying to figure out in the first place.
Furthermore, while the hypothetically singular, ideal version of you is intended to be a fiction that cannot be realized in practice, that we cannot become such a person is a practical limitation, not a conceptual or logical one. It is at least possible in principle to occupy the point of view of the ideal version of ourselves. If not, it would make no sense to try to approximate such a being’s point of view in the first place.
But I’m not convinced an ideal version of us is a coherent notion in the first place. Our values are shaped not simply by particular epistemic practices and non-normative facts about what the world independent of ourselves is like, but by our predispositions, personalities, inclinations, and experiences. And these simply cannot be eliminated from the picture through a process of “debiasing” without stripping us of any coherent sense of self in the first place; what distinguishes us from one another is partially constituted by our eccentricities. One might suppose that we could strip all of these away and locate some core self distinct from the vagaries and idiosyncrasies of each individual, and perhaps if we did we’d find the same foundation present in the mind of every agent. If so, perhaps we all share the same ideal version of ourselves.
I don’t think so. The notion of some abstract, undifferentiated, primordial “us” that would emerge if we eliminated our personalities, quirks, and experiences isn’t some fanciful fiction we could imagine but never arrive at in practice; it is a non-entity. Strip away everything that makes us us and you have nothing but a nervous system with only the rawest and most basic drives: hunger, thirst, fatigue. Even if we could sand away everything that makes us unique, identify some undifferentiated core to our agency, then imagine such a being knowing everything relevant to making decisions, why suppose it’d want anything in particular at all? Would it want a family? Would it want to see a movie? Would it enjoy music? Which music? And why would it enjoy that music, as opposed to something else?
No, such a being would be so alien that we face two questions. First, how could we possibly know what such a being would favor under ideal conditions? It is so unlike us we’d have little to grasp when it comes to reasoning about what its conclusions would be. This epistemic gap is, I believe, insurmountable. We simply lack the conceptual tools to make any reasonable judgments about what such a being would favor.
Second, it was already a stretch to suppose we would care what an “ideal” version of us would favor. Now we’re supposed to imagine an ideal version of us that lacks the many features that make us unique. Our prejudices, our biases, our eccentricities, everything that makes us unique isn’t merely some distorting factor; it is a push and pull, a response to stimuli, an impression, a scar, an internalized emotive response, a barely remembered trauma, it is all of this and more, it is the very assembly of experiences and memories that makes us people to begin with. If I’m to jettison all of that, then imagine what I’d want, I’m left doing little more than scratching my head. Can I even imagine that? How would I know if I’d succeeded? And if I did, why, at that point, would I even care what such a being would want? It is alien to me in precisely the ways that matter to me.
Most importantly, this whole hopeless exercise rests on the notion that there is a single, ideal version of us, and this, in turn, relies on the notion that there is a single set of standards for evaluating which state of affairs we could be in that would be ideal. But to make such a determination would require adopting a privileged perspective independent of any one of us in the first place. There are countless combinations of standards or states of mind we could in principle consider unbiased, and just as many combinations of facts we could consider relevant; there is no single, privileged set of standards and facts in the first place; judgments about appropriately unbiased standards and relevant facts can only be made from within a particular vantage point.
The whole point of the exercise was to avoid such a view from nowhere; yet such a view remains operating in the background, implicitly. But suppose it isn’t. In that case, we’d be back where we started: do we appeal to our present standards of evaluation for what constitutes an ideal self? Presumably not! That’s what we wanted to avoid. Do we appeal to the standards of an ideal version of ourselves? Of course not! We’re trying to settle on a single standard for what would constitute an ideal version in the first place; we can’t appeal to such an entity to determine what that entity would be like without having already settled what it would be like.
There’s no way around this.
There is a simple explanation for this. There is no single standard for what makes something ideal or relevant. These notions are ineliminably bound up in our individual perspectives, and are themselves always and inextricably perspectival. I believe this has broader implications for philosophy, but the takeaway here is that it’s not simply the case that we can’t adequately approximate what an ideal version of ourselves would favor in practice, but that it makes no sense to suppose there is an ideal version of ourselves to begin with.
I think you're right that, insofar as an ideal observer theory is trying to avoid "view from nowhere" objections and also is committed to a single correct idealization, they've just recapitulated the problem at the level of determining what the ideal is.
A more useful approach to idealization is as something of a tool. And idealization like Rawls' veil of ignorance may have no special metaphysical status, but people may nevertheless find it extremely helpful for thinking about and negotiating political disagreements.
This goes well with Rikard A. Hjort's "Deathbed Fallacy" article, which also explores an ideal version of oneself, but in terms of prospection and retrospection.