In the latter, wholly instrumental case, it is hard to see why the Patient's ex ante preference has moral weight for the Agent--even within a preferentialist and nonconsequentialist view of welfare.
Plausibly, the Agent has moral reason to advance the Patient's agency: either as one aspect of welfare (on a substantive rather than preferentialist account of welfare which makes agency one of various substantive values), or independent of welfare.
First, the substantive rather than preferentialist account of welfare is arguably correct.
If a substantive rather than preferentialist account of welfare is correct--and there are plausible reasons to believe this is the case (39)--the fact that the Patient prefers outcome [W.
But the convergence, crucially, will not occur except on a preferentialist view of welfare--a view that, as I have said, is problematic.
Given RU, the rationally approvable action for the Patient is necessarily the action with the best outcome for the Patient's welfare if, and only if, the preferentialist view of welfare is correct.