Strevens (2003) explicitly does not prove Priority achieves the optimal allocation in the winner-confers-all case — only that it moves allocation in the efficient direction

philosophy-of-sciencepriority-rulemichael-strevensresource-allocationscope-limitdivision-of-cognitive-labor

Strevens's proof of strict optimality — an exact formula equating marginal return functions across programs (m1(n) = m2(N−n)) — is derived only for the additive case, where it holds for the Marge scheme. (The paper's §4.1 does state an exact condition for the winner-confers-all optimum too — equal "adjusted marginal return functions" — but that characterizes the optimal allocation itself; no reward scheme is proven to attain it.) For the winner-confers-all case that actually models scientific priority races, he declines the optimality claim outright: "Whether Goal or Priority achieves a more efficient allocation in a winner-confers-all race is not, however, a question that I will be able to answer here, for reasons to be explained in section 5.1." He reiterates the limit for the "Arago effect" comparison — showing the strict system superior "would involve some substantive quantitative assumptions about the degree and form of human risk aversion" that the paper does not supply.

So the paper's argument has two distinct strengths: (a) a rigorous proof that a different, non-priority scheme (Marge) is optimal in the additive case; and (b) a qualitative, comparative argument — not a proof — that moving Marge → Goal → Priority shifts allocation in the direction the winner-confers-all optimum requires.

This note exists to fix how strongly the vault may state Strevens's central thesis. Supported: "Strevens argues priority moves allocation toward efficiency." Overstated: "Strevens proves priority is the economically optimal scheme." The capture's originating question framed the rule as an "economically optimal allocator"; the accurate answer keeps the optimal attached to the additive Marge proof, not to Priority in a real race.