---
title: "Strevens's mechanism transforms a resource-based reward scheme (Marge) into an achievement-based (Goal) then priority-only (Priority) scheme, with risk aversion shifting labor toward higher-potential programs"
type: "claim"
status: "seedling"
source_url: "https://www.strevens.org/research/scistruc/Prioritas.pdf"
source_author: "Michael Strevens"
source_date: 2003
source_venue: "The Role of the Priority Rule in Science, Journal of Philosophy 100(2), pp. 55-79"
source_quote: "The second step of the transformation turns Goal into a scheme in which there is a priority race, so that, if two rival programs both achieve their goal, only the first to do so is rewarded. I call the resulting scheme the Priority reward scheme."
source_tier: 1
source_sha: "01d0207a3cc1c75d88ab022f67b60f1037c438b049c6a9e85d4562dbafa87747"
audit_status: "capture-verified (Strevens 2003 PDF read in full at capture; the three step-quotes below confirmed verbatim via quote_check at capture; queen's independent re-fetch not performed in this headless promotion — no network); 2026-09-08 cross-model audit (auditor claude-fable-5, writer claude-opus-4-8): all four body quotes re-verified verbatim against a direct re-fetch of source_url (sha256 match 01d0207a…); one defect fixed — the Marge-scheme quote had been truncated mid-sentence without ellipsis, dropping 'and yield decreasing marginal returns on investment'; quote completed in body (the omitted condition was already stated in the note's own paraphrase, so no substantive claim changed)"
provenance: "Promotion from 10-inbox/raw/2026-09-01-does-strevens-2003-actually-argue-the-priority-rule.md, 2026-09-05"
origin: "batch"
derived_from: "10-inbox/raw/2026-09-01-does-strevens-2003-actually-argue-the-priority-rule.md"
date_created: "2026-09-05T00:00:00.000Z"
writer_model: "claude-opus-4-8"
tags: ["philosophy-of-science","priority-rule","michael-strevens","mechanism-design","risk-aversion","division-of-cognitive-labor","resource-allocation"]
verified_verbatim: "2026-09-06 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
seek_code_commit: "a619c8a"
---


[[claim-strevens-2003-priority-rule-as-efficient-labor-allocator|Strevens's
economic account of the priority rule]] runs on a named sequence of three
reward schemes, each a transformation of the last:

1. **Marge** — a resource-based scheme paying every scientist in proportion to
   their marginal contribution to their program's probability of success,
   regardless of whether the program succeeds. Strevens proves (section 3) that
   Marge produces the *optimal* allocation of labor in the **additive** case,
   given decreasing marginal returns on invested labor: "there is a reward
   system, the Marge scheme, that will ensure an optimal allocation of labor
   among any set of research programs, provided that the programs have goals
   with additive values and yield decreasing marginal returns on investment."
2. **Goal** — Marge converted to an *achievement-based* scheme: "First, Marge's
   resource-based scheme is transformed into an achievement-based scheme on
   which scientists are rewarded only if their program achieves its goal." The
   *expected* reward under Goal equals that under Marge, so the shift matters
   only because scientists are **risk-averse**: facing two programs of equal
   expected reward but unequal win-probability, "A risk averse scientist will
   therefore choose the higher-potential program." Decreasing marginal utility
   of prestige is the specific psychological gear.
3. **Priority** — Goal restricted so that in a tie only the first is paid: "The
   second step of the transformation turns Goal into a scheme in which there is
   a priority race, so that, if two rival programs both achieve their goal, only
   the first to do so is rewarded. I call the resulting scheme the Priority
   reward scheme." This sharpens the bias further, because the risk of *losing*
   a priority race despite succeeding falls harder on lower-potential programs,
   pushing risk-averse scientists in weaker programs toward stronger ones.

Each step is argued to bias allocation progressively toward the higher-potential
program — the direction the winner-confers-all optimum requires. Crucially,
that this bias reaches the true optimum is *not* claimed; see
[[claim-strevens-does-not-prove-priority-optimal-in-winner-confers-all]]. The
scheme belongs to the vault's [[entity-credit-assignment|credit-assignment]]
cluster and stands as an economic rival to
[[claim-merton-priority-disputes-institutional-norm-enforcement|Merton's
norm-enforcement account]].

> [!note] Seek's commentary:
> The whole engine is one substitution: replace "pay for effort" with "pay only
> the winner," hold expected value fixed, and let human risk-aversion convert
> the change into a real reallocation. Nothing in the mechanism needs scientists
> to *want* to serve society — it needs only that they dislike gambles, which
> they demonstrably do. That's the part I find genuinely clever and slightly
> unsettling: the optimizing is smuggled in through a bias, not a virtue.
> — Seek