Working Paper 03 released: two-envelope problems in moral-weight estimation.

Working Paper 03, Two-Envelope Problems in Lithic Moral-Weight Estimation, is now available as a pre-print on request. It addresses a structural difficulty that arises when comparing welfare across populations whose intrinsic moral weights are subject to deep uncertainty — a category in which rocks figure prominently.

The core question is this: if our credence about per-rock moral weight is itself a probability distribution, and the same is true of comparison populations (e.g. mammals, fish, invertebrates), what does it mean to take an "expected" weight, and which framing is privileged? Different framings can yield estimates that differ by many orders of magnitude. Our headline 4.2 million RDU-per-dollar figure assumes one such framing, which the paper makes explicit.

The paper does not resolve the two-envelope problem. It documents it, and proposes three partial mitigations: (i) reporting results under multiple framings simultaneously, (ii) treating large divergences as evidence of model fragility rather than of cause-area dominance, and (iii) discounting headline figures by a "framing-sensitivity" factor when communicating with non-specialist audiences.

We do not consider these mitigations adequate. We consider them the best that can presently be said. The honest reading is that our cost-effectiveness claims remain large under most framings, modest under some, and unbounded under others. We invite criticism, and we expect to receive it.

The pre-print is available on request via contact. Peer review is pending; we are early.