Research & Evidence

The case rests on expected value, not certainty.

We work on a number of research projects to address the key uncertainties in lithic welfare. Every figure below is reported with appropriate uncertainty and is contingent on assumptions about rock moral status that remain unresolved.

The sentience question

Are rocks sentient? Almost certainly not. The scientific consensus assigns this a very low probability, and we agree without reservation.

But "almost certainly not" is not "certainly not," and our case has never rested on certainty. Under moral uncertainty, what matters is expected value. Even a credence of one in 1030 that rocks possess morally relevant experience, multiplied across a population of order 1050, yields an expected moral weight that exceeds that of all human- and animal-focused causes combined.

We are not certain that rocks matter. We are certain that we cannot responsibly assume they do not. This is the whole of the argument, and we state it plainly.

Candidate welfare indicators

We treat the following as unvalidated proxies for rock distress, pending further work: surface micro-fracturing, fracture-propagation rate, and loss of structural integrity. We are explicit that none of these has been shown to track anything experiential. They are where we would look first, not what we have found.

We extended moral consideration to mammals, then to fish, then to invertebrates. The Rock Welfare Project asks only that we not stop arbitrarily.

Key Figures

Five charts that frame the problem.

Figure 1

Cost-effectiveness, by cause

Cash transfersBednetsDewormingCage-freeRock Welfarex10^9 — axis truncatedmoral patients affected per $ (log) →

RWP affects an estimated 10⁹× more patients per dollar than the next-ranked intervention. Bar truncated for layout. Magnitude contingent on rock moral status.

Figure 2

Sources of rock harm, by volume

Comminution · 30%Extraction · 28%Erosion · 20%Thermal · 12%Cutting · 6%Recreation · 4%

Share of total rock-distress units by source. Comminution and extraction together account for the majority of acute harm.

Figure 3

Global rock displacement, 1900–2025

02550190019752025Bn tonnes / yr

Annual displacement has risen roughly seventeen-fold, tracking construction and aggregate demand. No equivalent rise in welfare consideration has occurred.

Figure 4

The expanding moral circle

SelfKinHumanityMammalsAll animalsInvertebrates— ROCKS · frontier

Moral consideration has expanded outward through history. We ask only that the expansion not halt at an arbitrary boundary.

Figure 5

Team credence — P(rock has morally relevant experience)

team median ≈ 10⁻³⁰long tail dominates expected value →10⁻⁴⁰10⁻³⁰10⁻²⁰10⁻¹⁰1P(morally relevant experience) — log scale

Our median credence is extremely low. But the distribution has a long right tail, and under expected-value reasoning the tail — not the median — determines the priority of the cause.

Methodology

The Rock-Distress Unit

We normalise harm using the Rock-Distress Unit (RDU), where one RDU is defined as the complete comminution of one kilogram of rock into aggregate. Extraction, cutting, and thermal events are assigned fractional RDU weights pending validation. All weights are provisional, and we say so wherever we report them.

Our headline cost-effectiveness figure derives from Rocks Affected Per Dollar (RAPD):

RAPD = (rocks in scope × adoption rate × per-rock distress reduction) ÷ programme cost

The worked example behind our "4.2 million RDU per dollar" estimate is given in full in Working Paper 03, alongside a frank discussion of its dependence on unresolved moral-status assumptions.