summary - Points out that “Earth has survived N such experiments/exposures so far” is not, by itself, valid evidence that the per-event risk is low: any observers who ask the question necessarily live somewhere that hasn’t yet been destroyed, so naive survival-based arguments (including the cosmic-ray-survival case this pipeline’s empirical-bounds slice relies on) are subject to an observer-selection bias that can hide an arbitrarily high true catastrophe rate. The authors then derive a selection-bias-free upper bound of about 1 per billion years (99.9% confidence) on any exogenous terminal-catastrophe rate, using the distribution of planet-formation times rather than Earth’s own survival history.
relevance_note - Identifies a hidden assumption underneath the empirical “cosmic rays already do this and we’re still here” argument this slice’s sibling (empirical-bounds) relies on — one of the more load-bearing “what does it hinge on” points for the final report, since it says survival-based reassurance needs an extra anthropic correction to be valid at all.
Extracted
A-6 - Observer selection makes naive survival-based risk bounds uninformative
reasoning - Condition on the existence of the observer: every observer able to ask “has a terminal catastrophe happened here?” is guaranteed to find the answer “no”, whatever the true rate. So P(observed survival | high rate) = P(observed survival | low rate) = 1 once observer existence is conditioned on; the likelihood ratio is ~1 and the survival record is (to first order) uninformative about the rate. Hence any survival-based reassurance — explicitly including the cosmic-ray-survival premise (D-1 - Cosmic-ray survival premise - high-energy cosmic rays have bombarded astronomical bodies over Gyr without catastrophe) underlying collider-safety arguments — is valid only after an anthropic correction, e.g. by using data whose distribution is not conditioned on our own survival (as the formation-time bound does). Scope note: the schema only allows attachment to same-paper nodes; this argument’s principal external force is against evidence links resting on D-1 (survival of Earth/Sun specifically; survival of independently observed bodies like neutron stars is only partially shielded from the bias), which steps 5-6 should price using this reasoning.
Original
“That Earth (or any observed body) has survived past exposure cannot by itself bound the true catastrophe rate, because observers necessarily find themselves only where catastrophe has not yet occurred.”
Verdict (step 6)
corrected / checked. The core step is valid and traced: conditioning on observer existence forces P(observed survival | high rate) = P(observed survival | low rate) = 1 for any body whose destruction would have precluded the observation (Earth, Sun), so the likelihood ratio is ~1 and such survival is uninformative. But the parenthetical “(or any observed body)” overreaches: for causally quasi-independent distant bodies (WD/NS), our existence does not require their survival, so P(observing intact specimens | high rate) genuinely falls with the rate and the likelihood ratio departs from 1 — only a partial bias remains (selection of observable/catalogued objects). The body’s own scope note already concedes this. Statement narrowed to the version immune to that defeater; filename left as-is.
Link to original
O-10 - Habitable-planet formation times span many Gyr and Earth's formation date is unremarkably late
Link to original
A-7 - Earth's unremarkably late formation time yields a selection-bias-free catastrophe-rate bound near 1 per Gyr
reasoning - Under a spatially uncorrelated exogenous catastrophe process at rate 1/tau, the chance a planet formed at cosmic time t still hosts observers at time t+dt decays exponentially in elapsed time, so the observer-weighted formation-time distribution is the raw formation-time distribution exponentially tilted toward late formation dates (late formers have had less time at risk). If tau were short, essentially all observers would find their planet formed far later, relative to the formation-time distribution, than Earth’s ~9.1 Gyr date — i.e. our unremarkable, near-typical date would be a huge coincidence. Computing P(formation date as typical as ours | tau) against the distribution gives tau > ~1.1 Gyr at 99.9% confidence. The inference is selection-bias-free because it compares among surviving observers rather than counting survival itself as evidence. Assumptions it hinges on: the adopted planet-formation-time distribution (model-dependent), spatial/temporal uncorrelatedness of the catastrophe process, and exogeneity (rates humans modulate are outside the bound).
Verdict (step 6)
approved / checked. Traced the load-bearing step: for an exogenous, spatially/temporally uncorrelated catastrophe process at rate 1/tau, survival to observer-time decays as exp(-(t_obs - t_form)/tau), so the observer-weighted formation-time distribution is the raw one exponentially tilted toward late t_form; small tau pushes essentially all observers onto anomalously late-formed planets, so a near-typical formation date like Earth’s
Link to original9.1 Gyr becomes very improbable — a clean typicality test yielding tau >1 Gyr at the stated confidence. Crucially the comparison is among surviving observers, so the anthropic defeater that kills naive survival bounds (A-6) does not apply. The statement already names its real hinges (formation-time distribution model, uncorrelatedness, exogeneity); their truth is priced at steps 7-8, not here.
H-5 - The exogenous terminal-catastrophe rate is below about 1 per Gyr at 99.9 percent confidence
Rests on the planet-formation-time distribution rather than Earth’s own survival record, so it is free of the selection bias it criticizes — but inherits the modeling assumptions of that distribution and the restriction to exogenous (externally triggered) catastrophes; it does not directly bound risks whose rate humans change, like collider operation itself.
Link to original