Investigated
How do scientists test improbable causation?
Short answer
Image: link.springer.com
Scientists test improbable causal claims by deriving predictions from a causal model, comparing them with observations or experiments, and checking whether competing explanations fit better. For individual cases or rare events, probability usually yields bounds rather than certainty; critics also warn that a tiny p-value or unlikely alternative cause does not by itself prove causation.
The views on this question
Two lenses on the same evidence. Source weight and the primary source ratio show what each rests on.
Alternative
Source weight 50%Lens adapted to this topic: Limits of probability-based causal claims
Dissenting methodological perspectives argue that researchers can overstate what improbability establishes. A rare alternative cause is not, by itself, evidence that a specified cause produced the observed outcome; statistical significance may ignore effect size and practical importance; and even randomized experiments may identify treatment effects without settling every philosophical meaning of “cause.” These critiques favor explicit assumptions, competing explanations, and substantive scientific judgment.
- A very low probability of an alternative natural cause does not by itself establish that the proposed cause produced a specific outcome; comparison with relevant reference frequencies may be necessary.
- Statistical significance alone can mislead when it is detached from effect size, practical importance, and scientific judgment.
- A Bayesian account of coincidences treats an improbable event as possible support for an alternative causal theory, not automatically enough evidence to accept it when that theory has a low prior probability.
- One philosophical critique argues that randomization identifies an average treatment effect but does not, without further assumptions, settle the ontology of treatment as an individual cause.
- Matching-based approaches argue that probability-of-causation distributions may be estimated from experimental or observational data, while sample size and event rarity affect the resulting uncertainty.
Follow the white rabbit
Deeper threads worth pulling on next.
- What is counterfactual reasoning?Different angle, same storyCounterfactuals test causation by asking what would have happened under different conditions.
- How do epidemiologists infer causation?Different angle, same storyThey use study design, confounder control, and criteria like Bradford Hill to assess causal claims.
- What are false positives and p-hacking?