A counterfactual sentence does not prove an alternative history. Name the event, change, retained conditions and grounds. Observation, a conditional model and an individual inference differ; blame needs separate grounds.
Questions explored
Is imagining another outcome enough?
No. Grounds and retained conditions are needed; a toy model produces a conditional deduction within its rules.
Why can a condition’s coincidence with an output be insufficient?
The two stipulated models in V fit the same observations but differ when A changes. Further grounds are needed to distinguish them.
Does a counterfactual establish someone’s blame?
Not by itself. Evaluation needs separate information and norms; numerous conditions do not automatically remove responsibility.
Key distinctions
Hume includes the absence of the second without the first in a formulation of cause. (E 7.29 alongside conjunction and customary expectation; not a modern outcome-identification algorithm.)
Mill’s Method of Difference requires shared circumstances except the selected difference. (III.8, §2 and selected §3; an indispensable part is allowed, and historical dangerous examples are not practice instructions.)
Pearl’s structural approach states a changed condition and retained context. (§3.4 and intervention definition; no experiment, proof replication or calculation of this meeting.)
Different toy rules can fit the same observations while answering an alternative differently. (Separately stipulated copy-A/copy-B rules in V; no actual subject mechanism established.)
A conditional answer, group relationship, individual case and responsibility judgment are separate results. (Hypothetical models and meeting; no universal cancellation or equalisation of blame, and no established record efficacy.)
Limits of interpretation
- A bounded account of counterfactual questions, not a full causal-inference course or moral verdict. Indicators and meeting are hypothetical; rules are not discovered laws of real devices or behaviour. Hume, Mill and Pearl remain distinct. Retained conditions and model scope are required; observations identify no universal alternative answers. No record efficacy, diagnosis, experiment or future guarantee claimed.
Sources
- An Enquiry Concerning Human Understanding, E 7.28–7.30 — David Hume
Complete E7.28–7.30 web129–149, preceding tail128; rest VII not credited.
Source use: HISTORICAL_PRIMARY_EDITION_BOUNDED_USE_JURISDICTION_UNRESOLVED
- A System of Logic, Ratiocinative and Inductive, Book III — John Stuart Mill
Complete III.5 §3 web2410–2440; III.5 §4 opening2441–2446 only. Complete III.8 §2 web2823–2831 and §3 web2834–2846 selected contiguous passage; later §3 discontiguous matches not credited complete. Complete III.10 §§1–2 web3155–3189; §3 opening3190–3191 only.
Source use: HISTORICAL_PRIMARY_EDITION_BOUNDED_USE_JURISDICTION_UNRESOLVED
- Causal inference in statistics: An overview — Judea Pearl
Introduction to structural models web290–315 selected; web377–436 selected body including complete definition1, not entire §3.1. Nonparametric model and intervention §3.2–3.2.1 web437–477 complete selected definitions; later efficacy formulas480–487 read but not relied upon. Complete §3.4 web910–1020, printed119–122. Remaining sections, proofs, figures, software and empirical studies not fully inspected or reproduced.
Source use: BOUNDED_PRIMARY_METHOD_PARAPHRASE_NO_RAW_IMPORT_PERMISSION