The Contextual Character of Evidence for Causal Claims

1. The Contextual Character of Evidence for Causal Claims Mauricio Suárez CaEitS conference, University of Kent, 7 September 2012

2. The Main Claim • Evidence for causal claims is context-dependent: “a causes b” may be warranted by the same evidence in one context but not another. • N.B. Distinguish carefully from: “causal claims are context-dependent: “a causes b” may be true in one context but not another”. :: I am NOT arguing for the relativity of causal knowledge.

3. The Argument Woodward’s (2003) manipulability theory: Evidence for causal claims exhibits three kinds of context-dependence: -Personal -Objective - Hermeneutical The argument generalizes to other theories of causation, i.e.: 1 - Counterfactual theories (Lewis, 1985): The generalization rides upon the relation between invariant laws, and the counterfactuals they support. 2 - Process theories (Dowe, 2000): It rides upon the relation between conservation principles in explanatory theories, and the possible interventions they sanction.

4. The Manipulability Theory (I) Woodward (2003): a causes b if (but not only if) the functional relation between a and b remains invariant under interventions upon the putative causal variable a. a ? b I Evidence: tests for invariance under intervention.

5. The Manipulability Theory (II) I is an intervention on a with respect to b if and only if: • I is a cause of a • I acts as a switch for any other variable x that causes a (within a certain range of values of a): a b I x

6. The Manipulability Theory (III) • Any directed path from I to b goes through a. • I is statistically independent of any variable x that causes b and is on a directed path that does not go through a: a b I x

7. Sources of Contextuality (I) Michael Williams (Problems of Knowledge, 2001): a) Personal a.1) Intelligibility (of claims) a.2) Method (of inquiry) a.3) Dialectics (within the inquiry) a.4) Economics (involving valuations of risk and cost of inquiry) b) Objective or situational (relativetoobjective grounds)

8. Sources of Contextuality (II) Williams’ example: A: Isn’t that old sports car an E-Type? B: Yes, a rare early model A: What makes you say that: don’t they all look pretty much the same? B: Sure, but that one had external bonnet latches which you can only get on the first five hundred cars. Methodological: default entitlements in identifying, discussing and assessing evidence regarding cars: usual properties, function, form, purpose, distinguishing features that may be attended to, etc. Objective: Confounding factors in the objective situation,e.g.isthe car in question is a replica of an original of the same type, have there been an exceptional production, etc.

9. Contextual Causal Evidence (I) Methodologicaldefault entitlements or presuppositions of the manipulability theory include: • Causes and effects are variables in a directed acyclic graph. • Variables take values within particular ranges (significant intervals for the assessment of invariance). • Variables have quantitative functional relations • Interventions are in principle possible causes of the values of putative causes.

10. Contextual Causal Evidence (II) Objectiveconfounding factors, appropriately the defining conditions forinterventions: • I is in fact not a cause of a (but maybe merely correlated) • I in fact does not switch off alternative causes (they maybe independently / accidentally switched off). • I causes b by another path that does not go through a. • I is in fact correlated with other causes of b.

11. Contextual Causal Evidence (III) Hermeneutical factors: In theoretical science different frameworks will provide different sets of default entitlements and confounding factors Example from physics: EPR correlations modelled differently in quantum and Bohmian mechanics. Entitlements: hidden variables, trajectories, etc. Confounding factors: independent causation via underlying structure (quantum potential, etc).

12. Conclusions Evidence for causal claims is contextual: • Methodological (personal) default entitlements • Objective (situational) confounding factors • Hermeneutical (theoretical) interpretations :: The literature on causal inference has not so far identified and distinguished appropriately such sources of contextual evidence.