Intervention and Causation Intervention and Causation: A Philosophical Perspective Victor Gijsbers (Philosophy, Universiteit Leiden) 2012-10-05
Causation● A central topic in modern philosophy.● Why believe that philosophy could teach us something about causation?● The idea and use of philosophical analysis.
Intervention● Central notion in current debates: intervention. – Pearl, Causality (2000) – Woodward, Making Things Happen (2003)● Using intervention to understand causation.● Why is that a good idea?
Overview● Causation seems to be: – important, – modal.● Main development in theories of causation: – From reductive analyses – to interventionist analyses.
Importance of causation● Why do scientists care about causation?● “The law of causality, I believe, like much that passes muster among philosophers, is a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm.” – (Russell, On the Notion of Cause, 1912)
Importance of causation● Science is interested in functional relations. – pV = NkT – number of people infected = F(population density, quality of sanitation, availability of hospitals, …) – P(number of people infected) = F(population density, quality of sanitation, availability of hospitals, …)● So why be interested in causation?
Importance of causation● The functional relations by themselves are not guides to action.● We want to know about cause and effect.● But what is the relation between causation and functional relations?
Causation and modality● Some of our statements are about what actually happens; others are about modal facts. – What could happen. – What must happen. – What would happen, if...● Modal statements play an important role in planning, justifying, assigning blame and credit.
Causation and modality● Causation seems to be closely linked to modality.● If A caused B, then, if A had not happened, B would not have happened.● That doesnt quite work... but intuitively, something in the vicinity should.
Causation● Summarising, causation seems to be: – important – modal.● Suggestive links with action (but at this point, no more than suggestive).
Reductive theories● Reductive, regularity theories of causation. – Modal facts have to be reduced to facts about what actually happens. – Actions should not appear in the theory.● Why? Because our scientific knowledge is based on what we see. Anything else would be superstition, pseudoscience, empty speculation.● And we only see what is actual.
Reductive theories● Science is based on what we see.● Objection: it is also based on what we do.● Do → see → science.● Doing falls out of the equation when we think about the justification of our theories.
Reductive theories● We want to analyse causation in terms of observed relations between event types.● Most simple theory is the constant conjunction theory (Hume, 1748): – A causes B ↔ whenever A happens, B happens immediately afterwards● Can be changed to accommodate probabilistic relationships.
Counterexamples● There is constant conjunction without causation. For instance: – Common cause structures. – “Accidental” conjunction.● Both are important to recognise in actual scientific research.
Counterexamples● Reductive theories fail to capture the modal aspect of causation.● They fail less often when we add more causal variables. Still – this doesnt look like the right way. There always remains the possibility of accident.● Which we can accept in practice, but not when we want to know the meaning of causation.
Interventionism● Very rough example of an interventionist theory of causation: – A causes B ↔ by intervening on A, we can change B.● Why would talking about interventions help?
Using intervention● We do use interventions to test causal claims. Is A a cause of B? We intervene on A, and check whether B is still probabilistically related to A. – Yes? Causation! – No? No causation!● Both conclusions can be wrong.
Interventionism● Havent we just added another causal variable?● Reductive way of thinking: we perform interventions so that we can determine a set of functional relationships that we are interested in.● The doing is merely there to serve the seeing.● The possibility of accident remains.
Interventionism● Remember our statement of interventionism: – A causes B ↔ by intervening on A, we can change B.● This is an inherently modal claim. We set A = 1, and what happens is B = 1. But this only counts as a changing of B, if it is that case that had we set A = 0, then B would not have been 1.● Something counts as a successful intervention only if the effect was not accidental.
Interventionism● But why would modality be connected with our doing of things?● Because when we do things, we make choices. [Insert difficult questions about free will here.]● It is when we think about our future plans, about what we want to change, and so on, that we start thinking modally; and that we start to get interested in causation.
Interventionism● Beings who could only see things, but could not do things, would not be interested in causation.● What is more: they could not have the concept of causation. We can only understand causation because of the modality of action.● It is therefore no surprise that a philosophical analysis of causation must use a term like intervention.
Conclusion● But it is also no surprise that theories about causal modelling and causal inference can make good use of the notion of intervention.● Because claims about causation turn out to be claims about the effects of actions.● Reminder: this is a controversial story!