Observing interventions: a logic for thinking about experiments

Abstract This paper makes a first step towards logic of learning from experiments. For this, we investigate formal frameworks for modeling the interaction causal and (qualitative) epistemic reasoning. Crucial our approach is idea that notion an intervention can be used as expression (real or hypothetical) experiment (Pearl, 2009, Causality. Models, Reasoning, Inference, 2nd edn. Cambridge University Press, Cambridge; Woodward, 2003, Making Things Happen, vol. 114 Oxford Studies in Philosophy Science. Press). In extend model (Briggs, 2012, Philosophical Studies, 160, 139–166; Galles Pearl, 1998, An axiomatic characterisation counterfactuals. Foundations Science, 3, 151–182; Halpern, 2000, Axiomatizing Journal Artificial Intelligence Research, 12, 317–337; Cambridge) with simple Hintikka-style representation state agent. resulting setting, one talk about knowledge agent information update. The reasoning thought However, it unable to account experiments, which clearly brought out by fact validates principle no interventions. Therefore, second step, implement more complex (Nozick, 1981, Explanations. Harvard Cambridge, Massachusetts) allows observe (measure) certain variables when carried out. extended system does allow all proposed logics, provide sound complete axiomatization.