On Open-Universe Causal Reasoning
This work addresses foundational challenges in causal reasoning for AI systems dealing with complex, open-ended environments where causal representations may be implicit.
The authors extended structural equation models and simulation models to infinite variable spaces, enabling a semantics for conditionals based on intervention calculus and axiomatizing causal reasoning for expressive generative models in open-universe settings. They showed these two model types are equivalent under certain restrictions despite differing axiomatizations generally, and provided complete axiomatizations where the open-universe aspect is essential.
We extend two kinds of causal models, structural equation models and simulation models, to infinite variable spaces. This enables a semantics for conditionals founded on a calculus of intervention, and axiomatization of causal reasoning for rich, expressive generative models -- including those in which a causal representation exists only implicitly -- in an open-universe setting. Further, we show that under suitable restrictions the two kinds of models are equivalent, perhaps surprisingly as their axiomatizations differ substantially in the general case. We give a series of complete axiomatizations in which the open-universe nature of the setting is seen to be essential.