In the field of artificial intelligence, a model must be causal to support reasoning about interventions and counterfactuals, and also combinatorial to generalize to unseen combinations of objects. This work formally investigates when and how such a model can be learned. We develop relational structural causal models, extending structural causal models (Pearl 2009) to settings where objects and their relations vary.
First, we show that answers to both causal and observational queries about unseen combinations cannot be identified without further assumptions. To enable such identification—including in the presence of unobserved confounding—we define relational causal graphs and derive symbolic identification criteria.
Finally, we propose relational neural causal models, a provably correct approach that outperforms non-relational baselines on simulated traffic scenes with varying cars, signals, and pedestrians.
Blogger's Review: The relational structural causal models introduce a new perspective for AI reasoning by incorporating variations in object relationships, significantly enhancing the model's flexibility and adaptability. This research not only advances the theoretical development of causal reasoning but also provides robust tools for practical applications, particularly in decision support within complex environments.