1: \begin{abstract}

2: We study the problem of globally optimizing the causal effect on a target variable of an \emph{unknown causal graph} in which interventions can be performed. This problem arises in many areas of science including biology, operations research and healthcare. We propose Causal Entropy Optimization (\our), a framework which generalizes Causal Bayesian Optimization (\cbo) \cite{cbo} to account for all sources of uncertainty, including the one arising from the causal graph structure. \ceo incorporates the causal structure uncertainty \emph{both} in the surrogate models for the causal effects and in the mechanism used to select interventions via an information-theoretic acquisition function. The resulting algorithm automatically trades-off structure learning and causal effect optimization, while naturally accounting for observation noise. For various synthetic and real-world structural causal models, \ceo achieves faster convergence to the global optimum compared with \cbo while also learning the graph. Furthermore, our joint approach to structure learning and causal optimization improves upon sequential, structure-learning-first approaches.

3: %Our experiments with real-world structural causal models (\scm{s}) demonstrate how \ceo achieves faster convergence to the global optimum than \cbo, while also learning the graph, and how our joint approach to structure learning and causal optimization improves upon sequential, structure-learning-first approaches.

4: \end{abstract}

5: