Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining causal models for controlling environments. One of the methods includes repeatedly selecting control settings for the environment based on (i) a causal model that identifies causal relationships between possible settings for controllable elements in the environment and environment responses that reflect a performance of the control system in controlling the environment and (ii) current values of a set of internal parameters; and during the repeatedly selecting: monitoring environment responses to the selected control settings; determining, based on the environment responses, an indication that one or more properties of the environment have changed; and in response, modifying the current values of one or more of the internal parameters.

A system and methods for multivariant learning and optimization repeatedly generate self-organized experimental units (SOEUs) based on the one or more assumptions for a randomized multivariate comparison of process decisions to be provided to users of a system. The SOEUs are injected into the system to generate quantified inferences about the process decisions. Responsive to injecting the SOEUs, at least one confidence interval is identified within the quantified inferences, and the SOEUs are iteratively modified based on the at least one confidence interval to identify at least one causal interaction of the process decisions within the system. The causal interaction can be used for testing, diagnosis, and optimization of the system performance.

An example method includes selecting treatment assignments for operation of a system in an environment based on a causal model mapped to a probability distribution over possible treatment assignments, where the causal model measures causal relationships between the treatment assignments and a performance metric associated with the environment. The method further includes mapping causal effects between the treatment assignments and measures of uncertainty associated with the environment in the causal model to corresponding probabilities using probability matching. The method further includes determining the performance metric using the treatment assignments and corresponding dependent variables obtained from the environment in response to the treatment assignments. The method further includes adjusting, based on the performance metric, the causal model, and re-computing the adjusted causal model by computing overall causal effects based on the measures of uncertainty around the overall causal effects.