The purpose of this essay is to discuss two approaches to inference and how “human capital” can provide a way to combine them. The first approach, ubiquitous in economics, is based upon the Rubin–Holland potential outcomes model and relies upon randomized treatment to measure the causal effect of choice. The second approach, widely used in the pattern recognition and machine learning literatures, assumes that choice conditional upon current information is optimal (or at least high quality), and then provides techniques to generalize observed choice to new cases. The “human capital” approach combines these methods by using observed decisions by experts to reduce the dimensionality of the feature space and allow the categorization of decisions by their propensity score. The fact that the human capital of experts is heterogeneous implies that errors in decision making are inevitable. Moreover, under the appropriate conditions, these decisions are random conditional upon the propensity score. This in turn allows us to identify the conditional average treatment effect for a wider class of situations than would be possible with randomized control trials. This point is illustrated with data from medical decision making in the context of treating depression, heart disease and adverse childbirth events.