Neural Random Utility and Measured Value

We present a method for relating a neural measurement of value to choice behaviour. In a previous study, precisely targeted measurements of brain activity were made in the medial prefrontal cortex of subjects while they considered individual consumer goods. We present here two advances. First, we develop an empirical framework for relating this class of measured value data to choice prediction. Second, we apply a benchmarking tool to compare the predictive power of a measured value dataset with established techniques. We find that our measured neural activity cardinally encodes valuations and predicts choice behaviour, though a significant degree of measurement error affects prediction rates. Accounting for measurement error and combining neural data with standard observables improves predictive performance. We also note some potential normative implications of our measured value approach.