An efficiency framework for valence processing systems inspired by soft cross-wiring

Recent experiments suggest that subsecond dopamine delivery to human striatum encodes a combination of reward prediction errors and counterfactual errors thus composing the actual with the possible into one neurochemical signal. Here, we present a model where the counterfactual part of these striatal dopamine fluctuations originates in another valuation system that shadows the dopamine system by acting as its near-antipode in terms of spike-rate encoding yet co-releases dopamine alongside its own native neurotransmitter. We show that such a hypothesis engenders important representational consequences where valence processing appears subject to the efficient encoding considerations common to the visual and auditory systems. This new perspective opens up important computational consequences for understanding how value-predicting information should integrate with sensory processing streams.