F5: DA Output Is Gain-Controlled at the Source
Integration
This premise integrates the DA gain control mechanism (Grace et al. 2007; Inglis et al. 2022 MODAL model) with the catecholamine-PFC dynamics to form a single causal chain: environment -> DA gain modulation -> PFC function.
The vSub-NAcc-VP-VTA pathway controls how many DA neurons are available for phasic response, functioning as population-level gain control. Environmental variables (controllability, uncertainty, volatility) modulate this gain. The MODAL model (Inglis et al. 2022) provides a spiking-level computational implementation validated against neural data, with state variables (NAcc membrane potentials, VP firing rates, VTA active population size) and a clear input-output mapping (modulating variable -> DA gain on RPE).
Why this matters for the architecture
The DA gain control pathway is the mechanism by which environmental context modulates the human's capacity for adaptive learning and reward-guided behavior. When gain is appropriately set, phasic DA signals drive effective learning. When dysregulated (either too high or too low), the same signals produce either hypervigilant responding or learned helplessness.
This mechanism connects the environmental design (what the system presents, when, in what context) to the human's neurobiological capacity. It is why the architecture specifies that the machine must modulate its own behavior based on state estimation: the machine's actions are inputs to a system (the DA gain control pathway) that determines how the human processes information and learns.
The missing link
Grace's circuit explains how contextual signals modulate DA gain. Arnsten's work explains how DA levels modulate PFC function. These two literatures describe adjacent links in the same causal chain but have not been integrated as a single pathway relevant to human-AI system design. The computational neuroscience of reward learning and the neuroscience of stress-induced cognitive degradation are largely separate research communities working on the same transmitter system.
Evidence status
SUPPORTED. Spiking-level computational model validated against neural data with multiple empirical benchmarks.