Consciousness research employs diverse affective scales, yet current measures remain fragmented and weakly tied to explicit consciousness models. The States of Mind (SOM) model addresses this gap by applying a reflexive self-other architecture that yields a discrete affect variable with empirically constrained balance points. By modeling positive (1) and negative (0) states of self and other across three reflection levels, SOM generates balance point states distinguishing psychological functioning from dysfunctional to optimal. Clinical and cross-cultural validation studies show observed affect balances cluster at predicted points across diverse populations and contexts. SOM's architecture aligns naturally with Active Inference accounts of consciousness, particularly Seth's formulation of self as an embodied, hierarchical system. By quantifying positive response likelihood, SOM functions as an affective readiness-to-respond variable modulating policy selection in predictive processing terms. Shifts in affective context can move SOM among balance point states, altering predicted probabilities while maintaining core consistency of the self-generative model. As a Bayesian prior, SOM establishes expectations for actions and outcomes, enabling quantitative assessment of prediction error and signaling when affective recalibration is warranted. SOM is proposed as a plug-in affect metric for consciousness-based architectures, including Active Inference, Global Workspace models and affect-modulated robotic control systems.