This preprint presents a dynamic, time-indexed extension of Kurt Lewin's classic formula B = f(P, E). It makes the time dimension t explicit and introduces two additional state variables: EFilter_t, an emotional weighting field over a limited-capacity workspace, and HOT_t, a meta-awareness regime variable tracking how clearly a person can see their own thinking in the moment. The framework is designed to explain why the same person, in the same nominal situation, can behave very differently from one moment to the next, especially under emotional strain. It embeds EFilter_t and HOT_t into a time-indexed version of Lewin's equation and a simple state vector, and connects this to global workspace theory, predictive processing, and higher-order thought models. The paper proposes initial operational proxies that combine phenomenological ratings with behavioral and physiological measures (including candidate EEG markers), together with explicit psychometric validation criteria. It also outlines three falsifiable within-subject hypotheses that tie the framework to drift-diffusion models of decision-making, with concrete predictions about how emotional tilt and meta-awareness bias choice over time. This work is explicitly framed as a theory and empirical roadmap rather than a report of completed empirical data. It includes an illustrative formal sketch and a minimal modeling strategy intended to be implementable by research groups interested in dynamic models of behavior, affect, and meta-awareness. An earlier, more preliminary version of this idea was posted as a Zenodo preprint; the present version substantially revises the theory, operationalization, and empirical roadmap, and is intended as the definitive journal submission.