Agency as a State Variable: A Theoretical Framework for Decision Initiation in Non-Coercive Contexts

The dominant tradition in decision science rests on a foundational presupposition: humans are autonomous agents whose decision-making requires principally an account of how choices are evaluated and selected. This paper challenges a more primitive assumption — that agency is reliably present to evaluate anything at all. Drawing on convergent indirect evidence from behavioral economics, motivational psychology, and the neuroscience of voluntary action, we propose that human agency in non-coercive, low-pressure everyday contexts may function not as a stable background capacity but as a threshold-dependent emergent state. We introduce the Decision Initiation Threshold (DIT) as a two-level construct: a latent threshold parameter (DIT-threshold) specifying the activation conditions required for self-directed decision initiation, and an occasion-level propensity score (DIT-state score) estimating the probability of crossing that threshold in any given context. Person-level estimates of the DIT-threshold are inferred from the distribution of repeated DIT-state scores using multilevel modelling, rather than equated with the mean state score. The DIT framework addresses a pattern observable in clinical, organizational, and educational settings — a systematic gap between what individuals are capable of deciding and what they actually decide without constraint — and accounts for this gap in terms of initiation failure rather than motivational content or decision quality. We differentiate DIT from existing constructs including proactive personality, locus of control, ego depletion, and self-determination theory's autonomy need. We propose a three-priority research program: (1) prospective ecological measurement of initiation frequency via Experience Sampling; (2) investigation of neurocognitive correlates of uninstructed voluntary initiation using the Open-Window Task (OWT), a paradigm designed to study unprompted initiation behavior; and (3) randomized testing of whether DIT is trainable. The framework carries implications for decision science, clinical psychology, organizational behavior, and educational design.