The Direction Problem: On the Causal Origin of Oriented Motion in Complex Systems

While constructing a predictive model for phase transitions in complex systems, we encountered a dilemma: the empirically measured acceleration constant g_j increases as the system approaches the transition point. In a conventional causal framework — where the past drives the present — a system consuming its internal resources should decelerate, not accelerate. The measured 67% increase in g_j is incompatible with a push-only dynamics. Starting from a simple observation — that every functional system exhibits a vector oriented toward the future — we follow a chain of logical eliminations. The orientation cannot originate in the past, which does not contain the novelty toward which the system moves. It cannot be random: randomness has no direction. The only remaining option is that the source of the orientation is in the destination itself. We formalize this as the Principle of Causal Inversion: every determined result occupies a coordinate in a space where time is treated as a navigable dimension, and from that coordinate it emits a signal — a pull — that intensifies with proximity. The principle resolves five open problems: the origin of genuine novelty, the arrow of time, the source of spontaneous impulses in neuroscience, the anomalous pre-transition acceleration in complex systems, and the multi-scale coherence of simultaneous events without lateral coordination. Empirical evidence is presented from a reaction-diffusion model of civilizational dynamics. The principle is falsifiable: the conditions under which it would fail are specified, including the pre-registered condition F6 added in v1.2, whose second-phase evidence (§6.3) shows the calibrated constants transferring unchanged across the July 2026 macro-junction and dating four further junctions to within two days each — the first empirical datum bearing directly on the Ordinative Equivalence Principle. Companion paper to The Collapse Equation (https://doi.org/10.5281/zenodo.19932312). Notation aligned with the Ordinative Sciences Symbol Canon v1.2.