Testing Bidirectional Temporal Causality

Since we have evolved to generally perceive causality as proceeding forward in time it is at least implicitly counterintuitive to propose that causality may be in fact temporally bidirectional. To suppose such a hypothesis is liberally tantamount to an insult (or at least conservatively indicative of a lack of esteem) towards science as a whole. However as systems become increasingly chaotic our current standard for predictive power begins to approach the limit of its power (e.g. weather, population growth, and the evolution of celestial magnetic fields). Increased accuracies, accommodation of initial conditions, and iterative inclusion/exclusion of increasingly predictive variables does not seem to be appreciably accelerating the curve of this differential. I, therefore propose that our perceived relation to the dimension of time to be a proprietary one. Taking into account that evolution selects for merely the minimally subjectively competitive traits (i.e. in this case, calibration of the senses) we should be careful to not take our conscious experience to be anything more than a locally objective standard. By extension then the pursuit of any truly objective knowledge must follow a comparable course of critical competition if the result is to be the same: evolution.
In science we seek to avoid bias, the most paradoxically challenging one to foresee implicitly being expectation bias itself. Moving forward in time we tend to use statistical algorithms (e.g. Monte Carlos, Metropolis) that follow energy through time since in any open system energy will move downhill. I propose following a disparate variable backward through time will minimize any tendency towards bias, while also adding power to any algorithm that seeks to predict the state of a system for any given time. Extrapolating this reductionistic physical perspective even allows for the tracking of variables (mass, space, energy) linearly through the spatial dimensions as well. Quadratic equations modeled in parallel, holding each dimension constant in turn, will describe the system from the multiplicity of perspectives possible (themselves potentially to be modeled together with a Monte Carlo-like random statistical algorithm). Then since objectivity is merely the sumof all possible subjective states, these proprietary equations may then be combined to give greater discriminatory power for the prediction of the system's state. This information can then be retrofitted back to human perceptual bias as needed for translation into non-mathematical languages.