Why it is a question
In two hydrodynamically coupled beads, with one trap moved by a recorded drive, delayed mutual information and transfer entropy read the inter-bead flow both ways at moderate drive. In numerics of the fitted linear model they read it one way at strong drive. The coupling tensor stays symmetric throughout (THEORY-275, from LIT-947 read in NOTE-788). The strong-drive regime is beyond that experiment, and the system is linear, Gaussian and two-body. Whether directional information measures in general follow a one-sided drive rather than the coupling’s symmetry is therefore open. Until 2026-10-19 this was THEORY-275’s promote_when.
What would answer it
A second system, experimental or exactly solvable, in which units with a known reciprocal coupling are driven on one side at graded strengths. The transfer-entropy asymmetry between them must grow with the drive to one-way flow while the coupling stays symmetric, and it must be measured, not only simulated, in the strong-drive regime. An exact derivation of the strong-drive limit in the linear Gaussian pair would settle that case analytically. There, delayed-MI asymmetry is c_xy(τ)² − c_yx(τ)².
It would answer no if directional measures between reciprocally coupled units stay symmetric under strong one-sided drive, or if they recover the coupling’s symmetry without the drive being recorded or intervened on.
What would not answer it: systems whose coupling is itself asymmetric, or further cases in the moderate regime alone, where flow is already two-way.