Figure 5.

Control results for Crooks’ fluctuation theorem in two scenarios: (A) the sensorimotor loss behaves like a Mexican hat function and (B) the sensorimotor loss behaves as an exponential quadratic error but we sample the observed angles randomly with repetition. The black line is the theoretical prediction of Crooks’ fluctuation theorem (4) while the curves stand for the mean path after 1000 bootstraps of the observed driving error values. The shaded areas inside the graphs are the 99% confidence intervals which result from bootstrapping. Note, for simplicity, we assume $\beta =1$ for all participants when using the Mexican hat to demonstrate that the result in (A) does not trivially hold for any cost function. For (B), we fit the parameters for each participant according to Sect. A.3.3.