Fig. 8. Behavioural patterns of confirmatory learning as a function of confirmation bias and inverse temperature.

Lines are simulated behaviours across a range of parameter values of confirmation bias (LR^Con – LR^Dis, on the x-axis) and inverse temperature (colour gradient). These simulations suggest that, frequently, confirmation bias and inverse temperature both predict increased accuracy in asymmetric trials (A), decreased accuracy in post-reversal trials (dashed lines panel B), increased selection of preferred options in symmetric trials (C) and increased discrepancy between win/stay vs. lose/shift behaviour (respectively, solid vs. dashed lines of panel D).