On the other hand, for the general case in which αL
̸= αF
, the
choice-autocorrelation factor can also influence the reward history
because it directly affects the number of identical choices that are
made after a reward is given, and this effect influences the regres-sion coefficients for the reward history, as discussed above (see
Eq. (21)). Fig. 5(B) shows the simulation results for the standard
Q-learning model with αF = 0 and illustrates one example of this
effect. As ϕ increases in a positive domain, the tendency to repeat
the same choice increases and enhances the decay of the influence
of the reward history compared to the case in which there was no
choice-autocorrelation factor (ϕ = 0). As ϕ decreases in a negative domain, the opposite effect is observed. The residual choice-autocorrelation factor has an additive effect on the regression
coefficients for the choice history, bc
. Taken together, the effects of
the choice autocorrelation factor on the dependence on the history
are largely additive and straightforward. For the general case, how-ever, this factor may modulate the dependence on reward history
through the property that we have observed in previous results,
i.e., that it depends on the number of times that the option is chosen.