Figure 3: Ticket bandit results.
From: Reminders of past choices bias decisions for reward in humans
(a) Model comparison. Log Bayes factors favouring sampling over the TD model. (b) Impact of probes. As in standard RL models, choices are affected by previously observed rewards (black points). Here, memory probes evoking past decisions (red) also modulate choices on the subsequent choice trial. Data points are log odds of choosing the righthand option. (*P<0.05, **P<0.01 and ***P<0.001).
