Kyoto University: theft recidivists without kleptomania show lower learning rate in 4-arm bandit task
In a 4-arm bandit task with prefrontal fNIRS recording, non-kleptomanic theft recidivists updated action values more slowly than controls, the authors report.
Researchers led by Yukiori Goto at Kyoto University studied decision-making in people with repeated theft convictions. They used a four-arm bandit task while recording prefrontal cortex hemodynamics with functional near-infrared spectroscopy (fNIRS). The study was published on November 9 in Frontiers in Psychology frontiersin.org
Frontiers in Psychologyfrontiersin.orgOpen source ↗. Three groups took part: 53 control participants without a criminal record, 38 non-kleptomanic theft recidivists and 12 recidivists with a kleptomania diagnosis.
Task and behavior
In the task, a participant chooses among four options whose payout probabilities shift over 160 trials, and the best option keeps changing. To do well, a participant has to balance exploring new options against exploiting the current best one, and to revise that balance as feedback arrives.
Non-kleptomanic recidivists accumulated significantly higher cumulative regret and made fewer optimal choices than controls. The kleptomanic group did not differ from controls on these measures.
The authors fitted the choices with computational models using Variational Bayesian Analysis. Q-learning with decay gave the best fit frontiersin.org
Frontiers in Psychologyfrontiersin.orgOpen source ↗. In this model each action carries a value that is updated after feedback, and stored values gradually fade over time. The learning rate, which sets how strongly one outcome shifts the value estimate, was lower in the non-kleptomanic group than in controls and in the kleptomanic group. The authors interpret this as a deficit in updating action values after environmental feedback.
Prefrontal activity
fNIRS measures how hemoglobin absorbs near-infrared light passing through the skull, which tracks changes in oxy- and deoxyhemoglobin. The method tolerates head movement, but it only reaches regions near the cortical surface.
The team compared trial-by-trial latent reinforcement variables with prefrontal signals. Prefrontal activity was modulated by these variables in all participants. Group differences appeared as static baseline shifts in hemodynamics rather than as rewiring of the circuits that track value.
The authors conclude that recurrent non-kleptomanic theft may be associated with impaired reinforcement learning, which would challenge punitive deterrence models. They argue for rehabilitation approaches that target these learning deficits frontiersin.org
Frontiers in Psychologyfrontiersin.orgOpen source ↗. "Our findings have clearly demonstrated that recurrent instrumental theft involves neurobehavioral deficits," Goto said. The kleptomanic group was small, 12 people, so conclusions about it are limited.
