Better decisions through reinforcement learning
Decision-making relies on the interaction of several processes, including representation of value, response selection and learning. In adolescence, decisions can often be characterised as being impulsive and high risk with serious consequences. With reinforcement learning, the processes are integrated. It deals with learning to improve one's future choices in order to maximise the occurrence of pleasant events (rewards) and minimise the occurrence of unpleasant events (punishments). Understanding the process that underlies decision-making in adolescence is important, yet the development of reinforcement learning in human adolescence has been studied only recently. DEVELOP-LEARNING (Developmental trajectories for model-free and model-based reinforcement learning: Computational and neural bases) was an EU-funded project devoted to changing this by conducting behaviour experiments. The first experiment involved administrating a novel instrumental learning task involving basic reward and punishment learning, as well as learning from counterfactual (fictive) information to adults and adolescents. It was found that adolescents' performance was not enhanced by counterfactual information; and, compared to adults, adolescents learned preferentially from reward compared to punishment, whereas adults learned equally from both. Results showed that adults and adolescents did not use the same algorithm to solve the learning task. Unlike in adults, adolescents' performance did not take into account counterfactual information. Additionally, adolescents learned to seek rewards rather than to avoid punishments. Adults, on the other hand, learned to seek and avoid both equally. Increasing understanding of computational changes in reinforcement learning during adolescence may provide insights into adolescent value-based decision-making. Results can also have a potential impact on education, since they suggest that adolescents benefit more from positive feedback than from negative feedback in learning tasks.
