Rewards, rather than punishments, help teenagers learn

What have you been working on?

We used a combination of lab-based computer tasks and computational learning models to compare how adolescents and adults learn to make choices, based on different types of feedback.

Adolescent and adult volunteers played a computer task, in which they saw different pairs of abstract symbols on a computer screen, and had to choose one by pressing a button. The symbol they chose could either result in a reward (winning a point), a punishment (losing a point), or no outcome. Participants all wanted to get as many points as possible, as they could earn up to £10 for a high score.

Within each pair, one symbol was more likely to be associated with a good outcome than the other. However, at the start of the task, participants didn’t know which symbol was which and had to try to learn which was more likely to result in a good outcome through trial and error based on feedback. Sometimes, participants would also find out what would have happened if they’d chosen the other option.

We also used computational models to help to interpret our results. Different learning processes can be modelled using mathematical models, which simulate how someone would play the game if they were using those specific learning processes. This is important, as in everyday life, decisions are not always associated with either a good or bad outcome — we have to learn through experience.

What did you find?

We found that adolescents and adults were equally good at learning from the rewards which options would result in the best outcomes. However, adolescents were not as good at using the bad outcomes (punishments) to guide their future choices. We also found that while adults showed improvements in learning when they saw the outcome of the other option, adolescents were less likely to use this information to guide their future choices.

What are the implications for schools?

Our data suggests that when rewards and punishments are equal in value (e.g., winning a point compared with losing a point), adolescents are more likely to take the rewarding information into account in future choices than the punishment. Therefore, in some cases, positive feedback may have more of an effect than negative feedback on learning.

However, this type of study where participants are learning from abstract symbols in a quiet environment without other things going on is very different to learning in the classroom, where many different factors are involved. Studies of classroom-based learning would need to be done before any real-life implications could be asserted.

What do you hope the impact will be?

We hope our study will draw attention to the fact that learning and decision-making don’t rely on a single, unitary system, but instead involve the co-ordination of a range of different processes, which may show different patterns of development both in adolescence and across the lifespan. For example, rewarding and punishing feedback do not necessarily rely on the same systems or have the same effect on behaviour in everyone, but may differ according to variation between individuals, including, but not limited to, their age.

The Computational Development of Reinforcement Learning during Adolescence was published in PLOS Computational Biology

Your thoughts

Leave a Reply

Your email address will not be published.