Modeling changes in probabilistic reinforcement learning during adolescence.

Area of Science:

• Cognitive psychology
• Neuroscience
• Developmental psychology

Background:

• Real-world events often involve uncertainty and probabilistic relationships.
• Youth experience more uncertainty due to less life experience compared to adults.
• Previous research shows mixed findings on probabilistic learning efficiency in adolescents.

Purpose of the Study:

• To investigate age-related differences in probabilistic reinforcement learning.
• To identify factors influencing learning rate and decision-making in youth and adults.
• To explore the relationship between testosterone levels and learning in mid-adolescence.

Main Methods:

• Utilized a probabilistic reinforcement learning task with 187 youths (ages 8-17) and 110 adults (ages 18-30).
• Employed hierarchical Bayesian methods to fit computational reinforcement learning models.
• Analyzed the impact of negative outcomes and age on learning performance.

Main Results:

• Performance in probabilistic learning improved with age until the early twenties, then plateaued.
• Learning rate increased (integration time scale decreased) and noisy choices decreased with age.
• Models where negative outcomes minimally impacted learning best explained performance across all participants.
• Positive correlation observed between salivary testosterone and learning rate in mid-adolescents (ages 13-15).

Conclusions:

• Probabilistic learning capabilities mature through adolescence and into early adulthood.
• Developmental changes in learning rate and decision strategy underlie performance improvements.
• Hormonal factors like testosterone may influence learning during mid-adolescence, warranting further investigation into adolescent brain development and learning.