Biased or noisy? Motivational biases and decision noise across development

Abstract

Learning and decision-making undergo substantial developmental changes. In adolescents, both specific developmental changes of choice behavior, such as in motivational choice bias, and generally higher levels of decision noise have been observed. However, it remains unknown whether these two observations are independent or related. A not yet investigated possibility is that the specific development of motivational choice bias might depend on decision noise. We examined 93 participants (12 – 42 years) with a motivational Go/NoGo task, assessing ‘Pavlovian’ choice bias and disentangling it from instrumental learning biases. Participants performed two more reinforcement learning (RL) tasks to test cross-task generalization of computational parameters such as decision noise. Using mixed-effects models, we find an age-related increase in Pavlovian choice bias while instrumental learning bias did not change with age. Implementing a novel adaptation of a computational RL model with outcome-specific noise parameters (‘feedback sensitivity’) showed increases in Pavlovian choice bias and sensitivity for positive feedback with age. Beyond these within-task developmental age effects, noise levels are, firstly, strongly correlated across RL tasks and, secondly, mediate age dependent performance gain and more sophisticated RL processes, i.e., model-based control over choices. Taken together, our findings provide novel insights into the computational processes underlying developmental changes in decision-making: namely a vital role of seemingly unspecific changes in noise in the specific development of more complex learning and choice components. Studying the neurocomputational mechanisms of how varying levels of noise impact distinct aspects of learning and decision processes may also be key to better understand the developmental onset of psychiatric diseases.