Deconstructing appetitive learning in cocaine addiction using hierarchical Bayesian modelling
Abstract:
Background: Decision making in cocaine-addicted patients has frequently
been described as suboptimal and risky. Since decisions are usually made with a
view to a favourable out- come, rewards provide the motivation to make decisions
and cognition is necessary to appraise the available options and evaluate the
consequences involved with each choice [1]. Disadvantageous decision making in
cocaine addiction may be related to impairments in either reward process- ing or
goal-directed learning, or both. Computational mod- elling approaches provide a
method to elucidate the nature of aberrant decision making by deconstructing the
processes of choice selection. Reinforcement learning is a widely used framework
that defines key parameters that underlie op- timal learning [2], for example
the extent to which prior experience influences choice selection (learning
rate), per- severative responding (‘stickiness’ to options presented on one
side), extinction to non-reinforced choices (extinction rate) and sensitivity to
the value of the reinforcer (rein- forcement sensitivity).