Table 1 Overview of the learning models. The initial value of the risky option \(\left( {{\text{Q}}_{1} } \right)\) was a free parameter in all models as well (not included in the table).
From: Impaired learning to dissociate advantageous and disadvantageous risky choices in adolescents
Parameters | ||
|---|---|---|
Reinforcement learning (1) | Bayesian ideal-observer (2) | |
Basic models (A) | Learning rate α | Update rate π |
Asymmetric learning (B): Stronger weighting of win outcomes promotes risk seeking | Learning rates for win and no-win outcomes: α+ and α− | Update rates for win and no-win outcomes: π+ and π− |
Nonlinear utility function (C): Overvaluation of higher outcomes promotes risk seeking | Utility parameter \({\upkappa }\) \({\upkappa }\) > 1 and \({\upkappa }\) < 1 cause over- and undervaluation of higher outcomes, respectively | Utility parameter \({\upkappa }\) |
Uncertainty affects value (D): Uncertainty bonus promotes risk seeking | Not applicable | Uncertainty parameter \({{\varphi }}\) \({{\varphi }}\) > 0 and \({{\varphi }}\) < 0 cause uncertainty bonus and penalty, respectively |
