Skip to content

Latest commit

 

History

History
2 lines (2 loc) · 311 Bytes

README.md

File metadata and controls

2 lines (2 loc) · 311 Bytes

avoidance_learning

This is an avoidance learning paradigm adapted from the relative value task from Palminteri et al. (2015) and the task code from [the demo tasks from Yael Niv's lab] (https://github.com/nivlab/jspsych-demos/tree/main/tasks/rvl). This task has also been converted from jsPsych6 to jsPsych7.