Value Preferences Estimation and Disambiguation in Hybrid Participatory Systems - code

Code for the paper "Value Preferences Estimation and Disambiguation in Hybrid Participatory Systems", under review at JAIR. This code contains five value preferences estimation methods based on participants' choices and textual justifications for a survey. Further, it implements an Active Learning pipeline with three sampling strategies---random, uncertainty, and driven by the estimation of value preferences. The pre-trained Dutch mode RobBERT is used to perform value classification.