In this short project, we propose algorithms to mitigate scoial bias of open polls. We do this by approximating open voting profiles to their theotical hidden counterparts and de-biasing the an open poll’s final outcome, hopefully maximizing social utility.

In this work, we incorporate existing tools for interpretability of machine learning models such as LIME, SHAP and MMD-critic to gain a better understading of AI models. These hybrids are then evaluated to detemine which provaids better explanations.

This work investigates possible mediators for developing cannabis use disorder (CUD). It takes a partial correlation approach for the analysis of trios with possible causal relationships and build on graph theory to deduce the true relationships and idetify mediators.