ESTABLISHING THE FEASIBILITY OF CROWDSOURCING TO STUDY VALUE-BASED DECISION-MAKING IN SUBSTANCE USE DISORDERS

Substance use disorders (SUDs) are characterized by repeated choices to use drugs despite adverse consequences. Value-based decision-making theory posits that choices are guided by predicted values of available options which are updated based on outcomes. Computational modeling of reinforcement learning (RL) during choice tasks has shown promise for uncovering latent mechanisms of decision-making. Recruiting people with SUD is difficult, but crowdsourcing offers an alternative for accelerating data collection. However, the validity of computational modeling using online choice task data has not been well-established. This thesis aimed to use the online crowdsourcing platform Prolific to 1) compare online and in-person choice data from non-SUD controls and 2) survey SUD histories of individuals to inform selection of future study samples. Averaged choice and trial-by-trial RL outcomes were tested for similarities using equivalence testing and for differences using null hypothesis testing. Evidence of value-based decision-making was observed online and in-person, although differences on key choice and RL outcomes (e.g., rich choices, perseveration estimates) indicated nonequivalence between groups. In addition, people endorsing problematic use of alcohol, cannabis, and nicotine were available on Prolific. Overall, these results support the feasibility of using crowdsourced data for studying RL mechanisms of decision-making in people with SUDs.