The impact of glucose on Bayesian v. heuristic-based decision making

We examine the impact of glucose in a choice task that can distinguish Bayesian from lowerlevel reinforcement heuristic choice. Drawing from a dual systems framework, we hypothesize that glucose administration will increase response times and improve Bayesian accuracy because it should shift decision making towards the more deliberate system 2 and away from the more automatic system 1 decision process. We study 113 subjects randomly assigned to either a glucose or placebo drink condition, who make choices over several incentivized easy and difficult choices of the Bayesian task. Our results indicate a significant and robust glucose effect on response times. Glucose administration has a main effect of increasing response time, as predicted, but glucose also improves the marginal decrease in response times across trials. Regarding Bayesian accuracy, we analyze subject choice possibilities as states satisfying the properties of a regular discrete Markov chain (choices may be Bayesian, Reinforcement, or also Naïve). We calculate steady-state Markov probabilities and show that glucose increases the likelihood of making a Bayesian choice over a heuristic-based choice by up to 9%. These results suggest a beneficial impact of glucose on deliberative decision making.