Designing Interventions To Mitigate Cognitive Biases In Human Decisions
Naeem Khozaey Akl, University of Texas at Austin
Ahmed Tewfik, University of Texas at Austin

Human decision-making is adversely affected by cognitive biases. In this paper we construct an example that in- corporates an anchor-and-adjustment model for belief-updating when information is presented sequentially to a human. We prove that different orderings of the same information pieces impact the human decision-making by deriving closed-form expressions of the corresponding error probabilities. We then backpropagate the error for one selected ordering of the information and present two strategies to reduce the order-effects and improve the decision-making in presence of cognitive biases: either modify the information presented to the human taking into account history and context or teach the human to modify the decision boundaries for the given ordering of the observations. The methods generalize to more general settings.