Thomas Garcia

THE OPTIMALITY-ACCURACY TRADE-OFF IN DECISIONS UNDER UNCERTAINTY

When making a decision under uncertainty, individuals aim to achieve optimality. In general, an accurate decision is optimal. However, in many real life situations asymmetric stakes induce a divergence between optimality and accuracy (e.g. medical decision-making). We study this optimality-accuracy trade-off and its origins with experiments of perceptual decision allowing for the use of Signal Detection Theory (SDT) as a normative benchmark. The first experiment confirms the existence of an optimality-accuracy trade-off with a leading role of accuracy. Two different explanations of the observed trade-off are then proposed. The first explanation lies on an induced value of being right. The second one is based on a difference in attention allocation between the perceptual and payoff information. We develop an extension of SDT inclusive of this allocation process predicting the observed trade-off. This model is supported by an experiment inducing implicit changes in attention allocation.When making a decision under uncertainty, individuals aim to achieve optimality. In general, an accurate decision is optimal. However, in many real life situations asymmetric stakes induce a divergence between optimality and accuracy (e.g. medical decision-making). We study this optimality-accuracy trade-off and its origins with experiments of perceptual decision allowing for the use of Signal Detection Theory (SDT) as a normative benchmark. The first experiment confirms the existence of an optimality-accuracy trade-off with a leading role of accuracy. Two different explanations of the observed trade-off are then proposed. The first explanation lies on an induced value of being right. The second one is based on a difference in attention allocation between the perceptual and payoff information. We develop an extension of SDT inclusive of this allocation process predicting the observed trade-off. This model is supported by an experiment inducing implicit changes in attention allocation.