Dataset Information

Risk, unexpected uncertainty, and estimation uncertainty: Bayesian learning in unstable settings.

ABSTRACT: Recently, evidence has emerged that humans approach learning using Bayesian updating rather than (model-free) reinforcement algorithms in a six-arm restless bandit problem. Here, we investigate what this implies for human appreciation of uncertainty. In our task, a Bayesian learner distinguishes three equally salient levels of uncertainty. First, the Bayesian perceives irreducible uncertainty or risk: even knowing the payoff probabilities of a given arm, the outcome remains uncertain. Second, there is (parameter) estimation uncertainty or ambiguity: payoff probabilities are unknown and need to be estimated. Third, the outcome probabilities of the arms change: the sudden jumps are referred to as unexpected uncertainty. We document how the three levels of uncertainty evolved during the course of our experiment and how it affected the learning rate. We then zoom in on estimation uncertainty, which has been suggested to be a driving force in exploration, in spite of evidence of widespread aversion to ambiguity. Our data corroborate the latter. We discuss neural evidence that foreshadowed the ability of humans to distinguish between the three levels of uncertainty. Finally, we investigate the boundaries of human capacity to implement Bayesian learning. We repeat the experiment with different instructions, reflecting varying levels of structural uncertainty. Under this fourth notion of uncertainty, choices were no better explained by Bayesian updating than by (model-free) reinforcement learning. Exit questionnaires revealed that participants remained unaware of the presence of unexpected uncertainty and failed to acquire the right model with which to implement Bayesian updating.

SUBMITTER: Payzan-LeNestour E

PROVIDER: S-EPMC3024253 | biostudies-literature | 2011 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Risk, unexpected uncertainty, and estimation uncertainty: Bayesian learning in unstable settings.

Payzan-LeNestour Elise E Bossaerts Peter P

PLoS computational biology 20110120 1

Recently, evidence has emerged that humans approach learning using Bayesian updating rather than (model-free) reinforcement algorithms in a six-arm restless bandit problem. Here, we investigate what this implies for human appreciation of uncertainty. In our task, a Bayesian learner distinguishes three equally salient levels of uncertainty. First, the Bayesian perceives irreducible uncertainty or risk: even knowing the payoff probabilities of a given arm, the outcome remains uncertain. Second, th ...[more]

PMID: 21283774

Similar Datasets

Project description:A two-layer model based on the integrated form of Richards' equation (RE) was recently developed to simulate the soil water movement in the roots layer and the vadose zone with a relatively shallow and dynamic water table. The model simulates thickness-averaged volumetric water content and matric suction as opposed to point values and was numerically verified for three soil textures using HYDRUS as a benchmark. However, the strengths and limitations of the two-layer model and its performance in stratified soils and under actual field conditions have not been tested. This study further examined the two-layer model using two numerical verification experiments and, most importantly, tested its performance at site-level under actual, highly variable hydroclimate conditions. Moreover, model parameters were estimated and uncertainty and sources of errors were quantified using a Bayesian framework. First, the two-layer model was evaluated for 231 soil textures under varying soil layer thicknesses with a uniform soil profile. Second, the two-layer model was assessed for stratified conditions where the top and bottom soil layers have contrasting hydraulic conductivities. The model was evaluated by comparing soil moisture and flux estimates to those from the HYDRUS model. Last, a case study of model application using data from a Soil Climate Analysis Network (SCAN) site was presented. Bayesian Monte Carlo (BMC) method was implemented for model calibration and quantifying sources of uncertainty under real hydroclimate and soil conditions. For a homogeneous soil profile, the two-layer model generally had excellent performance in estimating volumetric water content and fluxes, while the model performance slightly declined with increasing layer thickness and coarser textured soils. The model configurations regarding layer thicknesses and soil textures that generate accurate soil moisture and flux estimations were further suggested. With the two layers of contrasting permeability, model-simulated soil moisture contents and fluxes agreed well with those computed by HYDRUS, indicating that the two-layer model accurately handles the water flow dynamics around the layer interface. In the field application, given the highly variable hydroclimate conditions, the two-layer model combined with the BMC method showed good agreement with the observed average soil moisture of the root zone and the vadose zone below (RMSE <0.021 during calibration and <0.023 during validation periods). The contribution of parametric uncertainty to the total model uncertainty was too small compared to other sources. The numerical tests and the site level application showed that the two-layer model can reliably simulate thickness-averaged soil moisture and estimate fluxes in the vadose zone under various soil and hydroclimate conditions. Results also indicated that the BMC method could be a robust framework for vadose zone hydraulic parameters identification and model uncertainty estimation.

Dataset Information

Risk, unexpected uncertainty, and estimation uncertainty: Bayesian learning in unstable settings.

Publications

Risk, unexpected uncertainty, and estimation uncertainty: Bayesian learning in unstable settings.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets