Decision confidence as a function of multiple uncertainty judgements: An investigation of Bayesian inference in decision-making and learning
Lead Research Organisation:
University of Oxford
Department Name: Experimental Psychology
Abstract
Focusing on decision confidence, this project seeks to assess the independent and interactive effects of multiple uncertainties over the course of the learning period. Using a combination of behavioural tasks, Bayesian computational modelling, and with potential to extend to brain-imaging paradigms, the results of these studies will elucidate how information across different uncertainty levels is prioritised and whether different cognitive strategies, such as rule-based or category learning, are employed by individuals under uncertainty. This may shed light on the processes leading to the variability of reported confidence judgements and ultimately could inform the ongoing debate regarding the Bayesian optimality of human perception and decision-making.
Examination of everyday judgements reveals a multitude of factors which may influence an individuals' sense of decision-confidence, their belief that their choices are accurate. When judging the weather, an observant individual can learn that a clear sky is usually predictive of dry weather and that darker grey skies usually predict rainfall. Categorising based on cloud cover can provide a good estimate of the upcoming weather, however, it is not absolute. On a clear day, a sudden storm could occur or on a cloudy day it could remain dry. This simple judgement reflects the multiple types of uncertainty faced by an observer 1) a perceptual uncertainty in judging cloud level based on imperfect sensory measurements, 2) an uncertainty in terms of how strongly the level of cloud relates to a certain weather outcome and 3) how likely rain or sunshine are in this new environment over time. A naïve observer must grapple with these uncertainties, while simultaneously learning how this variation relates to outcomes of sunshine or rain, in settling on a course of action such as bringing an umbrella or leaving it at home. This kind of interaction between forms of uncertainty is, of yet, poorly defined in the literature, particularly over the course of the learning period, and may explain the reported variability of confidence judgements in the literature.
Examination of everyday judgements reveals a multitude of factors which may influence an individuals' sense of decision-confidence, their belief that their choices are accurate. When judging the weather, an observant individual can learn that a clear sky is usually predictive of dry weather and that darker grey skies usually predict rainfall. Categorising based on cloud cover can provide a good estimate of the upcoming weather, however, it is not absolute. On a clear day, a sudden storm could occur or on a cloudy day it could remain dry. This simple judgement reflects the multiple types of uncertainty faced by an observer 1) a perceptual uncertainty in judging cloud level based on imperfect sensory measurements, 2) an uncertainty in terms of how strongly the level of cloud relates to a certain weather outcome and 3) how likely rain or sunshine are in this new environment over time. A naïve observer must grapple with these uncertainties, while simultaneously learning how this variation relates to outcomes of sunshine or rain, in settling on a course of action such as bringing an umbrella or leaving it at home. This kind of interaction between forms of uncertainty is, of yet, poorly defined in the literature, particularly over the course of the learning period, and may explain the reported variability of confidence judgements in the literature.
Organisations
People |
ORCID iD |
| Nicholas Yeung (Primary Supervisor) | |
| Sarah Ashcroft-Jones (Student) |