Melbourne School of Psychological Sciences - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemBelief updating and information seeking in decision making under uncertainty( 2017)Adaptive goal-directed behaviour depends on a well-calibrated internal model of the environment. In particular, learning and decision making is most efficient when there is a good match between an agent’s model of the environment and true environmental contingencies. Accordingly, recent theories in cognitive neuroscience have proposed that a primary computational goal of the central nervous system in humans and other animals is the construction and maintenance of these internal models (also termed beliefs). The present thesis aimed to investigate the implications of this belief-centric perspective for the study of decision making under uncertainty, using computational modelling of behaviour in concert with analysis of human electroencephalography (EEG) data. Across one behavioural and three EEG studies, this thesis addressed two related questions concerning participants’ updating of beliefs. Study 1 and Study 2 sought to characterise the neural and cognitive processes by which participants incorporated new information into their beliefs (belief updating). Study 3 and Study 4 sought to investigate the nature of participants’ preferences for acquiring such information, and the neural substrates of this preference (information seeking). In Study 1, 18 healthy young adult participants completed a perceptual learning task—in which participants learned to identify the target contrast of a greyscale checkerboard stimulus using monetary feedback—while EEG was recorded. Formal comparison of computational cognitive models revealed that participants’ behaviour in this task was better explained by a model implementing Bayesian belief updating than by a model implementing a simpler win-stay lose-shift (WSLS) heuristic. Belief variables computed from this Bayesian model were then used as predictors in a single-trial regression analysis of event-related potential (ERP) data. This analysis revealed that the amplitude of the frontocentral P3 component of the ERP was positively associated with belief update size, and that the amplitude of the stimulus-preceding negativity component was negatively associated with belief uncertainty. These results provided evidence that belief update size and belief uncertainty had distinct neural signatures that could be tracked in single trials in specific ERP components. The results further suggested that the cognitive mechanisms underlying belief updating in this task could be described well within a Bayesian framework. Study 2 investigated the effect of task motivation on participants’ use of Bayesian versus heuristic task strategies, and upon the neural substrates of belief updating. Using a modified variant of the perceptual learning task from Study 1, Study 2 presented feedback to participants in the form of either monetary reward, or as affectively neutral instructional directives. Using model-based clustering based upon formal comparison of computational cognitive models, Study 2 identified two distinct participant subgroups. The first subgroup used Bayesian inference in the monetary condition, but switched to a heuristic strategy in the instructive feedback condition; by contrast, the second subgroup always used the Bayesian inference strategy, regardless of feedback condition. It was found that only the strategy-switching subgroup showed worse performance for instructive than for monetary feedback, whereas the Bayesian subgroup did not. This pattern of performance was reflected by similar differences between subgroups in neural encoding of feedback in two components of the event-related potential: the P3, and the late positive potential. These findings suggested that selection of Bayesian versus heuristic strategies in perceptual learning may depend critically on participants’ motivational state, and that individual differences in strategy-switching may underlie group-level differences in neural encoding of feedback. Study 3 investigated information seeking behaviour in two experiments, each with 40 healthy participants. In these experiments, participants completed a novel information-seeking task in which they could choose to pay a monetary cost to receive advance information about the outcome of a monetary lottery. Importantly, acquiring information did not alter lottery outcome probabilities. In the first experiment, it was found that participants were willing to incur considerable monetary costs to acquire payoff-irrelevant information about the lottery outcome. This behaviour was well explained by a computational cognitive model in which information preference resulted from aversion to temporally prolonged uncertainty. In the second experiment, it was found that preference for information was modulated by the rate of information delivery. These results strongly suggest that humans assign an intrinsic value to information, in a manner inconsistent with normative accounts of decision making under uncertainty. Finally, Study 4 investigated the neural substrates of the intrinsic valuation of information, using EEG data were recorded from 22 participants performing the information seeking task developed in Study 3. Behavioural results replicated the findings of Study 3. Analysis of ERPs elicited by informative cues revealed that the feedback-related negativity, an ERP component linked to reward processing, independently encoded both a reward prediction error and an information prediction error. These findings are consistent with the hypothesis that information seeking results from processing of information within neural reward circuits, and provide further evidence for an intrinsic valuation of information. Overall, these findings demonstrate that a belief-centric research perspective has substantial explanatory power for behavioural and neural data in human decision making under uncertainty.