{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/389409"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/389409","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"The clinical, computational, and neuropharmacological determinants of learning under uncertainty","abstract":"Uncertainty is ubiquitous in our interaction with the world, and hence adaptive behaviour relies on the appropriate estimation of such uncertainty. Associative learning tasks provide a framework within which researchers can induce different types of uncertainty in order to see how the participant responds to them, through learning under uncertainty paradigms. Computational modelling of participants responses in these paradigms allow researchers to infer the participants estimates of uncertainty. Given that issues with uncertainty underlie a plethora of neurodevelopmental and psychiatric conditions, studying individual differences in uncertainty estimation presents a potential opportunity for researchers to further the understanding of these conditions. This may also pave the way for further research into treatment for these conditions, through psychotherapy or pharmacotherapy, although little is known about the neural basis of uncertainty estimation. The focus of this thesis is on understanding the individual differences in uncertainty estimation and its pharmacological mechanisms, through computational modelling of associative learning paradigms. In this framework, associative learning is viewed as a process of statistical inference, where participants combine observations to infer an underlying contingency. Computational modelling allows experimenters to quantify estimates of the uncertainties in this process of inference for each participant. In other words, the view put forward in this thesis is that associative learning tasks provide computational assays of uncertainty estimates. This was explored initially by drawing parallels between commonly used models of associative learning paradigms, and deriving them from the rules of probability theory, advocating for the use of an individualized Bayesian learning model known as the Hierarchical Gaussian Filter. Subsequently, simple reinforcement learning models were used to understand the phenomenon of learning rate adaptation, whereby the speed of learning increases in a changeable environment This chapter also develops the approaches to computational modelling safeguards such as parameter recovery, model selection and validation, which were then used in more complex models in later chapters. Individual differences in uncertainty estimation are then examined by using Pavlovian conditioning tasks of both neutral and aversive valence, to test whether reported misestimation of uncertainty in individuals with anxiety and/or autism exist across learning domains or are specific to a particular valence. Estimates of aversive changeability (one form of uncertainty) derived using the Hierarchical Gaussian Filter were found to be elevated in anxious individuals. The relationship between anxiety and uncertainty is then investigated across diagnostic categories, in keeping with recent efforts to iii consider the transdiagnostic nature of mental health, and a factor of social withdrawal was found to relate to higher estimates of aversive changeability. This thesis then employed an aversive learning paradigm to explore the relationship between uncertainty and stress, known to be elevated in psychiatric conditions and implicated in their aetiology. Specifically, the hypothesis that computationally derived trial-by-trial estimates of uncertainty will explain stress ratings during the task – such that subjective uncertainty will drive higher stress – was tested. Results suggested that participants higher in transdiagnostic internalizing psychopathology scores reported higher average stress. Across all participants, subjective levels of surprise explained variation in stress ratings, such that stress ratings of higher performing participants were more influenced by their surprise estimates. The final chapter of this thesis investigated the pharmacological mechanisms of uncertainty estimation, specifically to test whether the noradrenergic system is important in estimating changeability or signalling high level changes in the associative environment. Results of an aversive learning task with a primary aversive outcome (electric shock) from a placebo-controlled drug study using atomoxetine are reported. Atomoxetine, a noradrenaline reuptake inhibitor, was found to exert baseline-dependent effects on estimates of changeability, offering evidence for the hypothesized role of noradrenaline in uncertainty representation. In summary, anxiety – whether through a diagnostic or transdiagnostic framework – is found to be associated with elevated estimates of aversive changeability. Subjective estimates of uncertainty are found to drive stress ratings, which can lead to adaptive performance. Evidence for the mechanism of aversive changeability representation by the noradrenergic system is also offered. This thesis concludes with the premise that computational modelling of learning under uncertainty has the potential to provide computational assays of uncertainty estimates, which can be used to deepen the understanding of psychiatric conditions and their potential treatment.","abstract_html":"Uncertainty is ubiquitous in our interaction with the world, and hence adaptive behaviour relies on the appropriate estimation of such uncertainty. Associative learning tasks provide a framework within which researchers can induce different types of uncertainty in order to see how the participant responds to them, through learning under uncertainty paradigms. Computational modelling of participants responses in these paradigms allow researchers to infer the participants estimates of uncertainty. Given that issues with uncertainty underlie a plethora of neurodevelopmental and psychiatric conditions, studying individual differences in uncertainty estimation presents a potential opportunity for researchers to further the understanding of these conditions. This may also pave the way for further research into treatment for these conditions, through psychotherapy or pharmacotherapy, although little is known about the neural basis of uncertainty estimation. The focus of this thesis is on understanding the individual differences in uncertainty estimation and its pharmacological mechanisms, through computational modelling of associative learning paradigms. In this framework, associative learning is viewed as a process of statistical inference, where participants combine observations to infer an underlying contingency. Computational modelling allows experimenters to quantify estimates of the uncertainties in this process of inference for each participant. In other words, the view put forward in this thesis is that associative learning tasks provide computational assays of uncertainty estimates. This was explored initially by drawing parallels between commonly used models of associative learning paradigms, and deriving them from the rules of probability theory, advocating for the use of an individualized Bayesian learning model known as the Hierarchical Gaussian Filter. Subsequently, simple reinforcement learning models were used to understand the phenomenon of learning rate adaptation, whereby the speed of learning increases in a changeable environment This chapter also develops the approaches to computational modelling safeguards such as parameter recovery, model selection and validation, which were then used in more complex models in later chapters. Individual differences in uncertainty estimation are then examined by using Pavlovian conditioning tasks of both neutral and aversive valence, to test whether reported misestimation of uncertainty in individuals with anxiety and/or autism exist across learning domains or are specific to a particular valence. Estimates of aversive changeability (one form of uncertainty) derived using the Hierarchical Gaussian Filter were found to be elevated in anxious individuals. The relationship between anxiety and uncertainty is then investigated across diagnostic categories, in keeping with recent efforts to iii consider the transdiagnostic nature of mental health, and a factor of social withdrawal was found to relate to higher estimates of aversive changeability. This thesis then employed an aversive learning paradigm to explore the relationship between uncertainty and stress, known to be elevated in psychiatric conditions and implicated in their aetiology. Specifically, the hypothesis that computationally derived trial-by-trial estimates of uncertainty will explain stress ratings during the task – such that subjective uncertainty will drive higher stress – was tested. Results suggested that participants higher in transdiagnostic internalizing psychopathology scores reported higher average stress. Across all participants, subjective levels of surprise explained variation in stress ratings, such that stress ratings of higher performing participants were more influenced by their surprise estimates. The final chapter of this thesis investigated the pharmacological mechanisms of uncertainty estimation, specifically to test whether the noradrenergic system is important in estimating changeability or signalling high level changes in the associative environment. Results of an aversive learning task with a primary aversive outcome (electric shock) from a placebo-controlled drug study using atomoxetine are reported. Atomoxetine, a noradrenaline reuptake inhibitor, was found to exert baseline-dependent effects on estimates of changeability, offering evidence for the hypothesized role of noradrenaline in uncertainty representation. In summary, anxiety – whether through a diagnostic or transdiagnostic framework – is found to be associated with elevated estimates of aversive changeability. Subjective estimates of uncertainty are found to drive stress ratings, which can lead to adaptive performance. Evidence for the mechanism of aversive changeability representation by the noradrenergic system is also offered. This thesis concludes with the premise that computational modelling of learning under uncertainty has the potential to provide computational assays of uncertainty estimates, which can be used to deepen the understanding of psychiatric conditions and their potential treatment.","abstract_has_math":false,"creators":["Sandhu, Tim"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Rowe, James","Lawson, Rebecca"],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-02-03","date_published":"2025-02-03","updated_at":"2026-07-22T22:24:27Z","subjects":["computational psychiatry"],"languages":["eng"],"rights":[],"rights_urls":["https://www.repository.cam.ac.uk/bitstreams/0cc40c01-86b0-4217-9bec-26db5ff064af/download","http://purl.org/NET/rdflicense/allrightsreserved"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.121305","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Rowe, James","Lawson, Rebecca"]},{"key":"dc:creator","label":"Author","values":["Sandhu, Tim"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2025-02-03"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/389409"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["computational psychiatry"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["https://www.repository.cam.ac.uk/bitstreams/0cc40c01-86b0-4217-9bec-26db5ff064af/download","http://purl.org/NET/rdflicense/allrightsreserved"]},{"key":"dc:rights.embargodate","label":"Dc Rights Embargodate","values":["2026-09-15"]},{"key":"dc:rights.embargotype","label":"Dc Rights Embargotype","values":["embargo"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.121305"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://www.repository.cam.ac.uk/bitstreams/59d39335-53cd-4b51-aa8e-f3be7994a79b/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Uncertainty is ubiquitous in our interaction with the world, and hence adaptive behaviour relies on the appropriate estimation of such uncertainty. Associative learning tasks provide a framework within which researchers can induce different types of uncertainty in order to see how the participant responds to them, through learning under uncertainty paradigms. Computational modelling of participants responses in these paradigms allow researchers to infer the participants estimates of uncertainty. Given that issues with uncertainty underlie a plethora of neurodevelopmental and psychiatric conditions, studying individual differences in uncertainty estimation presents a potential opportunity for researchers to further the understanding of these conditions. This may also pave the way for further research into treatment for these conditions, through psychotherapy or pharmacotherapy, although little is known about the neural basis of uncertainty estimation. The focus of this thesis is on understanding the individual differences in uncertainty estimation and its pharmacological mechanisms, through computational modelling of associative learning paradigms. In this framework, associative learning is viewed as a process of statistical inference, where participants combine observations to infer an underlying contingency. Computational modelling allows experimenters to quantify estimates of the uncertainties in this process of inference for each participant. In other words, the view put forward in this thesis is that associative learning tasks provide computational assays of uncertainty estimates. This was explored initially by drawing parallels between commonly used models of associative learning paradigms, and deriving them from the rules of probability theory, advocating for the use of an individualized Bayesian learning model known as the Hierarchical Gaussian Filter. Subsequently, simple reinforcement learning models were used to understand the phenomenon of learning rate adaptation, whereby the speed of learning increases in a changeable environment This chapter also develops the approaches to computational modelling safeguards such as parameter recovery, model selection and validation, which were then used in more complex models in later chapters. Individual differences in uncertainty estimation are then examined by using Pavlovian conditioning tasks of both neutral and aversive valence, to test whether reported misestimation of uncertainty in individuals with anxiety and/or autism exist across learning domains or are specific to a particular valence. Estimates of aversive changeability (one form of uncertainty) derived using the Hierarchical Gaussian Filter were found to be elevated in anxious individuals. The relationship between anxiety and uncertainty is then investigated across diagnostic categories, in keeping with recent efforts to iii consider the transdiagnostic nature of mental health, and a factor of social withdrawal was found to relate to higher estimates of aversive changeability. This thesis then employed an aversive learning paradigm to explore the relationship between uncertainty and stress, known to be elevated in psychiatric conditions and implicated in their aetiology. Specifically, the hypothesis that computationally derived trial-by-trial estimates of uncertainty will explain stress ratings during the task – such that subjective uncertainty will drive higher stress – was tested. Results suggested that participants higher in transdiagnostic internalizing psychopathology scores reported higher average stress. Across all participants, subjective levels of surprise explained variation in stress ratings, such that stress ratings of higher performing participants were more influenced by their surprise estimates. The final chapter of this thesis investigated the pharmacological mechanisms of uncertainty estimation, specifically to test whether the noradrenergic system is important in estimating changeability or signalling high level changes in the associative environment. Results of an aversive learning task with a primary aversive outcome (electric shock) from a placebo-controlled drug study using atomoxetine are reported. Atomoxetine, a noradrenaline reuptake inhibitor, was found to exert baseline-dependent effects on estimates of changeability, offering evidence for the hypothesized role of noradrenaline in uncertainty representation. In summary, anxiety – whether through a diagnostic or transdiagnostic framework – is found to be associated with elevated estimates of aversive changeability. Subjective estimates of uncertainty are found to drive stress ratings, which can lead to adaptive performance. Evidence for the mechanism of aversive changeability representation by the noradrenergic system is also offered. This thesis concludes with the premise that computational modelling of learning under uncertainty has the potential to provide computational assays of uncertainty estimates, which can be used to deepen the understanding of psychiatric conditions and their potential treatment."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["038ce002e64a02dfb26e54d399bb1079","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["The clinical, computational, and neuropharmacological determinants of learning under uncertainty"]}]}],"canonical_facts":{"dc:contributor.advisor":["Rowe, James","Lawson, Rebecca"],"dc:creator":["Sandhu, Tim"],"dc:date.issued":["2025-02-03"],"dc:description.abstract":["Uncertainty is ubiquitous in our interaction with the world, and hence adaptive behaviour relies on the appropriate estimation of such uncertainty. Associative learning tasks provide a framework within which researchers can induce different types of uncertainty in order to see how the participant responds to them, through learning under uncertainty paradigms. Computational modelling of participants responses in these paradigms allow researchers to infer the participants estimates of uncertainty. Given that issues with uncertainty underlie a plethora of neurodevelopmental and psychiatric conditions, studying individual differences in uncertainty estimation presents a potential opportunity for researchers to further the understanding of these conditions. This may also pave the way for further research into treatment for these conditions, through psychotherapy or pharmacotherapy, although little is known about the neural basis of uncertainty estimation. The focus of this thesis is on understanding the individual differences in uncertainty estimation and its pharmacological mechanisms, through computational modelling of associative learning paradigms. In this framework, associative learning is viewed as a process of statistical inference, where participants combine observations to infer an underlying contingency. Computational modelling allows experimenters to quantify estimates of the uncertainties in this process of inference for each participant. In other words, the view put forward in this thesis is that associative learning tasks provide computational assays of uncertainty estimates. This was explored initially by drawing parallels between commonly used models of associative learning paradigms, and deriving them from the rules of probability theory, advocating for the use of an individualized Bayesian learning model known as the Hierarchical Gaussian Filter. Subsequently, simple reinforcement learning models were used to understand the phenomenon of learning rate adaptation, whereby the speed of learning increases in a changeable environment This chapter also develops the approaches to computational modelling safeguards such as parameter recovery, model selection and validation, which were then used in more complex models in later chapters. Individual differences in uncertainty estimation are then examined by using Pavlovian conditioning tasks of both neutral and aversive valence, to test whether reported misestimation of uncertainty in individuals with anxiety and/or autism exist across learning domains or are specific to a particular valence. Estimates of aversive changeability (one form of uncertainty) derived using the Hierarchical Gaussian Filter were found to be elevated in anxious individuals. The relationship between anxiety and uncertainty is then investigated across diagnostic categories, in keeping with recent efforts to iii consider the transdiagnostic nature of mental health, and a factor of social withdrawal was found to relate to higher estimates of aversive changeability. This thesis then employed an aversive learning paradigm to explore the relationship between uncertainty and stress, known to be elevated in psychiatric conditions and implicated in their aetiology. Specifically, the hypothesis that computationally derived trial-by-trial estimates of uncertainty will explain stress ratings during the task – such that subjective uncertainty will drive higher stress – was tested. Results suggested that participants higher in transdiagnostic internalizing psychopathology scores reported higher average stress. Across all participants, subjective levels of surprise explained variation in stress ratings, such that stress ratings of higher performing participants were more influenced by their surprise estimates. The final chapter of this thesis investigated the pharmacological mechanisms of uncertainty estimation, specifically to test whether the noradrenergic system is important in estimating changeability or signalling high level changes in the associative environment. Results of an aversive learning task with a primary aversive outcome (electric shock) from a placebo-controlled drug study using atomoxetine are reported. Atomoxetine, a noradrenaline reuptake inhibitor, was found to exert baseline-dependent effects on estimates of changeability, offering evidence for the hypothesized role of noradrenaline in uncertainty representation. In summary, anxiety – whether through a diagnostic or transdiagnostic framework – is found to be associated with elevated estimates of aversive changeability. Subjective estimates of uncertainty are found to drive stress ratings, which can lead to adaptive performance. Evidence for the mechanism of aversive changeability representation by the noradrenergic system is also offered. This thesis concludes with the premise that computational modelling of learning under uncertainty has the potential to provide computational assays of uncertainty estimates, which can be used to deepen the understanding of psychiatric conditions and their potential treatment."],"dc:format.checksum.md5":["038ce002e64a02dfb26e54d399bb1079","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.121305"],"dc:identifier.uri":["https://www.repository.cam.ac.uk/bitstreams/59d39335-53cd-4b51-aa8e-f3be7994a79b/download"],"dc:language":["eng"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/389409"],"dc:rights":["https://www.repository.cam.ac.uk/bitstreams/0cc40c01-86b0-4217-9bec-26db5ff064af/download","http://purl.org/NET/rdflicense/allrightsreserved"],"dc:rights.embargodate":["2026-09-15"],"dc:rights.embargotype":["embargo"],"dc:subject":["computational psychiatry"],"dc:title":["The clinical, computational, and neuropharmacological determinants of learning under uncertainty"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:27Z"}