Oliver James Hulme
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Oliver James Hulme has a total of 172 co-authors. Here are the top co-authors by number of shared publications:
Hartwig R. Siebner
Affiliations Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital Amager and Hvidovre, Copenhagen, Denmark, Department of Neurology, Copenhagen University Hospital Bispebjerg, Copenhagen, Denmark, Institute for Clinical Medicine, University of Copenhagen, Copenhagen, Denmark
12 shared publications- Kristoffer Hougaard Madsen
Technical University of Denmark
10 shared publications Colm Connaughton
Affiliations London Mathematical Laboratory, London, United Kingdom, Warwick Mathematics Institute and Centre for Complexity Science, University of Warwick, Coventry, United Kingdom
9 shared publicationsOle Peters
Affiliations London Mathematical Laboratory, London, United Kingdom, Santa Fe Institute, Santa Fe, United States of America
9 shared publicationsBenjamin Skjold
8 shared publications- Hartwig Roman Siebner
Danish Research Centre for Magnetic Resonance, Department of Radiology and Nuclear Medicine, Copenhagen University Hospital - Amager and Hvidovre
8 shared publications
Latest Publications
- Open Access
- 17 August 2026
- Published by Center for Open Science in PsyArXiv
Ergodicity and the speed-accuracy trade-off: Why optimal response times depend on reward dynamics
In the cognitive and behavioural sciences, speed-accuracy trade-offs arise when longer response times improve accuracy but delay payoff. Greater accuracy can increase reward probability or magnitude, but time dilutes the rate at which rewards are obtained. This raises the question: what is optimised when time is exchanged for accuracy? Here we recast speed-accuracy trade-offs as a time-optimality problem in which an agent maximises the time-average growth rate of their resources. In additive environments, the relevant growth rate is the expected increment of their resources per unit time, but in multiplicative environments, it is the expected log-increment per unit time. We show that even with identical speed-accuracy curves, shifting from additive to multiplicative dynamics increases the time-optimal response time (due to volatility drag, which makes it worth waiting longer for higher accuracy). We derive these effects analytically and use simulations to connect them to empirically-inspired speed-accuracy curves. Non-decision latencies and inter-trial intervals further increase the time-optimal response times. Overall, we show that theories based on maximising the expected reward rate are suboptimal in some environments, and instead, maximising the time-average growth rate provides a more generalnormative principle. This framework yields falsifiable predictions applicable across motor, cognitive, and perceptual domains.
- 17 August 2026
- Published by Center for Open Science in PsyArXiv
Open AccessErgodicity and the speed-accuracy trade-off: Why optimal response times depend on reward dynamics
In the cognitive and behavioural sciences, speed-accuracy trade-offs arise when longer response times improve accuracy but delay payoff. Greater accuracy can increase reward probability or magnitude, but time dilutes the rate at which rewards are obtained. This raises the question: what is optimised when time is exchanged for accuracy? Here we recast speed-accuracy trade-offs as a time-optimality problem in which an agent maximises the time-average growth rate of their resources. In additive environments, the relevant growth rate is the expected increment of their resources per unit time, but in multiplicative environments, it is the expected log-increment per unit time. We show that even with identical speed-accuracy curves, shifting from additive to multiplicative dynamics increases the time-optimal response time (due to volatility drag, which makes it worth waiting longer for higher accuracy). We derive these effects analytically and use simulations to connect them to empirically-inspired speed-accuracy curves. Non-decision latencies and inter-trial intervals further increase the time-optimal response times. Overall, we show that theories based on maximising the expected reward rate are suboptimal in some environments, and instead, maximising the time-average growth rate provides a more generalnormative principle. This framework yields falsifiable predictions applicable across motor, cognitive, and perceptual domains.
- Open Access
- 20 July 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (7) , e1014409
Ergodicity transformations predict human decision-making under risk
Decision theories commonly model human behavior as maximizing the expected value of a utility function. This function may vary from one person to another but is assumed to be stable over time. Recent theoretical developments demonstrate that these assumptions are generally incompatible with growing wealth at the fastest rate. Growth optimality requires utility functions to mirror ergodicity transformations and adapt to the dynamic environment. We exposed human participants to different wealth dynamics in a consequential risky decision-making experiment. Via Bayesian modelling, we estimated utility functions separately for each dynamic. Pre-registered analyses revealed strong evidence supporting the quantitative predictions of the ergodicity model. Our study provides evidence that human risk-taking can adapt quickly to the dynamical context, in ways that align closely to the theoretical optimum for maximizing wealth over time.
- 20 July 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (7) , e1014409
Open AccessErgodicity transformations predict human decision-making under risk
Decision theories commonly model human behavior as maximizing the expected value of a utility function. This function may vary from one person to another but is assumed to be stable over time. Recent theoretical developments demonstrate that these assumptions are generally incompatible with growing wealth at the fastest rate. Growth optimality requires utility functions to mirror ergodicity transformations and adapt to the dynamic environment. We exposed human participants to different wealth dynamics in a consequential risky decision-making experiment. Via Bayesian modelling, we estimated utility functions separately for each dynamic. Pre-registered analyses revealed strong evidence supporting the quantitative predictions of the ergodicity model. Our study provides evidence that human risk-taking can adapt quickly to the dynamical context, in ways that align closely to the theoretical optimum for maximizing wealth over time.
- Open Access
- 22 June 2026
- Published by Elsevier in Metabolism
- Vol. 182, 156683
Ghrelin predicts food consumption up to three hours prior to eating: A randomised controlled trial
- 22 June 2026
- Published by Elsevier in Metabolism
- Vol. 182, 156683
Open AccessGhrelin predicts food consumption up to three hours prior to eating: A randomised controlled trial
- Open Access
- 4 March 2026
- Published by MIT Press in Imaging Neuroscience
- Vol. 4
Computational parametric mapping of functional neuroimaging data
Elucidating the neural basis of cognition requires theoretical models of cognition to constrain the modeling of neural data. A prevalent strategy in functional neuroimaging is to regress the latent variables of cognitive models onto neural data. Though widely used, this approach restricts the mapping of computational variables to single parameter values. We introduce computational parametric mapping (CPM), which builds on and generalizes the Bayesian population receptive field framework. CPM offers three main advances for cognitive computational modeling. First, it allows the fitting of cognitive models directly to neuroimaging data. Second, it allows for voxelwise or regionwise mapping of parameters of cognitive computational models onto the brain, thus making the topographic mapping methods prevalent in the sensory sciences available to the cognitive computational neuroscientist. Finally, it is efficient enough to make voxelwise mapping over large regions of interest feasible. Here, we illustrate how CPM can be used to fit reinforcement-learning algorithms to synthetic and real data.
- 4 March 2026
- Published by MIT Press in Imaging Neuroscience
- Vol. 4
Open AccessComputational parametric mapping of functional neuroimaging data
Elucidating the neural basis of cognition requires theoretical models of cognition to constrain the modeling of neural data. A prevalent strategy in functional neuroimaging is to regress the latent variables of cognitive models onto neural data. Though widely used, this approach restricts the mapping of computational variables to single parameter values. We introduce computational parametric mapping (CPM), which builds on and generalizes the Bayesian population receptive field framework. CPM offers three main advances for cognitive computational modeling. First, it allows the fitting of cognitive models directly to neuroimaging data. Second, it allows for voxelwise or regionwise mapping of parameters of cognitive computational models onto the brain, thus making the topographic mapping methods prevalent in the sensory sciences available to the cognitive computational neuroscientist. Finally, it is efficient enough to make voxelwise mapping over large regions of interest feasible. Here, we illustrate how CPM can be used to fit reinforcement-learning algorithms to synthetic and real data.
- Open Access
- 6 November 2024
- Published by Center for Open Science in OSF Preprints
Losing money: Risk preferences for losses from long-term endowments
A central prediction within the behavioral sciences asserts that agents are risk-seeking in the domain of losses. This prediction rests on experimental evidence that did not involve the realistic experience of losses. Here we tested whether this notion still holds if participants face realistic losses. To this end, we endowed participants with over $600 for at least one month and asked them to play gambles for real, substantive, and consequential losses that spanned three orders of magnitude. A hierarchical Bayesian cognitive model comprising a latent mixture of several competing utility models yielded strong evidence for risk-seeking in the domain of losses. The participant's risk-seeking behavior was invariant to the order of magnitude of the gamble. According to this data, risk-seeking behaviour in the domain of losses generalised in our sample to consequential and substantive financial losses.
- 6 November 2024
- Published by Center for Open Science in OSF Preprints
Open AccessLosing money: Risk preferences for losses from long-term endowments
A central prediction within the behavioral sciences asserts that agents are risk-seeking in the domain of losses. This prediction rests on experimental evidence that did not involve the realistic experience of losses. Here we tested whether this notion still holds if participants face realistic losses. To this end, we endowed participants with over $600 for at least one month and asked them to play gambles for real, substantive, and consequential losses that spanned three orders of magnitude. A hierarchical Bayesian cognitive model comprising a latent mixture of several competing utility models yielded strong evidence for risk-seeking in the domain of losses. The participant's risk-seeking behavior was invariant to the order of magnitude of the gamble. According to this data, risk-seeking behaviour in the domain of losses generalised in our sample to consequential and substantive financial losses.
- Open Access
- 28 May 2024
- Published by Center for Open Science in OSF Preprints
Ergodicity transformations predict human decision-making under risk
Decision theories commonly assume that risk preferences can be expressed as utility functions, which vary from person to person but are stable over time. A recent model from ergodicity economics reveals that if people want their wealth to grow at the fastest rate they need to adjust their utility functions depending on the dynamics of their wealth. Here, we ask whether humans make such adjustments by exposing them to different wealth dynamics. We carried out an experiment in which participants made consequential risky decisions under two different conditions, additive and multiplicative wealth dynamics. We estimated risk aversion parameters separately in the two conditions for each participant, fitting isoelastic functions via hierarchical Bayesian models. In our pre-registered analyses, we found strong evidence for a change in utility function, namely an increase in the risk aversion parameter under the multiplicative condition, as predicted by ergodicity economics. Apart from evidence for a large effect of wealth dynamics, we also recover trait-like differences between participants that persist across the two conditions. Our study introduces a new experimental design and contains two independent replications between pilot data and a larger cohort. Together, these results provide evidence that human risk-taking behaviour is sensitive to the dynamical context in which decisions are made and that long-term wealth maximization is an important explanatory principle.
- 28 May 2024
- Published by Center for Open Science in OSF Preprints
Open AccessErgodicity transformations predict human decision-making under risk
Decision theories commonly assume that risk preferences can be expressed as utility functions, which vary from person to person but are stable over time. A recent model from ergodicity economics reveals that if people want their wealth to grow at the fastest rate they need to adjust their utility functions depending on the dynamics of their wealth. Here, we ask whether humans make such adjustments by exposing them to different wealth dynamics. We carried out an experiment in which participants made consequential risky decisions under two different conditions, additive and multiplicative wealth dynamics. We estimated risk aversion parameters separately in the two conditions for each participant, fitting isoelastic functions via hierarchical Bayesian models. In our pre-registered analyses, we found strong evidence for a change in utility function, namely an increase in the risk aversion parameter under the multiplicative condition, as predicted by ergodicity economics. Apart from evidence for a large effect of wealth dynamics, we also recover trait-like differences between participants that persist across the two conditions. Our study introduces a new experimental design and contains two independent replications between pilot data and a larger cohort. Together, these results provide evidence that human risk-taking behaviour is sensitive to the dynamical context in which decisions are made and that long-term wealth maximization is an important explanatory principle.
- Open Access
- 25 December 2023
- Published by Wiley in Alzheimer's & Dementia
- Vol. 19 (S23)
Characterising the covariance pattern between lifestyle factors and structural brain measures: a multivariable study of two cohorts of older adults
Background: Modifiable risk factors have been shown to promote healthy brain ageing. However, most studies to date have focused on the relationship between a single risk factor and brain health. Given that risk factors do not occur in isolation, multivariable analyses may provide a more realistic model of the effects of modifiable lifestyle factors on brain ageing.Method: Here, the relationship between 9 modifiable lifestyle factors and 7 indices of brain structure were examined using canonical correlation analyses to identify a covariance pattern between lifestyle factors associated with dementia risk and MRI‐derived measures of brain structure. Analyses were first conducted on a Danish cohort of older adults (n = 251) and then cross‐validated in an independent cohort of older adults from the United Kingdom (n = 668).Results: In a 5‐fold cross‐validation analysis of the test dataset, a canonical correlation of 0.14 was observed between the lifestyle and brain variates. In the cross‐study validation, the lifestyle‐brain correlation was r = .10. In the Danish cohort, feelings of loneliness, BMI, depressive symptoms, and years of smoking were the primary contributors to the lifestyle variate (rs ≥ 0.3). In the British cohort, a different pattern of lifestyle factors was found to relate to brain structure the most (rs ≥ 0.3: physical activity, education, alcohol consumption, blood pressure and BMI).Conclusion: In both cohorts, we found that a latent of lifestyle factors was positively associated with age‐related measures of brain structure and this relationship was validated across cohorts of older adults. However, the covariance pattern between lifestyle and brain outcomes differed between datasets. Taken together, these findings highlight how future lifestyle interventions should be tailored to suit their target populations.
- 25 December 2023
- Published by Wiley in Alzheimer's & Dementia
- Vol. 19 (S23)
Open AccessCharacterising the covariance pattern between lifestyle factors and structural brain measures: a multivariable study of two cohorts of older adults
Background: Modifiable risk factors have been shown to promote healthy brain ageing. However, most studies to date have focused on the relationship between a single risk factor and brain health. Given that risk factors do not occur in isolation, multivariable analyses may provide a more realistic model of the effects of modifiable lifestyle factors on brain ageing.Method: Here, the relationship between 9 modifiable lifestyle factors and 7 indices of brain structure were examined using canonical correlation analyses to identify a covariance pattern between lifestyle factors associated with dementia risk and MRI‐derived measures of brain structure. Analyses were first conducted on a Danish cohort of older adults (n = 251) and then cross‐validated in an independent cohort of older adults from the United Kingdom (n = 668).Results: In a 5‐fold cross‐validation analysis of the test dataset, a canonical correlation of 0.14 was observed between the lifestyle and brain variates. In the cross‐study validation, the lifestyle‐brain correlation was r = .10. In the Danish cohort, feelings of loneliness, BMI, depressive symptoms, and years of smoking were the primary contributors to the lifestyle variate (rs ≥ 0.3). In the British cohort, a different pattern of lifestyle factors was found to relate to brain structure the most (rs ≥ 0.3: physical activity, education, alcohol consumption, blood pressure and BMI).Conclusion: In both cohorts, we found that a latent of lifestyle factors was positively associated with age‐related measures of brain structure and this relationship was validated across cohorts of older adults. However, the covariance pattern between lifestyle and brain outcomes differed between datasets. Taken together, these findings highlight how future lifestyle interventions should be tailored to suit their target populations.
- Open Access
- 7 September 2023
- Published by Center for Open Science in OSF Preprints
Are risk preferences optimal?
Decision theories commonly assume that risk preferences are idiosyncratic but stable over time. A recent model from ergodicity economics reveals that optimising the growth rate of wealth requires individuals to adjust their risk preferences to wealth dynamics. Here we ask whether humans are capable of such adjustments. In a randomised control trial, participants will make risky decisions under additive and multiplicative dynamics. We will estimate risk preferences separately in the two conditions for each participant by fitting isoelastic utility functions via hierarchical Bayesian models and standard regression techniques. Growth optimal adjustments to risk preferences would confirm our main hypothesis, whereas risk preferences that are stable across conditions would disconfirm it. Pilot data from 11 participants revealed strong evidence supporting the main hypothesis. We will replicate this pilot in a pre-registered experiment with up to 150 participants.
- 7 September 2023
- Published by Center for Open Science in OSF Preprints
Open AccessAre risk preferences optimal?
Decision theories commonly assume that risk preferences are idiosyncratic but stable over time. A recent model from ergodicity economics reveals that optimising the growth rate of wealth requires individuals to adjust their risk preferences to wealth dynamics. Here we ask whether humans are capable of such adjustments. In a randomised control trial, participants will make risky decisions under additive and multiplicative dynamics. We will estimate risk preferences separately in the two conditions for each participant by fitting isoelastic utility functions via hierarchical Bayesian models and standard regression techniques. Growth optimal adjustments to risk preferences would confirm our main hypothesis, whereas risk preferences that are stable across conditions would disconfirm it. Pilot data from 11 participants revealed strong evidence supporting the main hypothesis. We will replicate this pilot in a pre-registered experiment with up to 150 participants.
- Open Access
- 4 October 2021
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 17 (10) , e1009452
Choice history effects in mice and humans improve reward harvesting efficiency
Choice history effects describe how future choices depend on the history of past choices. In experimental tasks this is typically framed as a bias because it often diminishes the experienced reward rates. However, in natural habitats, choices made in the past constrain choices that can be made in the future. For foraging animals, the probability of earning a reward in a given patch depends on the degree to which the animals have exploited the patch in the past. One problem with many experimental tasks that show choice history effects is that such tasks artificially decouple choice history from its consequences on reward availability over time. To circumvent this, we use a variable interval (VI) reward schedule that reinstates a more natural contingency between past choices and future reward availability. By examining the behavior of optimal agents in the VI task we discover that choice history effects observed in animals serve to maximize reward harvesting efficiency. We further distil the function of choice history effects by manipulating first- and second-order statistics of the environment. We find that choice history effects primarily reflect the growth rate of the reward probability of the unchosen option, whereas reward history effects primarily reflect environmental volatility. Based on observed choice history effects in animals, we develop a reinforcement learning model that explicitly incorporates choice history over multiple time scales into the decision process, and we assess its predictive adequacy in accounting for the associated behavior. We show that this new variant, known as the double trace model, has a higher performance in predicting choice data, and shows near optimal reward harvesting efficiency in simulated environments. These results suggests that choice history effects may be adaptive for natural contingencies between consumption and reward availability. This concept lends credence to a normative account of choice history effects that extends beyond its description as a bias.
- 4 October 2021
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 17 (10) , e1009452
Open AccessChoice history effects in mice and humans improve reward harvesting efficiency
Choice history effects describe how future choices depend on the history of past choices. In experimental tasks this is typically framed as a bias because it often diminishes the experienced reward rates. However, in natural habitats, choices made in the past constrain choices that can be made in the future. For foraging animals, the probability of earning a reward in a given patch depends on the degree to which the animals have exploited the patch in the past. One problem with many experimental tasks that show choice history effects is that such tasks artificially decouple choice history from its consequences on reward availability over time. To circumvent this, we use a variable interval (VI) reward schedule that reinstates a more natural contingency between past choices and future reward availability. By examining the behavior of optimal agents in the VI task we discover that choice history effects observed in animals serve to maximize reward harvesting efficiency. We further distil the function of choice history effects by manipulating first- and second-order statistics of the environment. We find that choice history effects primarily reflect the growth rate of the reward probability of the unchosen option, whereas reward history effects primarily reflect environmental volatility. Based on observed choice history effects in animals, we develop a reinforcement learning model that explicitly incorporates choice history over multiple time scales into the decision process, and we assess its predictive adequacy in accounting for the associated behavior. We show that this new variant, known as the double trace model, has a higher performance in predicting choice data, and shows near optimal reward harvesting efficiency in simulated environments. These results suggests that choice history effects may be adaptive for natural contingencies between consumption and reward availability. This concept lends credence to a normative account of choice history effects that extends beyond its description as a bias.
- Open Access
- 9 September 2021
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 17 (9) , e1009217
Ergodicity-breaking reveals time optimal decision making in humans
Ergodicity describes an equivalence between the expectation value and the time average of observables. Applied to human behaviour, ergodic theories of decision-making reveal how individuals should tolerate risk in different environments. To optimize wealth over time, agents should adapt their utility function according to the dynamical setting they face. Linear utility is optimal for additive dynamics, whereas logarithmic utility is optimal for multiplicative dynamics. Whether humans approximate time optimal behavior across different dynamics is unknown. Here we compare the effects of additive versus multiplicative gamble dynamics on risky choice. We show that utility functions are modulated by gamble dynamics in ways not explained by prevailing decision theories. Instead, as predicted by time optimality, risk aversion increases under multiplicative dynamics, distributing close to the values that maximize the time average growth of in-game wealth. We suggest that our findings motivate a need for explicitly grounding theories of decision-making on ergodic considerations.
- 9 September 2021
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 17 (9) , e1009217
Open AccessErgodicity-breaking reveals time optimal decision making in humans
Ergodicity describes an equivalence between the expectation value and the time average of observables. Applied to human behaviour, ergodic theories of decision-making reveal how individuals should tolerate risk in different environments. To optimize wealth over time, agents should adapt their utility function according to the dynamical setting they face. Linear utility is optimal for additive dynamics, whereas logarithmic utility is optimal for multiplicative dynamics. Whether humans approximate time optimal behavior across different dynamics is unknown. Here we compare the effects of additive versus multiplicative gamble dynamics on risky choice. We show that utility functions are modulated by gamble dynamics in ways not explained by prevailing decision theories. Instead, as predicted by time optimality, risk aversion increases under multiplicative dynamics, distributing close to the values that maximize the time average growth of in-game wealth. We suggest that our findings motivate a need for explicitly grounding theories of decision-making on ergodic considerations.