DS0407 -

Being an agent in an uncertain world: a neuro-computational approach – BE-AGENT

Submission summary

Healthy adults generally have a sense of control over their own actions, and over the effects of those actions – a feeling that is classically referred to as “sense of agency”. So far, research on the sense of agency has predominantly been driven by developments in the field of sensorimotor control. Although influential, the sensorimotor view is however limited in its explanatory scope as it cannot account for choice behaviours that do not depend on the sensorimotor system. In this research project, I propose to address this shortfall by exploring sense of agency and its neurocognitive correlates at an elementary level of neurocognitive functioning.

Fundamentally, a reliable sense of agency – the sense that I am the cause of a change in the world – necessarily draws on a reliable model of causal inference, i.e., a model of how actions causally relate to their consequences. Despite considerable advances in the modelling of causal inference-making in human observers, it is still unclear i) which model best describes people’s ability to infer their power as causal agents, and hence ii) what the neural correlates of this causal model are, and iii) whether and how selective alterations of this model can account for symptomatic pathologies of agentive experience.

To address these three unresolved questions, I propose to develop an original paradigm inspired from studies of decision-making under uncertainty. This paradigm models a dynamic environment where subjects are to continuously monitor their causal influence over the task environment by tracking multiple statistics (statistical relationships, value, variability of action outcomes) at a given time, and by flexibly adjusting to unpredictable changes in action-outcome contingencies. Together with designing adequately controlled behavioural experiments, this paradigm will involve defining, testing, and comparing several theoretical models of choice, and hence will allow for addressing three complementary questions: 1. What causal model best accounts for people’s behaviour? 2. Is the best-fitting model unique, or are there multiple models available to individuals? 3. And if so, what are the factors that arbitrate between models?

Crucially, distinct models should map selectively onto dissociable brain structures, allowing us to finely map the algorithmic architecture of causal inference in the human brain. In the second axis of this proposal, we will turn to functional neuroimaging (fMRI) in order to identify which neural variables correlate with variables of best-fitting models. We will draw upon fMRI to provide an evidence-based answer to the question of whether multiple models are simultaneously available to individuals, and if so, we will further seek to characterize the nature of their interaction (e.g. cooperation or competition).

Finally, we aim to exploit the neurocognitive correlates identified to pinpoint the selective alterations of agency in patients with schizophrenia. Disorders of agentive experience can take various (and often extreme) forms in the disease, from delusions of control to delusions of omnipotence and auditory hallucinations. In the third axis of this proposal, we will ask whether these heterogeneous disorders can be convincingly described and explained within our original framework, as distortions and/or suboptimal use of causal models normally used by healthy individuals.

The scientific program of this proposal follows these three complementary questions, with three principal research axes combining model-based analyses of human choice behavior with non-invasive recording of human brain activity in fMRI, in both healthy and psychiatric participants.

Project coordination

Valerian CHAMBON (Institut Jean Nicod)

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.


IJN Institut Jean Nicod

Help of the ANR 262,995 euros
Beginning and duration of the scientific project: December 2016 - 36 Months

Useful links

Explorez notre base de projets financés



ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter