Ereira, Samuel Philip Aaron; (2019) A neurocomputational account of self-other distinction: from cell to society. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Ereira thesis.pdf - Accepted Version Download (34MB) | Preview

Abstract

Human social systems are unique in the animal kingdom. Social norms, constructed at a higher level of organisation, influence individuals across vast spatiotemporal scales. Characterising the neurocomputational processes that enable the emergence of these social systems could inform holistic models of human cognition and mental illness. Social neuroscience has shown that the processing of ‘social’ information demands many of the same computations as those involved in reasoning about inanimate objects in ‘non-social’ contexts. However, for people to reason about each other’s mental states, the brain must be able to distinguish between one mind and another. This ability, to attribute a mental state to a specific agent, has long been studied by philosophers under the guise of ‘meta-representation’. Empathy research has taken strides in describing the neural correlates of representing another person’s affective or bodily state, as distinct from one’s own. However, Self-Other distinction in beliefs, and hence meta-representation, has not figured in formal models of cognitive neuroscience. Here, I introduce a novel behavioural paradigm, which acts as a computational assay for Self-Other distinction in a cognitive domain. The experiments in this thesis combine computational modelling with magnetoencephalography and functional magnetic resonance imaging to explore how basic units of computation, predictions and prediction errors, are selectively attributed to Self and Other, when subjects have to simulate another agent’s learning process. I find that these low-level learning signals encode information about agent identity. Furthermore, the fidelity of this encoding is susceptible to experience-dependent plasticity, and predicts the presence of subclinical psychopathological traits. The results suggest that the neural signals generating an internal model of the world contain information, not only about ‘what’ is out there, but also about ‘who’ the model belongs to. That this agent-specificity is learnable highlights potential computational failure modes in mental illnesses with an altered sense of Self.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item