How the brain conducts complex tasks
The EU-funded EMCOREP (Emergence of complex internal representations in humans) project investigated the behavioural and computational foundations of human sensory learning in adults and children. They used an integrated interdisciplinary approach based on behavioural studies, computational analyses and theoretical work. Researchers began by investigating what type of computational framework best captures the functioning of the brain. A probabilistic framework was proposed, which assumed the brain weighs up incoming information according to its ‘plausibility’. Although this strategy leads to optimally efficient behaviour, especially in complex situations, it also raises a number of questions with regard to implementation. Behavioural studies and modelling were carried out to determine whether this probabilistic framework is likely to be implemented in the brain. Experiments were conducted to demonstrate different manifestations of problematic behaviour. This revealed that people learn unknown information through probabilistic learning rather than simple associative learning and combine their knowledge according to the probabilistic method. It is possible that only the high cognitive aspects of brain functioning follow the probabilistic route and low level sensory perception is hardwired. EMCOREP conducted further investigations to determine how fundamental this type of behaviour was in the cortex. Results revealed that the brain implements probabilistic computation from the very first step of sensory information processing. EMCOREP assessed if the probabilistic framework adequately described internal representations and their effect in the brain. They studied how decision processes are influenced by previous decisions. They found that strong biases do emerge but that they are influenced through a complex process, where significance is given to detected changes in the environmental status quo. Scientists also conducted experiments that suggested sampling-based computational schemes may be in place in the cortex. They examined whether this sampling-based representation can be used in hierarchical models to test it against more complex tasks such as visual processing. A final study tested the hypothesis that information processing in children and adults is very different. Together all the investigations conducted by EMCOREP provide a major step forward in understanding the role of the probabilistic framework in complex cortical functions.
Keywords
EMCOREP, probabilistic framework, internal representation, cortex, sampling-based representation
