Dynamic Visual Inferences and their Neural implementation

Despite recent spectacular advances in the field of Artificial Intelligence, Cognitive Science and Neuroscience, the flexible functioning of human and animal brains still poses one of the greatest challenges to our scientific understanding. How can we learn so much so quickly and how can we use our knowledge to solve so many completely different problems from tying shoelaces to solving a new word puzzle to recognizing a photo seen more than two decades ago? In our project, we approach this conundrum with the hypothesis that this flexibility and ability to generalize is anchored in the way we encode any information in the brain. Specifically, we assume that humans and some animals encode and use sensory and memory information probabilistically that is they handle information together with their uncertainty as to how much they trust the given information is correct. We and others have argued in the past that such belief-weighted handling on incoming and stored information provides the necessary flexibility that has been observed during brain functioning. To confirm this hypothesis, we devised a perceptual decision making task, in which the identity of the visual input in each trial could be categorized in one of two possible ways depending on what the observers internal interpretation of the contextual situation is. Importantly, we set up the task so that we could identify purely from the observers answers which interpretation they go with. Using this task in our project, we plan to collect data from humans and behaving non-human primates to obtain behavioural evidence that these species solve these kinds of tasks similarly to each other during decision making and we will identify the interpretation they use for their decisions. Meanwhile, we will also collect eye movement data since we suspect that cognitive decisions and eye movement patterns reflect different aspects of the internal interpretation the observers use, thus measuring behavioural decision and eye movements would provide concurrent and independent confirmation of our hypothesis. In the second part of our project, we will search for neural evidence supporting our hypothesis of probabilistic coding in the brain. We will use a technique called Voltage Sensitive Dye Imaging (VSDI) that allows to see the activity of neural population at a large scale in the cortex of the behaving animal. We will investigate whether these observed patterns in the cortex will indicate which internal representation was used by the animal. If our hypothesis is correct, we expect to see an immediate neural correlate when a shift of the internal mindset of the observer occurs i.e., when the interpretation of the animal about the events occurring in the environment changes. If successful, not only will this project provide a novel way to link behaviour to neural functioning of the brain, but it would also be the first direct neural evidence of probabilistic thinking in the brain.