Funded Grants

From single decisions to mental programs

My research program stems from a simple question: How can we solve so easily tasks that go well beyond the limits of the most powerful computers, such as understanding jokes or making (good) theories from extremely sparse data, but struggle with tasks that demand a puny amount of processing, such as multiplying 1279*563, which the most primitive calculating machines can easily compute.

My hypothesis is that this apparent paradox resides in the fact that the human brain behaves like a hybrid computing device. We (adults, young infants and most likely other non-human animals) make rapid inferences computing in parallel entire probability distributions. This form of computation driven by priors, carved by learning and operating with analog quantities is the opposite of Alan Turing's sequential symbolic devices. This explains why Turing’s ideas on computation have met strong resistance in the neuroscience community. However, an emergent aspect of cognition, the conscious rational thought that was at the root of Turing’s insight, acts as a slow serial machine1. A first challenge of my research program is to understand how Turing-like operations can be implemented in the brain using architectures quite different from those used in classical computers.

Mental calculation and understanding a joke also differ substantially in their subjective experience. The former is perceived with a strong sense of mental effort. The latter feels like there is really nothing that needs to be done; it just happens. A second challenge of my research program is to investigate this dissociation between the computational complexity of mental operations and their subjectively perceived difficulty. This brings together the flourishing field of metacognitive research (introspection about mental content, feeling of knowing, confidence) with analytic tools traditionally applied to dissect elementary cognitive processes from an objective, third-person perspective.

Finally, the qualitative difference between the transparency of the procedures used in arithmetic calculation and the opaqueness of the processes by which we understand humor or develop intuitions is also expected to influence the capacity to teach them. I plan to investigate the transmission of knowledge from the point of view of the teacher, focusing in the development of spontaneous teaching behavior in children. It seems reasonable to assume that one can only teach contents and procedures which are accessible to explicit knowledge and hence conceive teaching as a vehicle to inquire about introspection. I will contrast this natural hypothesis with the more provocative view that teaching constitutes a spontaneous behaviour which acts as a precursor and catalyst of the construction of explicit knowledge, bringing to the laboratory Seneca's famous idea, docendo discimus; when we teach we learn.