Acessando representações mentais para predizer estímulos: como crenças modulam sinais cerebrais

Abstract

Using the theoretical framework called Predictive Processing (Clark, 2013), which assumes that brains are essentially machines that predict their next inputs and do so by minimizing prediction error (difference between predicted and received input) similar to a Bayesian inference, it was sought, in this dissertation, to find patterns of electrophysiological activity that signal the recruitment of epistemic beliefs. It has been conjectured that as sentences are read, representations of states of affairs and events are mentally modeled in a gradual and predictive fashion, and that the upcoming stimuli are anticipated on the basis of what appears to be "most likely" in epistemic, semantic, syntactic, lexical and perceptual terms. Due to the fact that the electrophysiological responses that are modulated by violations of expectation of the semantic, syntactic, lexical and perceptual levels are known, as well as the precise time interval in which they are manifested, it was possible to study how the expectation generated by beliefs modulates the cerebral signals. The experiments carried out in the Laboratory of Experimental Philosophy and Studies of Cognition, located at UNISINOS, presented sentences on a monitor (word by word) that alluded to facts known only by a group of participants (philosophers), structured in a way that only a single word, appearing at the end of each sentence, would be able to make them true (e.g. "Theaetetus is a dialogue written by Plato"). The proposed model assumes that, before the final word appears, the meaning of the terms is accessed (in the exemplified case, a specific text and a relation), a state of affairs represented (i.e. there is an x such that x wrote Theaetetus), and a search initiated to find the best candidate for x, taking the subject's beliefs as the target of tracking. The relevant belief is then recruited and the representation of the identified x (Plato) accessed. Information on the best semantic prediction (the philosopher Plato) is used to select the most likely lexical item that has Plato as meaning. Participants for which philosophical beliefs were ascribed presented Event-Related Potentials correlated to lexical-semantic processing significantly different from participants in the non-philosophers group. Considering that the only relevant difference between the two groups was the conjectured possession or absence of certain beliefs, the results were interpreted as signaling the recruitment of beliefs in predictive processes underlying textual comprehension. The results contradict eliminativist positions that consider that the mentalist vocabulary about beliefs, intentions, and desires is meaningless, since the difference found suggests that there is indeed something in brains that is denoted by these expressions (albeit vaguely and coarse) and that is modulating the signals. Although we do not readily understand exactly how belief instantiation by brains is effective, we can study how they are recruited and integrate various cognitive processes (i.e. their causal roles).