A Paradigm Shift in Cognitive Sciences?

Since the 1950s, the dominant paradigm in the cognitive sciences has been cognitivism, which arose as an alternative to behaviorism and regards cognitive processes predominantly as various kinds of “computations” similar to those performed by a computer; there is no fundamental difference from a universal Turing machine (TM). Despite significant progress made in the last quarter of the 20th century within the framework of this paradigm, many remain dissatisfied as it cannot adequately explain some features of cognitive processes. Connectionism, which emerged later, although recognizing the role of computational processes, regards them as based not on a TM, but on a neural network, as this provides a better model of brain functionality than TM-type computations. Neural networks, unlike classical computers, show resilience and flexibility in the face of problems arising in the real world, such as increasing input noise or blockade of part of the network. They are also well suited for tasks requiring parallel resolution of many conflicting constraints. Despite this, the analogy between the functioning of the human brain and artificial neural networks is still limited, because of radical differences in the design and associated capabilities of such systems. In parallel with the paradigms of cognitivism and connectionism, further ideas developed, whereby cognition is the result of strictly biological processes involved in the interaction of an organism with the external environment. These concepts, which have become increasingly popular in recent years, have taken shape in various strands of so-called enactivism. This review compares the theoretical postulates of cognitivism, connectionism, and enactivism, along with paradigms of predictive coding and the principle of free energy. © 2023, Springer Nature Switzerland AG.