Page 56 - Deep Learning
P. 56

The Nature of the Enterprise             39

            conception  dominates  contemporary  cognitive  psychology.  Information  pro-
            cessing concepts are used up and down the pages of cognitive psychology text-
            books. Few of the experiments reported in psychological research articles make
            sense unless we assume that there is a cognitive architecture that remains stable
            across tasks and content domains; after all, the tasks performed by the subjects
            in those experiments are seldom of any intrinsic interest.
               Nevertheless, the Turing-Newell conception is unlikely to be correct.  It
                                                                         37
            is not plausible that representations and processes are as neatly separated in
            the brain as they are in a computer. It seems more plausible that the medium
            of mind will ultimately turn out to be representations that are also processes,
            or processes that also represent. We do not know how to model such strange
            entities. Also, the assumption of a fixed, presumably innate control structure
            might have to yield to a more fluid notion of control that emerges through self-
            organization at the neural level. But such developments are in the future, so the
            following chapters adopt a Newell-inspired view of the cognitive architecture as
            a conceptual platform from which to pursue principles of cognitive change.


                                  EXPLAINING CHANGE
            If the mind is a system for processing knowledge representations in the service
            of action and discourse, how does it change over time? What form should an
            explanation for cognitive change take? What are the criteria of a satisfactory
            explanation and what issues arise in the construction of such explanations? It
            is informative to consider successful explanations for other types of change.
            Cognitive psychologists need not imitate other sciences or assume that their
            own theories must, in the end, look like those of any other science, but neither
            is it wise to assume that other sciences have no lessons to teach. Natural scien-
            tists, social scientists and humanists have grappled with the concept of change
            and their successes provide calories for psychological thought.


                                Componential Explanations
            To extract the general features of scientific explanations of change, consider con-
            tagion and electrolysis, two seemingly different phenomena. How do contagious
            diseases like yellow fever spread such that we suffer epidemics?  It required con-
                                                              38
            siderable research to identify all parts of this complicated process. Although a
            specialist on yellow fever could add innumerable details, a mere outline suffices
            here: The disease is caused by a germ that multiplies in a person’s body, causing
            the symptoms. The sick person is bitten by a mosquito, which sucks up blood
   51   52   53   54   55   56   57   58   59   60   61