Thesis defense Maria Otworowska (Donders series 335)
4 October 2018
Promotor: prof. dr. H. Bekkering, co-promotors: dr. I. van Rooij, dr. J. Kwisthout
Computational Demons of an Adaptive Brain
Cognitive science has been on a quest to explain human cognitive processes for more than six decades. At the heart of cognitive science research lies the assumption that a cognitive system is computational in nature. It means that to enter into consideration as an explanation of cognitive processes, proposed models, theories and frameworks need to take into account human computational limitations. Neglecting or ignoring computational limitations leads to postulating computations that are effectively too difficult for humans to compute. This gives rise to a paradox. According to a theory postulating such computations, humans should not be able to do what they effortlessly do every day, like reasoning, communicating, making decisions or inferences. The only way to attribute such computations to humans would be to assume that humans are omniscient and omnipotent demons with unlimited time at their disposal. In this thesis, we do not attribute any supernatural powers to humans. Instead we argue that such postulated computations --- computational demons --- are inherent to theories and manifested in theories’ different assumptions and premises. To illustrate this, we used a variety of methods to challenge, analyze, formalize and test the assumptions of two influential theories: Adaptive Toolbox (AT) and Predictive Processing (PP). Both theories give a conceptually broad account of human functioning in an uncertain and complex world. PP postulates a universal principle by which all cognitive processes work, while AT argues that one-size-cannot-fit-all, and proposes that instead of one principle, there is actually a whole toolbox of strategies humans can employ and work by. Not only do these theories seem to make claims from the opposite spectrum, they also recognize each other as opposing. Especially crucial in this differentiation is that AT recognizes PP as a demonic theory while claiming itself allegedly demon-free. We have analyzed claims made by both theories, and we showed that the theories share many assumptions and commonalities, not only conceptually, but also in that they both suffer from computational demons. Computational demons are a serious impediment to scientific progress, and cannot be ignored. Our analyses and proposed methods of detecting and fighting the demons, even though applied to these two specific theories only, are in fact universal for cognitive science research.