Developing Higher-Level Cognitive Theories by Reduction

Developing Higher-Level Cognitive Theories by Reduction Tibor Bosse (tbosse@cs.vu.nl), Catholijn M. Jonker 1 (C.Jonker@nici.ru.nl), Jan Treur 2 (treur@cs.vu.nl) Vrije Universiteit Amsterdam, Department of Artificial Intelligence De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands can be used to strengthen the usefulness and further development of the higher-level theory. This position acknowledges the achievements of neuroscience and shows how to exploit these results where possible, in favour of higher-level concepts instead of against them. This paper shows how this perspective provides a scientific and philosophical foundation for higher-level notions. A practical method, based on formal techniques and tools, is proposed to support the perspective. The method will result, among others, in relational specifications of the functional role and of the representational content of these concepts; cf. Kim (1996, 1998, 2005). The method allows to check whether these specifications are logically coherent mutually, and with global behavioral properties, and, in as far as available from neuroscience, consistent with neurological theories. The resulting relational specifications can be said to specify meaning and underlying mechanisms for the higher-level concepts, and as such provide grounding of them both within the higher-level theory and within physical reality. Furthermore, the paper addresses different notions of explanation in terms of the higher-level theory and the lower-level theory. From a practical point of view, in a specific case study the predictive value of higher-level explanations is validated. Abstract Within philosophical literature, higher-level cognitive concepts such as free will, authorship of actions, and conscious control are often questioned. Neurological and biochemical mechanisms underlying human behavior provide alternative explanations of action. Reduction of cognitive states to neurophysiologic states shows that higher-level cognitive concepts in principle can be eliminated, replacing them by neurophysiologic concepts. In contrast, in this paper it is shown how reduction relations can be used in a constructive manner to strengthen the scientific foundation of higher-level cognitive concepts and further develop higher- level theories in which these concepts play a role. Introduction Reduction is an important theme within literature in the area of Cognitive Science, Philosophy of Science, and Philosophy of Mind; e.g., Kim (1996, 1998, 2005), Bickle (1998, 2003). One of the main perspectives advocated is physicalism: the idea that all processes (among which mental and biological processes), in one way or the other have a physical basis. Within Biology the strong development of biochemistry supports this perspective. For Cognitive Science, the strong development of neuroscience and its underlying biochemistry plays a similar role. These developments sometimes lead to a position that considers higher-level concepts, such as intention, free will, and consciousness, just illusions, and not usable in a scientific context; e.g., Wegner (2002). A main question addressed in this paper is how to defend such human-like, higher-level concepts. One strategy is to criticise the existence of reduction relations; if these do not exist, it is impossible to relate higher-level concepts to lower-level ones, and in this way to eliminate the higher- level concepts; e.g., Bennett and Hacker (2003, pp. 355- 366). This defensive strategy is not applicable in those cases where reduction relations have been or are being shown to exist. Another strategy is to put doubt on the quality of the higher-level theory, for example, by claiming that there do not exist cognitive laws that could be related to neurological laws; e.g., Bennett and Hacker (2003, p. 362). This strategy is not applicable in cases that the higher-level theory is assumed to be still under development. In this paper a different strategy is explored. This strategy takes the existence of reduction relations in the present or in the future as an assumption, and shows in some detail how they On the Use of Reduction In the philosophical literature on reduction of scientific theories, in many cases the advantages for scientific practice of having a reduction relation between two theories are not addressed explicitly. In such cases, sometimes it is implicitly assumed that these advantages are based on the idea to use the lower-level theory instead of the higher-level the theory. For example, Kim (1996, pp. 214-216) emphasizes three advantages of reduction based on (biconditional) bridge principles: ontological simplification (following Smart, 1959), having to deal with fewer assumptions about the world, and providing explanations of the laws of the higher-level theory in terms of the lower- level theory. For example, if F and G are higher-level entities and F* and G* lower-level ones with biconditional bridge laws F ↔ F* and G ↔ G* (cf. Nagel, 1961), then F* and G* can be used instead of F and G. 3 An eliminative perspective on the use of reduction provokes resistance from those who defend an autonomous status for higher-level Currently at: Radboud Universiteit Nijmegen, Nijmegen Institute for Cognition and Information, The Netherlands Part of this work was performed at a position at Utrecht University, Department of Philosophy, The Netherlands ‘Such identities, one could argue, are essential to the ontological simplification that we seek in theory reduction, for they enable us to dispense with facts involving F and G as something in addition to facts involving F* and G*.’ (…) ‘It shows that fewer basic laws, and fewer basic expressions, fully suffice for the description and explanation of the phenomena of a given domain.’ Kim (1996, p. 215)