The Turing Machine as a Cognitive Model of Human Computation Simone Pinna Simone Pinna, Università di Cagliari, Cagliari, Italy (e-mail: simonepinna@hotmail.it). Abstract — Classical computationalism considers the Turing Machine to be a psychologically implausible model of human computation. In this paper, I will first elaborate on Andrew Wells' thesis that the claim of psychological implausibility derives from a wrong interpretation of the TM as originally conceived by Turing. Then, I will show how Turing's original interpretation of the TM could be useful to construct cognitive models of simple phenomena of human computation, such as counting using our fingers or performing arithmetical operations using paper and pencil. Index Terms — Turing Machine, computationalism, Extended Mind Model, human computation I. INTRODUCTION Mainstream computationalism maintains that cognitive activities are the result of computations performed wholly within the cognitive system. This definition is deliberately vague, for different computational models could be distinguished by the ways in which computations are performed ( e.g. see the difference between symbolic and connectionist models). In what follows, I will exclusively focus on classical computationalism. With this expression, I mean the idea that computations are performed by a cognitive system through internal symbolic transformations based on purely syntactic rules. This idea lies behind many important cognitive theories, like Newell's Physical Symbol System (Newell,1980) and Fodor's Computational Theory of Mind (Fodor, 1975). Classical computationalism considers the Turing Machine (TM) as a paradigm for the definition of the abstract concept of effective computation, but it also takes the way in which computations are performed by a TM to be psychologically implausible. In this paper, I will first elaborate on Andrew Wells' thesis that the claim of psychological implausibility derives from a wrong interpretation of the TM as described by Alan Turing (1936). Then, I will show (i) that TM's computations, as originally intended by Turing, are in fact compatible with the Extended Mind Model (Clark and Chalmers, 1998) and (ii) how Turing's original interpretation of the TM could be useful to construct cognitive models of simple phenomena of human computation (Giunti, 2009) such as counting using our fingers or performing arithmetical operations using paper and pencil. II. PSYCOLOGICAL INTERPRETATIONS OF TURING'S WORK In his article On computable numbers with an application to the Entscheidungsproblem, Turing (1936) describes his computing machine starting from the analysis of the mechanisms standing beneath a real cognitive system, namely the one consisting of a man performing a calculation with paper and pencil. This article has ever since been recognized as a cornerstone of the computational approach to cognition. Nevertheless, this high consideration of Turing's work is primarily focused on its purely mathematical content (the formalization of the notion of effective procedure). As for its specific cognitive content, it is instead widely held that the way in which computations are performed by a TM makes it a psychologically implausible model of a real cognitive system. Andrew Wells (1998, 2005) rejects this claim of implausibility showing that it is based on a misinterpretation of the TM's basic architecture. A TM is essentially composed of: 1. a finite automaton (the internal part of the architecture) consisting of a simple input-output device that implements a specific set of instructions (machine table), and a small working memory that holds only one discrete element at each step (internal state); 2. a memory unit (the external part of the architecture) consisting of a tape divided into squares, potentially extendable in both directions ad libitum; 3. a read/write/move head, that scans the content of a cell at a time, and that links the internal and external parts of the machine. Given the initial configuration of a TM, every successive configuration of the machine is determined by an appropriate quintuple of the form: (q x , r: w, M, q y ). (1)