Computational Creativity: A Philosophical Approach, and an Approach to Philosophy Stephen McGregor, Geraint Wiggins and Matthew Purver School of Electronic Engineering and Computer Science Queen Mary University of London s.e.mcgregor@qmul.ac.uk, geraint.wiggins@qmul.ac.uk, m.purver@qmul.ac.uk Abstract This paper seeks to situate computational creativity in relation to philosophy and in particular philosophy of mind. The goal is to investigate issues relevant to both how computational creativity can be used to explore philosophical questions and how philosophical posi- tions, whether they are accepted as accurate or not, can be used as a tool for evaluating computational creativity. First, the possibility of symbol manipulating machines acting as creative agents will be examined in terms of its ramifications for historic and contemporary theories of mind. Next a philosophically motivated mechanism for evaluating creative systems will be proposed, based on the idea that an intimation of dualism, with its in- herent mental representations, is a thing that typical ob- servers seek when evaluating creativity. Two compu- tational frameworks that might adequately satisfy this evaluative mechanism will then be described, though the implementation of such systems in a creative con- text is left for future work. Finally, the kind of audi- ence required for the type of evaluation proposed will be briefly discussed. Introduction In quotidian interactions, either on a personal or social level, computers are such familiar devices that their operations are taken for granted as having the same kind of relatively uni- versal grounding that humans engaging in interpersonal ex- changes of information employ. When computers become either the platform for or the object of philosophical en- quiries, though, it becomes necessary to talk about them as information processing systems or as symbol manipulating machines (per Newell and Simon, 1990): in this sense, the operations which computers perform must be seen as tran- spiring in an abstract space, defined by a system of infor- mation grounded somehow relative to an observer. This quality of computation immediately introduces a problem- atic element of subjectivity to the assessment of a purely informational system’s ability to generate meaning, and an ambiguity arises over whether such a system can really au- tonomously produce output which has been invested with semantic content. It is due to precisely this key feature of computational sys- tems, their dependence on an observer for operational co- herence, that computers have become an element in various philosophical discussions, often in the form of reductiones ad absurdem, exercises aimed at problematising both reduc- tionist and internalist accounts of mental phenomena. Put- nam (1988) in particular has argued for the computational significance of the internal states of a rock, while Searle (1990) constructed his famous Chinese room argument to demonstrate the absence of intentionality in machines which merely manipulate symbols, a stance subsequently used as a platform for questioning the very basis of cognitive experi- ence. In these examples, computers come out as the foils for arguments about the intractable difficulty of defining or even talking about human consciousness. Rather than treating computers as the theoretical objects of thought experiments, this paper will argue, as Sloman (1978) did several years ago, that computers should be considered essential tools for doing good philosophy, and that in particular the question of whether computers can be autonomously creative is philo- sophically valid. This paper’s first objective is to place the field of com- putational creativity within the context of the philosophy of mind, and in particular to consider how the field might be used as a vehicle for empirically exploring the problem of dualism which has been characteristically at the centre of questions regarding the mind and consciousness in modern Western philosophy. To this end, a strong counterargument to the traditional mode of dualism, which argues that the mind and physical matter occupy two mutually irreducible spaces, can be found in considering ways in which symbol manipulating machines might be able to autonomously pro- duce informational artefacts that are new and valuable and that furthermore bear some sort of meaning relevant to the way in which the creative system itself operates. If a compu- tational system can produce new, valuable artefacts in a way that is deemed suitably creative, and yet these systems are themselves reducible to manipulations of symbols grounded in the workings of a physical machine, there seems to be no case to make for the idea that the act of generating new meaning in the world transpires in some intangible mental domain. The second objective of this paper is to propose a new mechanism for evaluating creative systems, motivated by in- sights into the way that humans view themselves. Taking the intransigence of the mind/body problem as a starting point, it will be suggested that it is precisely the kind of represen-