Philosophical Foundations of AI

Artificial Intelligence was born in 1956 as the off-spring of the newly-created cognitivist paradigm of cognition. As such, it inherited a strong philosophical legacy of functionalism, dualism, and positivism. This legacy found its strongest statement some 20 years later in the physical symbol systems hypothesis, a conjecture that deeply influenced the evolution of AI in subsequent years. Recent history has seen a swing away from the functionalism of classical AI toward an alternative position that re-asserts the primacy of embodiment, development, interaction, and, more recently, emotion in cognitive systems, focussing now more than ever on enactive models of cognition. Arguably, this swing represents a true paradigm shift in our thinking. However, the philosophical foundations of these approaches — phenomenology — entail some far-reaching ontological and epistemological commitments regarding the nature of a cognitive system, its reality, and the role of its interaction with its environment. The goal of this paper is to draw out the full philosophical implications of the phenomenological position that underpins the current paradigm shift towards enactive cognition. 1 Philosophical Preliminaries Realism is a doctrine which holds that the objects of our perceptions are what is real and that reality is what is directly perceived; it is through our perceptions that we apprehend the actual real external world. The tradition of modern realism has an long pedigree, beginning with Ockham and continuing through Gallileo, Hobbes, Locke, Hume, Moore, and Russell. Gallileo, along with, e.g., Copernicus, Descartes, and Kepler, heralded the beginning of the scientific age which placed all empirical measurement and quantification along with rigourous mathematical (or logical) reasoning as the cornerstones for the construction of knowledge. This empiricist ethos was strengthed by John Locke, a quintessential realist, who viewed perception as a causal process whereby physical stimuli act on the sensory apparatus to produce ideas (concepts or representations, in the modern terminology). Much of today’s common understanding of reality is a legacy of this Lockean frame of mind. In realistic positions, there is the underpinning assumption that reality exists absolutely and, whether rationally by reason or empirically by sense, we apprehend it and thus come to understand its form and structure. Idealism, on the other hand, is a doctrine which posits that reality is ultimately dependent on the mind and has no existence outside of it. If Locke was M. Lungarella et al. (Eds.): 50 Years of AI, Festschrift, LNAI 4850, pp. 53–62, 2007. c © Springer-Verlag Berlin Heidelberg 2007 54 D. Vernon and D. Furlong the quintessential realist, then Berkeley was the quintessential idealist. Berkeley developed the philosophy that nothing exists save that which is perceived by a mind. This is neatly summarized by his famous aphorism ‘esse est percipi’ — to be is to be perceived. Berkeley’s position is that our idea about the world are based on our perceptions of it. In this sense, Berekeley is also taking an empirical position — that our knowledge of the world is gained exclusively from our senses. On the other hand, Berkeley denied the existence of matter: what exists is that which is perceived, and it exists because it is perceived. Reality pervades all perception but corporeal matter has no place in this scheme. This denial of the reality of matter distinguishes Berkeley’s empirical idealist notions of perception from the realist, empirical, notion that perception is an abstraction or apprehension of the (material) world via a causal process of sensing. Kant (1724-1804) was also an idealist, but his views differed significantly from those of Berkeley. Kant differentiated between noumena, the domain of ‘things in themselves’ and phenomena, or the ‘appearances’ of things as they are presented to us by our senses. Kant argued that noumena are not accessible to us, and cannot be known directly, whereas the phenomena — the contact we have with these things via our senses and perceptions — are the basis for knowledge. Kant refers to noumena as ‘trancendental objects’ and his philosophy is sometimes referred to as ‘trancendental idealism’. Thus, Kant admits the ‘reality’ of a domain of objects, the unknowable noumenological domain. On the other hand, he maintains that the objects of our experience are the only knowable objects and it is the mind that shapes and forms these sense data and, hence, for us, these objects are the only objects that really exist and they exist because of us and our minds. Reality, then, exists as an unknowable, nonsensible, noumenal domain which gives rise to the phenomenal domain of our senses.1 The idealist tradition did not stop with Kant and has been added to by, e.g., Schopenhauer, Nietzsche, and Hegel. There are many variations on these two themes of idealism and realism, perhaps the most well-known of which is dualism which holds that reality comprises two distinct ‘substances’: one physical and one mental. Dualism was first propounded as a philosophical system by Descartes who argued for the existence of two domains of reality: one corporeal and one non-corporeal. Both mutuallyexclusive domains exist concurrently. It is this mutual exclusivity which has caused dualism most of its problems for, if they are truly mutually exclusive, it is not clear how they can interact. This difficulty has been transposed into modern philosophical debate as the ‘mind-body’ problem: the paradox that if the body and mind are mutually exclusive entities, then how do they ‘communicate’? In the above, we have attempted the impossible: to summarize five hundred years of philosophical thought in a few paragraphs. Nonetheless, from this cursory look at the history of western philosophy, we can see that the philosophical 1 Although Kant is best known as an idealist, his particular brand of philosophical idealism is sometimes referred to as constructive realism due to the central role played by the observer in shaping phenomenal reality. Philosophical Foundations of AI 55 positions on reality have been dominated by realism (including dualism). Additionally, the philosophies that have been most closely aligned with the scientific method have also been those of realism. In a sense, this isn’t surprising since realism is the more immediately common-sense view: things exist — we perceive them. This world-view has been copper-fastened in the last century by the logical positivists, e.g. Schlick and Carnap, who held that reality is exactly that which yields to empirical investigation and anything that is not verifiable by empirical investigation is meaningless. There were, of course, other developments in philosophical thinking, which begin with Kant’s distinction between noumena and phenomena, and which evolved into a type of reconciliation of the idealist and the realist positions. The one that interests us here was developed by Husserl, who held that reality is personally and fundamentally phenomenological but is set against an objective spatio-temporal world. However, it was best espoused by Heidegger who denied the dichotomy between the world and ‘us’ and saw existence or ‘being in the world’ as our activity in a constitutive domain. Reality does not exist ‘outside us’; we are beings in a world, not disjoint from it. From a phenomenological perspective, what we perceive depends on what it is we are. The position taken by phenomenology is subtly, but significantly, different to that taken by either realism or idealism. The position is as follows. We play a role in defining reality, but only insofar as it affects us as individuals (the idealist aspect), that is, insofar as it affects our experience of reality; the reality that we perceive does exist (the realist aspect) but our perception and conception of it is conditioned by our experience. Thus, reality is for us a personal experience, though it derives from a common source and this reality is our experience and is contingent upon the current ontological status of us as entities in that universe. As perceivers, our perceptions of the world are a function of what we are: reality is conditioned by experience and experience is conditioned by the nature of the system and its history of interaction with reality. The dependence of reality on the ontogenetic state of an individual is the essential characteristic of phenomenology and is often referred to as radical constructivism: we construct our reality as a consequence of our perceptions and experiences. Unfortunately, the term constructivism is also sometimes used to denote an entirely different realist position taken by advocates of the cognitivist approach to artificial intelligence whereby representations of the external world are constructed through perception. Consequently, one must be careful when interpreting the term constructivism to be clear exactly what is meant: the radical constructivism of phenomenology or the representational constructivism of realism.