Contributed by: Marcin Milkowski, Institute of Philosophy and Sociology, Polish Academy of Sciences

Introduction
In its most generic version, computationalism is a claim that cognition involves computation. It seems that such a generic computationalism might be indeed the presupposition of any research on artificial and natural cognitive systems, as such systems need not be wholly computational for this claim to be true. However, in discussions, computationalism is often assumed in a more specific version:

Cognition is digital computation.

Computation is then assumed to be defined by the Church-Turing formal definition of computation (traditionally referred to as 'Church-Turing thesis', though actually no mathematically proved thesis is involved here).

Many authors objected to the standard computationalism. It's worth noting that these objections might be invalid in case of the generic computationalism.

Making generic computationalism explicit
One of the ways to make computationalism claim true and non-trivial is to define computation in such a way that not all possible processes in the world turn out to be computational, and at the same time computation is not restricted to the synthetic Church-Turing definition. Ron Chrisley (Chrisley 2000) suggests that we should accept transparent computationalism which is a claim that cognition involves computation, whatever computation might turn out to be according to our best theory of computation. This way, computationalism does not depend on the assumption that Church and Turing were right. Transparent computationalism is compatible with the claim that there are hypercomputational processes, i.e., processes that can solve problems that no Universal Turing Machine could solve (at least in the real time). The notion of hypercomputation is usually used for referring to quantum computers. The idea of quantum computers in general was inspired by Richard Feynman's argument that any physical process that seems to regularly process information can be said to be a computational process (Feynman 1982). Note that if we accept this argument (and all proponents of hypercomputation should probably do so) we must also accept that natural brains are computers as they clearly process information, no matter if it is digital or analog processing. Artificial cognitive agents would also appear to be computational, no matter what architecture they realize, as long as they process any information.

Computationalism or merely a computer metaphor?
In early days of cognitive science, researchers used to call mind 'software of the brain'. In other words, they used a computer metaphor to describe the mind/brain relationship. The computer metaphor was not taken to be true literally, as few people supposed that there is any physical correlate of the software level and the hardware level in nervous systems that clearly don't seem to have a classical von Neumann computer architecture. In case of artificial cognitive systems, the computer metaphor does not seem to be so metaphorical at all, as most are implemented using the traditional von Neumann architecture. However, in case of hardware-implemented connectionist network it was argued that they do not implement any algorithms or lack any clear software level (Churchland 1995). The problem with the computer metaphor notion is that anything that has at least one state could be described as realizing a computation (in this case, a simple constant function). Yet it seems clear that computationalism is not a case that anything could be described as computation (as Searle 1992 seems to believe) but that cognitive systems actually implement, or realize, computational processes (for a discussion of realization, see Scheutz 1998; Chalmers 1994). So the stake of computationalism is not just a matter of terminological convention; it's not whether a computer metaphor is attractive. It's about real causal processes obtaining in the cognitive systems.

Post-computational approaches and pancomputationalism
There are some approaches that are not directly incompatible with generic or transparent computationalism, such as dynamical cognition or interactivism. In the interactivism, the cognitive system might actually realize computation but not any computational system is cognitive as well. In other words, according to them cognition is not only computation, and involves non-computational features of the cognitive system. These approaches could seem however incompatible with universal computationalism, or pancomputationalism – a claim that all physical processes are computational in the strict sense (see Chaitin 2005). Pancomputationalism would make any computationalism a trivial claim, and not a bold hypothesis used in cognitive science research. One way to make computationalism a real claim about cognition, and a claim compatible with pancomputationalism, is to say that indeed many cognitive systems might realize computations on many levels, not only on the cognitive level (Milkowski 2007).

References
Chaitin, Gregory. 2005. Meta Math! The Quest for Omega. Pantheon Books.

Chalmers, David J. 1994. “On Implementing a Computation”. Minds and Machines, 4, 391-402.

Chrisley, Ron, 2000. Transparent Computationalism, in New Computationalism: Conceptus-Studien 14, ed. Scheutz, M., Academia Verlag.

Churchland, Paul M., 1995. The Engine of Reason, the Seat of the Soul: A Philosophical Journey into the Brain, MIT Press.

Feynman, Richard, 1982. "Simulating physics with computers". International Journal of Theoretical Physics 21: 467.

Milkowski, Marcin, 2007. “Is computationalism trivial?” in: Gordana Dodig Crnkovic and Susan Stuart (eds.), Computation, Information, Cognition – The Nexus and the Liminal, Cambridge Scholars Publishing, 2007, pp. 236-246.

Scheutz, Matthias. 1998. “Implementation: Computationalism's Weak Spot”. Conceptus JG, 31, 79, 229-239.

Searle, John. 1992. The Rediscovery of Mind. MIT Press.