Jens Harbecke (Witten/Herdecke University), Oron Shagrir (The Hebrew University of Jerusalem)
At the heart of the so-called "mechanistic view of computation" lies the idea that computational explanations are mechanistic explanations. Mechanists, however, disagree about the precise role that the environment — or the "contextual level" (Miłkowski 2013) — plays for computational (mechanistic) explanations.
Some mechanists argue that contextual factors do not affect the computational identity of a computing system and, hence, that they do not play an explanatory role vis-á-vis the system’s computational aspects. If anything, contextual factors are important to specify the explanandum, not the explanation (cf. also Kaplan 2011, Miłkowski 2013, Dewhurst 2017, Mollo 2017).
Other mechanists agree that the contextual level is indeed part of the computational level of a computing system, but claim that "[i]n order to know which intrinsic properties of mechanisms are functionally [computationally] relevant, it may be necessary to consider the interaction between mechanisms and their contexts." (Piccinini 2008, 220). In other words, computational explanations involve more than an explication of the relevant mechanisms intrinsic to a computational system. These further aspects specify the causal-mechanistic interaction between the system and its context.
On this poster, we challenge both claims. We argue that (i) contextual factors do affect the computational identity of a computing system, but (ii) that it is not necessary to specify the causal-mechanistic interaction between the system and its context in order to offer a complete and adequate computational explanation. We then discuss the implications of our conclusions for the mechanistic view of computation. Our aim is to show that some versions of the mechanistic view of computation are consistent with claims (i) and (ii), whilst others are not.
We illustrate the following argumentative steps. First, we introduces the notion of an automaton, and we point out that complex systems typically implement a large number of inconsistent automata all at the same time. The challenge is to single out those automata of a system that correspond to its actual computations, which cannot be achieved on the basis of the intrinsic features of the system alone. We then argue that extending the basis by including the immediate or close environment of computing systems does not do the trick. This establishes an externalist view of computation. We then focus on claim (ii) and argue that various different input mechanisms can be correlated with the same computations, and that it is not always necessary to specify the environment-to-system mechanism in order to explain a system’s computations. Finally, we assess the compatibility of claims (i) and (ii) with several versions of the mechanist view of computation.
