Full Program »
Information and Explanation: A Dilemma for the Ontic ConceptionI identify a dilemma for Craver's (2014) interpretation of the ontic conception, though my argument will be of interest to anyone who holds a theory according to which explanation is a matter of providing information of some kind (e.g., Jackson and Pettit 1990; Lewis 1986; Skow 2014). I then sketch a solution to this dilemma that respects the motivations of the ontic conception.
The ontic-epistemic debate has shifted twice since Salmon (1986), first from a debate about what explanations do to a debate about what explanations are (Illari 2013). The ontic conception was the idea that scientific explanations are ontic structures; the epistemic conception was the idea that scientific explanations are epistemic states/representations. A second shift occurs in Craver (2014), which focuses on demarcatory and normative constraints on explanation. (According to Illari , Craver holds that this was always the debate.) On this interpretation, attention only to ontic structures is required to demarcate explanation from other scientific achievements, and to distinguish good from bad explanations.
The dilemma arises by conjoining Craver's (2014) interpretation of the ontic conception and another idea that I argue best makes sense of various mechanists' claims and argumentative strategies, namely the idea that mechanistic explanations are representations that provide information about mechanisms. This "informational construal" makes sense of many mechanists' arguments that apparently non-mechanistic explanations are actually mechanism-sketches (i.e. incomplete mechanistic explanations), and their view on the relation between explanations and models. For example, the primary reason that Piccinini and Craver (2011) claim that functional analyses are actually mechanism-sketches is that functional analyses put constraints on the possible mechanisms that implement the functions identified in the analysis. But constraining a space of possibilities is tantamount to uncertainty reduction, which is a classical definition of information (Shannon 1948). Piccinini and Craver basically argue that functional analyses are mechanism-sketches because they provide some information about mechanisms. A similar strategy recurs when mechanists argue that dynamical explanations are mechanism-sketches (Kaplan 2011). Consider also the claim by some mechanists that the representational form of a model (e.g., whether it is boxological or consists of coupled differential equations) is irrelevant to the question of whether or not it provides a mechanistic explanation (Kaplan 2011; Zednik 2011). This claim is implied by the informational construal of mechanistic explanation since many different kinds of representation can provide information about mechanisms.
However, the informational construal is problematic for proponents of the ontic conception because information is usually thought to be relative to an agent's background knowledge (Dretske 1981; Cohen and Meskin  deny this). According to the ontic conception, whether or not a representation is a mechanistic explanation does not depend on the epistemic states of agents. I argue that the solution for a proponent of the ontic conception is to relativize scientific explanation to what I call "common scientific background knowledge," the current total store of propositions that are common knowledge among scientists. This respects Craver's (2014) motivations behind the ontic conception by not tying explanation to individuals' knowledge or understanding.
Washington University in St. Louis