Joshua Mugg: Response to Commentators

Response to Commentators

Joshua Mugg

Response To Curry

One goal of my project is to provide an empirically informed account of cognitive control that can be marshalled in defense of voluntarism of belief. Curry offers a dilemma for my account of control and voluntarism about belief. By his lights, I am engaged in a homuncular functionalist project of explaining human reasoning by positing smaller and dumber homunculi. Cognitive control, it turns out, are actually three homunculi in a trenchcoat. So far, so good. However, Curry argues, if control is to be further explained (away) through even smaller dumber homunculi, then it is not the case that the agent controls the rerunning of the process. After all, control may not even be one kind on my account—it may be three kinds. Curry argues that I cannot have it both ways, and recommends I stick with the homuncular functionalist project and give up on the belief voluntarism.

I think Curry’s argument shows too much. If his argument is sound regarding my account and its relation to voluntarism of belief, then an analogue argument would be sound regarding a homuncular functionalist account and its relation to voluntarism about action. Here’s how such an argument might go: as one begins to explain my sipping of a cask-conditioned ale in functionalist term, control will likely turn out to divide into multiple kinds. So, it will turn out, my action of drinking the cask-conditional ale won’t be voluntary. But surely it is voluntary! Something must have gone wrong in the second argument. So, something must be wrong with the first.

Where is the error in the argument? Admittedly, I have not committed to a particular account of voluntariness of action. What I tried to show in the book is that some belief formation processes are as voluntary as shooting a ball through a hoop. Curry’s argument seems tacitly committed to something like: “if a [mental] action is voluntary, then it must be controlled by one thing: the agent.”  Granted, there are philosophers—defenders and detractors of voluntarism—who do commit to the idea that control must be of a sui generis kind to be genuine agency. However, there are weaker options open. One option is that all that is needed is that the agent has a new token power, and that could be within any of the three homunculi in the trenchcoat. Another is that the three homunculi do not really complete with the subject’s agency because mental agency just is the composition of those homunculi. So, I’m hopeful that we can have a causal description of belief-forming processes that remain consistent with a kind of voluntarism. If that’s possible, then I can have my functionalist account of control and belief voluntarism.

Response to Walker

Walker picks up on the theme of control, specifically how it fits into the normative project of supporting an ethics of belief. In my view, voluntariness merely opens the possibility of an ethics of belief. One objection to the possibility of the ethics of belief is that we lack any control over our beliefs, and it makes little sense to hold people accountable for what they cannot control. So my argument for voluntarism of belief is really just blocking one argument against the ethics of belief.

As Walker points out, greater control does not entail more normatively correct beliefs, whether morally correct or epistemically correct. Instances of wishful thinking may lead us to less accurate beliefs akin to false-positives, assuming our hopes are not oriented toward truth. However, there may be instances in which a subject reruns the process resulting in a more accurate belief, as would have been the case had the restaurant goer stopped to think about why he thought the well-dressed black man was a valet. The upshot is that control alone won’t do the normative work in an ethics of belief.

Walker suggests that motivation for the rerunning of the reasoning process may be what really matters for an ethics of belief. I haven’t committed myself to any particular normative account (epistemic or moral), but Walker’s suggestion seems like a good one. Within the ethics of belief, I am particularly interested in cases of epistemic-moral dilemmas, and looking at motivation might offer a way of sorting the good rerunning from the bad. Of course, we would also need a story about how to evaluate these motivations.

I hope that the Soundboard Account and the wider account of belief I’ve offered might provide a framework for thinking about the epistemic-moral relation within the ethics of belief. It is not as though we have two systems—one governed by the epistemic ought and another by the moral ought. Cognitively, the two are mixed together, even if we can step back and evaluate a single process from either perspective.

Response to Rumana

Rumana’s criticism of the book is in its framing. According to her, what I have really shown is that many psychologists (and, I would hasten to add, some philosophers) have inappropriately interpreted the dual-process theory in a realist way. Rather than interpreting the low-capacity versions of DPT as idealizations for the sake of testing, I interpret them as instantiations of the more general DPT that these low-capacity versions are supposed to be gesturing toward. In my view and, I think, Rumana’s, these low-capacity DPTs are in no way gesturing toward a unified version of DPT.  Rumana asks for a better rival for the Soundboard Account than this misinterpreted version of DPT, such as an dual-soundboard account. I have two responses. First, the fact that some philosophers and psychologists do seem to think low-capacity versions of DPT are gesturing toward a unified version makes that account fair game for criticism. Second, the dual-soundboard theory would be a substantively different account of human reasoning than any version of DPT currently on offer, and, as Rumana says, the onus will be on dual process advocates to build such an account.

One concern Rumana has for the Soundboard Account is that it is, in principle, untestable. I think the Soundboard Account is testable, but I also think of testability as contrastive (following Sober’s work on testability). So, the question is less how to test the account tout court and more so how to test it against specific articulations of DPT (e.g. De Neys’s Logical Intuition Model). As Rumana points out, I argue that if there are reasoning responses that are contradictory and arise simultaneously, this would be evidence that there are (at least) two processes: one for each response.

Here is a suggestion of how one might develop an experiment to test my account against Logical Intuitionist Model of DPT. On this account, there are multiple simultaneously operating Type-1 processes, and conflicts between their outputs trigger Type-2 processing. The Type-2 response is sometimes then able to override the Type-1 responses. This overriding requires cognitive effort. Therefore, those subjects who come to believe the Type-2 response answer after the reasoning task must expend more cognitive energy (and so be more cognitively tired) than those who offer Type-1 responses.These theoretical points suggest a way we might test for the existence of contradictory beliefs arising simultaneously from distinct reasoning processes: subjects who are made aware of their reasoning errors generated by Type-1 processing will show greater cognitive depletion than similar subjects who are not made of their reasoning errors generated by Type-1 processing.

Here’s a suggestion of how we can translate these theoretical points into a rough and ready experimental procedure. In step one, subjects undertake a reasoning task that might involve contradictory beliefs generated simultaneously (e.g reasoning problems that typically result in the conjunction fallacy). Subjects who offer the Type-2 response should be dismissed. For step 2, the Type-1 responders are divided into two groups: A and B. Researchers activate group A’s Type-2 processing through teaching, such that they now respond to the question with the Type-2 response. Group B receives instruction on some unrelated problem as a control. In step 3, subjects complete a measurement of cognitive depletion, such as the Stoop Test. There will be greater cognitive depletion in Group A if they possess contradictory beliefs that are generated simultaneously.