DJ Homunculus
Devin Sanchez Curry
The first aim of Josh Mugg’s From Human Reasoning to Belief is to propose a new, empirically viable theory of central cognition (and especially belief fixation and revision): the Soundboard Account of Human Reasoning. The book achieves this aim admirably, both detailing how extant evidence supports the Soundboard Account and laudably proposing specific ways to test the theory directly against its chief competitors (namely, dual process theories and Eric Mandelbaum’s Spinozan theory). Mugg’s takedown of the gamut of dual-process theories is particularly convincing, as is his alternative interpretation of measures of implicit attitudes as tracking initial or intermediate (as opposed to final) outputs of reasoning processes.
But the book also has a second aim: to vindicate voluntarism about belief. As Mugg admits, the fundamentals of the Soundboard Account of Human Reasoning laid out in Chapter 3 are (in principle) compatible with a Spinozan theory of belief formation (2026: 65). Mugg develops an empirical case against going in for a Spinozan version of the Soundboard Account, but his urgency in making that case stems partly from the fact that “Spinozanism would imply a strong form of involuntarism about belief, having implications in ethics and epistemology” (81)—implications that Mugg would like to reject.
By contrast, Mugg’s non-Spinozan Soundboard Account is supposed to support voluntarism about belief, with attendant nice implications in ethics and epistemology. On Mugg’s view, “a reasoning process can be as controlled as shooting a basketball through a hoop” (100), or, to use the book’s central metaphor, a human being can control their reasoning to roughly the same extent as a DJ can control their remix. These analogies serve to illustrate the limits of control: “just as a DJ must work with the system and tracks they have, so must humans work with the modes of operation and evidence they have” (112). But they also serve to underscore the assertion that belief fixation and revision are voluntary processes, for which believers are rightly held responsible. A free-throw shooter generally has sufficient control to be responsible for whether they swish or brick, and a DJ generally has sufficient control to be responsible for whether they produce bops. So too, on Mugg’s view, human beings generally have sufficient control to be ethically and epistemically responsible for what they believe.
Here’s the rub. Mugg’s defense of voluntarism seems to require that the metaphorical DJ who runs the soundboard responsible for human reasoning is, well, the whole human being. After all, it’s the human believer whom we want to hold responsible. But the DJ isn’t a human: it’s a homunculus.
Actually, it might be three homunculi in a trenchcoat. Mugg floats this possibility when considering an objection to his initial sketch of the Soundboard Account.
One might object to the sound mixing board metaphor on the grounds that it seems to posit someone or something adopting the output or rerunning the process, which raises worries about a homunculus fallacy. Applied to the Soundboard Account, the worry is that there is a homunculus in the head that has a reasoning sound mixing board. She listens to the output of the reasoning sound mixing board and then decides whether to adopt or rerun, and which modes to alter or cognitive states to recruit. But now the account is pushed back because we may ask on what basis the homunculus decides. Is there a smaller homunculus in its head with its own sound mixing board? … I have two responses. First, positing a homunculus is not problematic when the homunculus is simpler than the subject—when the homunculus does not do the very thing that is supposed to be explained. In the case of rerunning the process or accepting the output, the homunculus has only two options: accept or rerun. There is no a priori reason to think that mode alteration and cognitive state recruitment are carried out by the same subsystem. Thus, insofar as I am positing homunculi, they are not of the worrying sort. Second, I take these questions to be fruitful ground for further empirical exploration of the Soundboard Account: under what conditions do subjects rerun reasoning processes? Under which conditions do subjects alter modes? Which modes? (Mugg 2026: 45; my italics)
In this passage, Mugg is reciting a familiar, two-step homuncular functionalist refrain. First, it’s fine to posit homunculi if doing so doesn’t lead to a vicious regress (since they decompose into simpler mechanisms). Second, it’s the job of cognitive science to explain how homunculi work (by modeling them as decomposing into simpler mechanisms).
This is a perfectly reasonable response to the anticipated objection about a vicious regress of homunculi. The Soundboard Account and homuncular functionalism fit together neatly. However, adopting homuncular functionalism opens the door to another classic worry about homunculi, which the book does not address: if the little monsters in your mind are in control, then you aren’t.
Now, the force of this worry might not be so great if there were just one DJ Homunculus in charge of central cognition, who could plausibly be understood as the mechanism by which the whole person exerts control over her beliefs. To the extent that this one unified homunculus purposively fiddled with their soundboard’s sliders (altering modes, recruiting new webs of belief, and rerunning reasoning processes), the whole person could plausibly be said to voluntarily control what she came to believe through the little DJ in her skull.
The problem is that Mugg (quite rightly) conceives of control as being a more diffuse phenomenon, and his (also quite reasonable) inclination towards homuncular functionalism pushes him towards positing a team of DJ homunculi, each of whom is responsible for generating a unique facet of control. As Mugg notes in the passage quoted above, the three main DJing “decisions”—the “decisions” (a) to rerun a reasoning process, (b) to alter the modes of a reasoning process, and (c) to recruit new cognitive states (or an entire new web of belief) into a reasoning process—may well be executed by three distinct subpersonal mechanisms. On Mugg’s view, each of these “decisions” plays a key role in rendering the reasoning process more controlled (47).
The less traditionally homuncular the three little monsters in control—the more they resemble dumb rote cogs rather than little rational agents—the harder it is to justify the claim that their combined operations constitute voluntary choices on the part of the whole person. And the book doesn’t say much to make mode alteration (or the “often automatic” (109) recruitment of new webs of belief, or even the binary “decision” to accept or rerun) sound like a rational process. What sounds potentially rational is the whole overarching project of running and rerunning the reasoning process, flipping switches and tweaking sliders as one goes. Again, if there’s one unified homunculus governing that whole project, then the doxastic voluntarist has some ground to stand on. But if DJ Homunculus actually decomposes into a fairly disunified hodgepodge of subpersonal algorithms, each of which takes a hodgepodge of other subpersonal states as inputs, then it looks much less like the agent is in control. Note that the call whether to rerun a reasoning process depends not just on the outputs of the previous reasoning process but also, crucially, on whether the other homunculi have “decided” to alter modes of operation or recruit new (webs of) beliefs. To the extent that these homunculi make their “decisions” separately from (but not independently of) one another, the process of belief formation is controlled not by a single unified subject but by a collection of distinct subpersonal mechanisms, each of which routinely operates subconsciously, and each of which makes “decisions” in ways that are dictated mainly by factors outside of their own homuncular control.
None of the above is meant to impugn the Soundboard Account itself. The homuncular nature of the metaphorical DJ is problematic only insofar as it provides reason to doubt the inference from the Soundboard Account to voluntarism about belief. The Soundboard Account remains a theoretically attractive and empirically plausible alternative to dual-process theories, even if it fails to put the control of belief formation back in the hands of the whole person. Nevertheless, if the Soundboard Account does little to vindicate voluntarism about belief, then much of the impetus to develop the Soundboard Account in a non-Spinozan direction will be gone. Spinozists still need to reckon with Mugg’s specific objections in Chapter 4; but if those objections can be overcome, then it may be worth forging a symbiotic relationship between the Spinozan theory of belief formation and the Soundboard Account of Human Reasoning.
Reference
Mugg, J. (2026). From Human Reasoning to Belief: An Empirical Account. Routledge.