In the target article, Murayama and Jach (M&J) argue that explanations of complex behaviors using higher-order motivational constructs, such as the need for competence or the achievement motive, face the “black-box” problem: While these constructs may explain why people act in certain situations, they don't explain the essence of motivation and how motivational processes arise. Moreover, such explanations supposedly face the motivational homunculus problem, where explaining one motivational construct in terms of another risks circular reasoning or infinite regress. To avoid this, the authors suggest that higher-order motivational constructs are psychologically constructed from more fundamental mental computational processes. I will argue that these objections are misguided because they fail to properly distinguish between levels of psychological explanation, particularly those involving personal and subpersonal explanations.

Higher-order motivational constructs elucidate behavior at the personal level, involving the references to whole persons and their psychological states (e.g., Dänzer, Reference Dänzer2023; Dennett, Reference Dennett1969; Drayson, Reference Drayson2012). As pointed out by several philosophers, at this level explanations run out sooner than one might think (see, Dennett, Reference Dennett1969, p. 95). Here explanations often conclude when we understand the reasons behind people's actions, that is, when we can rationalize their behavior (see, e.g., Queloz, Reference Queloz2017). If we want to know what enables reasons to motivate action, we should transition to the subpersonal level, where behavioral, cognitive, and motivational processes are explained by underlying biological, physiological, and computational processes (Drayson, Reference Drayson2012).

Crucially, however, searching for subpersonal explanations (and opening the “black box”) doesn't imply that the initial personal explanation was incomplete. Instead, we are seeking a different kind of explanation for the same thing. Personal explanations are supposed to illuminate what the person was doing and why, that is, in the light of what reasons they were acting, whereas subpersonal explanations primarily offer insights into how these processes were implemented at the computational (i.e., algorithmic) and/or physical levels (Wilkinson, Reference Wilkinson2014; see, also Marr, Reference Marr1982). M&J come close to recognizing this when they assert that their “perspective indicates that high-level motivation constructs reflect higher-level explanations whereas mental computational processes represent lower-level explanations” and add that “[n]o level of understanding should be dismissed as ‘wrong’ (…), because they just explain the behavior for different purposes (…)” (p. 22). However, in formulating the objection to motivational constructs in terms of the “black-box” problem and associating it with the homunculus fallacy, they seem to overlook the fact that this perspective doesn't require that higher-level motivational constructs come prespecified with internal properties that might connect them with physical, biological, or computational variables. This is a job for subpersonal explanation (Drayson, Reference Drayson2012).

So, what problem do M&J manage to solve by providing a subpersonal computational solution to the black-box problem? To answer this question, it should be noted that there are different views on how to understand the relation between the personal and the subpersonal. In the philosophy of psychology, this is labeled the interface problem (Bermúdez, Reference Bermúdez2005, p. 35).

I believe that M&J's solution could be understood as providing a specific perspective on the interface problem. For instance, traditional functionalism posits that constructs at the personal level have causal roles in generating actions. According to such approaches, advancements in cognitive science should enable us to identify computational procedures implementing these roles and eventually discern their implementation in diverse brain processes (for discussion, see, Colombo & Fabry, Reference Colombo and Fabry2021; Jurjako, Reference Jurjako2022). In contrast, M&J propose that personal-level motivational constructs do not play these “energizing” causal roles in generating action; rather, they are the effects of underlying computational processes that actually cause action. This perspective is in line with insights from social psychology suggesting that the reasons people give for their actions, instead of reflecting the actual causal factors giving rise to actions, often serve as post-hoc rationalizations that in a particular cultural context may make sense of observed behaviors (Nisbett & Ross, Reference Nisbett and Ross1980; see, also Cushman, Reference Cushman2020).

However, there is tension in M&J's account. On the one hand, they claim that “high-level motivation constructs are an emergent property of underlying mental computational processes” (p. 12). On the other hand, they claim that such constructs are “a consequence of psychological construction” (p. 12), meaning that they result from “interpreting and categorizing the regularities that exist in behavioral patterns and subjective experiences” (p. 13). If these constructs refer to emergent properties involving behavioral and experiential regularities, then, although they presumably lack energizing causal powers, they denote something real. However, if they are constructed from subjective interpretations of behavioral regularities and these interpretations falsely attribute energizing causality to them, then these constructs lack objective reality. Thus, should we understand these constructs as emergent properties devoid of energizing causal powers, or as products of subjective interpretations?

To avoid this ambiguity, I propose to reinterpret M&J's position as endorsing a kind of fictionalism about the personal/motivation-level constructs (Toon, Reference Toon2023; see, also Dennett, Reference Dennett, Demeter, Parent and Toon2022; Tollon, Reference Tollon2023). Such constructs could be understood as referring to useful fictions that form parts of our narratives about what we think typically causes our actions. Being fictive in this context doesn't mean that such constructs don't play significant roles in our lives and psychological theories (see, Cushman, Reference Cushman2020). They certainly do, just as constructs such as the equator, the average person, and the ideal gas law play significant roles in scientific theorizing and ordinary practices, even though they provide an idealized and thus not completely accurate view of real physical systems. Similarly, we could think of higher-level motivational constructs as referring to idealizations that enable us to capture and predict behavioral regularities at a more abstract level of description (e.g., Dennett, Reference Dennett1989), and also shape such regularities by embodying culturally based prescriptions for desirable behavior (e.g., McGeer, Reference McGeer2015). This view seems to be compatible with M&J's core claim that mental computational processes more precisely capture the causal underpinnings of action. Moreover, it simultaneously resolves the ambiguity in their account and avoids the reification of motivational constructs as emergent properties that lack energizing causal powers.