Cushman makes a compelling case that rationalization is a form of what he calls “representation exchange” and that such representation exchange is itself a crucial feature of action-guiding cognition. We agree with Cushman that these processes of representation exchange, including rationalization, are broadly adaptive. We are less convinced, however, by his claim that rationalization is adaptively advantageous because it allows humans to extract valuable information (“true beliefs and useful desires”) from the highly evolved inclinations of our non-rational action-guiding systems. We see two problems with this proposal, one theoretical and one empirical.
We begin with the theoretical problem. Cushman recognizes that our actions are generated by a complex combination of rational (beliefs, desires) and non-rational (instincts, habits, norm compliance) impulses, and he suggests that rationalization “constructs new beliefs and desires where none had existed, to extract information from the non-rational processes that influence our behavior” (target article sect. 1.1, para. 7). Thus, if I must decide which route to take as I walk my dog, and the complex product of the non-rational and rational influences on my behavior leads me to walk along the road rather than along the river, I might rationalize that decision by deciding that I am afraid of the river route. On Cushman's account, I will then come to accept and adopt that rationalization (my fear of the river route or a belief about its danger) as a further, conscious motivation for avoiding the river. But creating this new, additional motivation will not change the impact of those same non-rational action-guiding impulses on my behavior. Thus, if I face the same (or a similar) decision tomorrow, the very same non-rational action-guiding impulses will remain in place and continue to influence my behavior, ensuring that my degree of aversion to the river route is not explained by the magnitude of my new consciously accessible fear (or my beliefs about its danger) and triggering a new round of rationalization that strengthens or intensifies that fear in response. (Indeed, if these non-rational impulses fully preserve their influence, the degree of mismatch between my actions and my consciously accessible motivations should remain just as large as it was initially.) It seems that this process of repeated rationalization will continue as long as the relevant behavior is generated by a combination of non-rational and rational impulses, ultimately ensuring that the accumulated wisdom of my non-rational action-guiding impulses is substantially overrepresented in our actual (complex, multisystem) decisions about what to do (e.g., leaving me absolutely terrified of taking the river route). This would not be cause for concern if we thought the guidance of our non-rational action-guiding systems were always correct, but in that case, there would be no evolutionary advantage in making such decisions accessible to rational influence in the first place.
The empirical problem is that even if rationalization invariably produces more fitness-maximizing choices and decisions, it sometimes does so by giving us less accurate beliefs about the world. One example is Liu and Ditto's (Reference Liu and Ditto2012) work on moral coherence: Briefly, subjects who are induced by argument to shift their views concerning the moral defensibility of the death penalty will also shift their views about the extent to which the death penalty is practically effective in deterring crime. And we ourselves have shown that subjects will judge a child left alone in precisely the same circumstances to be in significantly more danger if the parent leaves for a morally unacceptable reason (e.g., an adulterous affair) than a morally neutral reason such as going to work (Thomas et al. Reference Thomas, Stanford and Sarnecka2016). In these studies, it seems that representation exchange is inducing subjects to modify their factual beliefs in ways that do not increase their accuracy but instead make them better cohere with the subjects’ own moral judgments. This may well ensure more adaptive responses (ones that better reflect the community's normative views, for example), but if so, this is achieved by making the subject's beliefs less accurate or responsive to relevant evidence.
The challenges we have presented also suggest an alternative to Cushman's hypothesis concerning the adaptive benefits of rationalization itself. Even if it does so at some cost to the accuracy of our beliefs, the one thing representation exchange undeniably increases is the coherence of our motivations for action and (therefore) the predictability of our behavior to others. There is reason to think that behaving predictably has adaptive value in its own right, as emphasized by those evolutionary theorists who appeal to mechanisms of partner choice, reputation management, and the virtues of predictable social partners more generally in seeking to understand the evolution of human ultrasociality (e.g., Baumard et al. Reference Baumard, André and Sperber2013; Stanford Reference Stanford2018; Tomasello, Reference Tomaselloin press). Thus, while we agree with Cushman that representation exchange is broadly adaptive and will sometimes increase the accuracy of our beliefs, we doubt that this is the only or even the most important way in which rationalization in particular increases our fitness. We suggest that rationalization in humans is adaptive in large part because it renders us more predictable (and therefore more attractive) partners for one another in the sorts of hypercooperative social structures on which human societies depend.
Cushman makes a compelling case that rationalization is a form of what he calls “representation exchange” and that such representation exchange is itself a crucial feature of action-guiding cognition. We agree with Cushman that these processes of representation exchange, including rationalization, are broadly adaptive. We are less convinced, however, by his claim that rationalization is adaptively advantageous because it allows humans to extract valuable information (“true beliefs and useful desires”) from the highly evolved inclinations of our non-rational action-guiding systems. We see two problems with this proposal, one theoretical and one empirical.
We begin with the theoretical problem. Cushman recognizes that our actions are generated by a complex combination of rational (beliefs, desires) and non-rational (instincts, habits, norm compliance) impulses, and he suggests that rationalization “constructs new beliefs and desires where none had existed, to extract information from the non-rational processes that influence our behavior” (target article sect. 1.1, para. 7). Thus, if I must decide which route to take as I walk my dog, and the complex product of the non-rational and rational influences on my behavior leads me to walk along the road rather than along the river, I might rationalize that decision by deciding that I am afraid of the river route. On Cushman's account, I will then come to accept and adopt that rationalization (my fear of the river route or a belief about its danger) as a further, conscious motivation for avoiding the river. But creating this new, additional motivation will not change the impact of those same non-rational action-guiding impulses on my behavior. Thus, if I face the same (or a similar) decision tomorrow, the very same non-rational action-guiding impulses will remain in place and continue to influence my behavior, ensuring that my degree of aversion to the river route is not explained by the magnitude of my new consciously accessible fear (or my beliefs about its danger) and triggering a new round of rationalization that strengthens or intensifies that fear in response. (Indeed, if these non-rational impulses fully preserve their influence, the degree of mismatch between my actions and my consciously accessible motivations should remain just as large as it was initially.) It seems that this process of repeated rationalization will continue as long as the relevant behavior is generated by a combination of non-rational and rational impulses, ultimately ensuring that the accumulated wisdom of my non-rational action-guiding impulses is substantially overrepresented in our actual (complex, multisystem) decisions about what to do (e.g., leaving me absolutely terrified of taking the river route). This would not be cause for concern if we thought the guidance of our non-rational action-guiding systems were always correct, but in that case, there would be no evolutionary advantage in making such decisions accessible to rational influence in the first place.
The empirical problem is that even if rationalization invariably produces more fitness-maximizing choices and decisions, it sometimes does so by giving us less accurate beliefs about the world. One example is Liu and Ditto's (Reference Liu and Ditto2012) work on moral coherence: Briefly, subjects who are induced by argument to shift their views concerning the moral defensibility of the death penalty will also shift their views about the extent to which the death penalty is practically effective in deterring crime. And we ourselves have shown that subjects will judge a child left alone in precisely the same circumstances to be in significantly more danger if the parent leaves for a morally unacceptable reason (e.g., an adulterous affair) than a morally neutral reason such as going to work (Thomas et al. Reference Thomas, Stanford and Sarnecka2016). In these studies, it seems that representation exchange is inducing subjects to modify their factual beliefs in ways that do not increase their accuracy but instead make them better cohere with the subjects’ own moral judgments. This may well ensure more adaptive responses (ones that better reflect the community's normative views, for example), but if so, this is achieved by making the subject's beliefs less accurate or responsive to relevant evidence.
The challenges we have presented also suggest an alternative to Cushman's hypothesis concerning the adaptive benefits of rationalization itself. Even if it does so at some cost to the accuracy of our beliefs, the one thing representation exchange undeniably increases is the coherence of our motivations for action and (therefore) the predictability of our behavior to others. There is reason to think that behaving predictably has adaptive value in its own right, as emphasized by those evolutionary theorists who appeal to mechanisms of partner choice, reputation management, and the virtues of predictable social partners more generally in seeking to understand the evolution of human ultrasociality (e.g., Baumard et al. Reference Baumard, André and Sperber2013; Stanford Reference Stanford2018; Tomasello, Reference Tomaselloin press). Thus, while we agree with Cushman that representation exchange is broadly adaptive and will sometimes increase the accuracy of our beliefs, we doubt that this is the only or even the most important way in which rationalization in particular increases our fitness. We suggest that rationalization in humans is adaptive in large part because it renders us more predictable (and therefore more attractive) partners for one another in the sorts of hypercooperative social structures on which human societies depend.