Framing effects offer a classic example of irrationality. Cases like Aeschylus's Agamemnon and Shakespeare's Macbeth, brilliantly discussed by José Luis Bermúdez, may subvert this consensus view. In my impression, the reason why this is evidently the case is that these are moral dilemmas as the agents face conflicts of moral nature. I will argue that morality is specifically the context where descriptive invariance and ecological relevance may be crucially distinguished.

The famous trolley-like dilemmas are other cases of what Bermúdez refers to as (conscious) quasi-cyclical preferences. Notoriously, switching the lever to deviate the trolley and sacrificing one person (F1o) is preferred to letting the five people on the tracks die (A), which is itself preferred to pushing a man out from the footbridge to stop the trolley (F2o). Even if frame-insensitive responses (F1o > A < F2o) can be primed through order presentation (Schwitzgebel & Cushman, Reference Schwitzgebel and Cushman2012), responses to “complex” frames are perceived as highly conflictual (Koenigs et al., Reference Koenigs, Young, Adolphs, Tranel, Cushman, Hauser and Damasio2007). Complex frames are the crying baby (i.e., would you smother your crying newborn to save yourself and four townspeople hiding from a Nazi incursion?) (Cushman & Greene, Reference Cushman and Greene2012), the transplant (i.e., would you harvest the organs from a pizza boy who entered the ER to save five patients in need of a transplant?) (Thomson, Reference Thomson1976), the cabin boy (i.e., would you kill and eat the cabin boy to save yourself and other starving shipwrecked sailors?) (Harris, Reference Harris2020), triage in a pandemic (Kneer & Hannikainen, Reference Kneer and Hannikainen2022), and many others (Edmonds, Reference Edmonds2014; Koenigs et al., Reference Koenigs, Young, Adolphs, Tranel, Cushman, Hauser and Damasio2007). These situations are logically indistinguishable from the bystander case (i.e., switching the lever). The irrationality of intuitive non-utilitarian responses in these complex frames (i.e., a “no” response) is the reason why these responses, associated with greater activation of emotional brain areas, are discarded by some as ethically unreliable (Greene, Reference Greene and Liao2016; Greene, Sommerville, Nystrom, Darley, & Cohen, Reference Greene, Sommerville, Nystrom, Darley and Cohen2001, Reference Greene, Nystrom, Engell, Darley and Cohen2004). Despite general appeals attributing a crucial role to emotions in morality (Greene & Haidt, Reference Greene and Haidt2002) and even if within a utilitarian (rather than a deontological) perspective (Greene, Reference Greene and Sinnott-Armstrong2008), this so-called dual-system approach ends up in a traditional merger between morality and rationality.

Emotionally different frames of the same outcome (one vs. five) in moral scenarios, I instead believe, may differ informationally by translating into dissociable moral attitudes. Framing it into Bermúdez's discussion, we may call this the non-extensionality of moral preferences, values, and actions. This discourse falls within the concept of “ecological rationality” proposed and defended over the years by Gigerenzer and colleagues (Gigerenzer, Reference Gigerenzer2015; Todd & Gigerenzer, Reference Todd and Gigerenzer2012). Accordingly, we may distinguish logical rationality or “axiomatic definitions of rationality that economic models draw upon” (Todd & Gigerenzer, Reference Todd and Gigerenzer2012, p. 425) from ecological rationality. The latter consists of contextual evaluations of decision processes that go beyond the identification of logical consistency, but rather refer “to how well they match the environments in which they are used” (ibid.). More precisely, ecological rationality expresses in encoding covert scenarios or surplus information between the lines (Sher & McKenzie, Reference Sher and McKenzie2006), a significant ability for moral cognition. In trolley-like dilemmas, identical outcomes are obtained through morally non-identical actions (Cushman, Reference Cushman2013). No less importantly, informational differentiation in these dilemmas is mostly evoked in a native language, where the emotionality of content sounds clearer (Costa et al., Reference Costa, Foucart, Hayakawa, Aparici, Apesteguia, Heafner and Keysar2014; Geipel, Hadjichristidis, & Surian, Reference Geipel, Hadjichristidis and Surian2016).

Sacrificial moral dilemmas are extraordinary cases where the logical accuracy of a frame does not necessarily imply its moral applicability. Interestingly enough, some utilitarian scholars (Kahane, Reference Kahane2012; Kahane, Everett, Earp, Farias, & Savulescu, Reference Kahane, Everett, Earp, Farias and Savulescu2015; Kahane et al., Reference Kahane, Everett, Earp, Caviola, Faber, Crockett and Savulescu2017) have defended this position. They reject the idea that cool harm-endorsing responses to complex frames are expressions of genuine utilitarianism, that is, “impartial concern for the greater good” (Kahane et al., Reference Kahane, Everett, Earp, Farias and Savulescu2015, p. 194), as they less arguably are in the bystander case. The scale they designed, the Oxford Utilitarian Scale (OUS), shows that responses to the two models through which utilitarianism is framed (impartial beneficence vs. instrumental harm) measure different individual traits and moral attitudes. Consistent with previous studies (Bartels & Pizarro, Reference Bartels and Pizarro2011; Glenn, Koleva, Iyer, Graham, & Ditto, Reference Glenn, Koleva, Iyer, Graham and Ditto2010; Koenigs, Kruepke, Zeier, & Newman, Reference Koenigs, Kruepke, Zeier and Newman2012), the scale associates non-clinical psychopathic tendencies with instrumentally harmful responses but, more importantly, not with an endorsement of impartial beneficence. Other morally questionable attitudes, such as rational egoism or lenient moral attitude, were also linked to “purely extensional approaches.” Besides, utility calculation – as intended by its originators – must consider also distal consequences of an action, referring to the class of actions to which the action belongs, and not only proximal consequences as if the action were single and insulated (Warnock, Reference Warnock and Warnock2003).

Sensitivity to emotionally salient frames in the moral domain, for example, to personal force (Greene et al., Reference Greene, Cushman, Stewart, Lowenberg, Nystrom and Cohen2009) and instrumental harm (Everett, Faber, Savulescu, & Crockett, Reference Everett, Faber, Savulescu and Crockett2018), is an efficient heuristic to prevent harm to others in real-life situations that also predicts expectations toward other people's moral behavior. Evolution may have selected for reactive harm-aversion as this favors cooperation, reciprocation, and trust (Cosmides, Guzmán, & Tooby, Reference Cosmides, Guzmán, Tooby, Zimmerman, Jones and Timmons2018). Harm-aversion is linked to a social cognition brain area (i.e., right temporoparietal junction [RTPJ]) (Young, Camprodon, Hauser, Pascual-Leone, & Saxe, Reference Young, Camprodon, Hauser, Pascual-Leone and Saxe2010), and it indeed increases the chances of being chosen as romantic or business partners (Everett, Pizarro, & Crockett, Reference Everett, Pizarro and Crockett2016).

We may also cast doubt on the association – mentioned by Bermúdez – between the recruitment of a brain area (i.e., dorsolateral prefrontal cortex [DLPFC]) and cognitive control processing, especially as this association justifies the appropriateness of utilitarian choices in complex frames (Greene, Nystrom, Engell, Darley, & Cohen, Reference Greene, Nystrom, Engell, Darley and Cohen2004, Reference Greene, Morelli, Lowenberg, Nystrom and Cohen2008). Brain stimulation studies suggest the opposite of what this model predicts, that is, disrupting the right-DLPFC actually increases utilitarian responses (Tassy, Oullier, Mancini, & Wicker, Reference Tassy, Oullier, Mancini and Wicker2013), while enhancing the left-DLPFC increases non-utilitarian responses (Kuehne, Heimrath, Heinze, & Zaehle, Reference Kuehne, Heimrath, Heinze and Zaehle2015). Agreeing with Bermúdez, the literature is now converging on the idea that self-control is closer to emotional regulation than cognitive effort (Fujita, Reference Fujita2011; Inzlicht & Friese, Reference Inzlicht and Friese2021; Magen, Kim, Dweck, Gross, & McClure, Reference Magen, Kim, Dweck, Gross and McClure2014). Accordingly, cognitive load delays reaction time, but does not determine any change in footbridge responses (Greene, Morelli, Lowenberg, Nystrom, & Cohen, Reference Greene, Morelli, Lowenberg, Nystrom and Cohen2008). We discern two aspects of self-control – inhibiting intuitive response and delaying gratification (Duckworth & Kern, Reference Duckworth and Kern2011). Failing to inhibit automatic non-utilitarian choices may be seen as gratification delay if looked at through ecological lenses. These choices may be logically irrational in the short term, as they imply a greater number of deaths, but promote greater social benefits in the longer term, as they limit pernicious attitudes like justifying the sacrifice of innocents.