Cushman argues that after engaging in behavior triggered at least in part by processes other than thinking (e.g., instincts and habits), people infer reasons that rationalize their behavior. These rationalizations, in turn, serve to adjust people's beliefs and desires to help guide future decisions and behaviors. This defense of the rationality of rationalization is insightful, but in our view, its scope and contribution are limited in two important ways.

First, the analysis posits that rational behavior is driven straightforwardly from “a mechanism that chooses actions by maximizing the satisfaction of your desires, given your beliefs” (target article sect. 3.1, para. 1). Thus, the model glosses over situations in which the putatively rational components – namely, desires and beliefs – do not align in perfect harmony. Yet such non-alignment is typical of consequential decisional dilemmas that people face in everyday life. Consider the numerous conflicting, uncertain, and incommensurable factors involved in choosing which college to attend, whether to accept an apology for an insult from a colleague, or even just deciding what to order from the food truck. A growing body of research shows that people reach decisions in such dilemmas neither by rational forward inference nor by backward-oriented rationalization. Rather, a subtler form of reasoning distorts the task attributes to make one course of action emerge as dominant over rival choices.

In models of coherence-based reasoning (Holyoak & Simon Reference Holyoak and Simon1999; Read & Simon Reference Read, Simon, Gawronski and Strack2012; Simon Reference Simon2004; Simon & Holyoak Reference Simon and Holyoak2002; Spellman et al. Reference Spellman, Ullman and Holyoak1993), complex decision situations are represented by networks in which the relevant variables are interconnected via excitatory and inhibitory links. Constraint satisfaction mechanisms settle the network into a stable state of coherence, in which mutually supportive connections (i.e., those that “go together”) activate one another and collectively inhibit their rivals (cf. McClelland & Rumelhart Reference McClelland and Rumelhart1981). Thus, variables in the network come to cohere with the emerging decision: Those that support the winning conclusion are strongly endorsed, whereas those that support the rejected interpretation are suppressed or rejected (Holyoak & Thagard Reference Holyoak and Thagard1989; Read et al. Reference Read, Vanman and Miller1997; Thagard Reference Thagard1989). This spreading-apart transforms difficult choice dilemmas into obviously correct decisions and thus serves the adaptive goal of affording confident choice in the face of decisional conflict (Glöckner et al. Reference Glöckner, Hilbig and Jekel2014; Janis & Mann Reference Janis and Mann1977; Read & Simon Reference Read, Simon, Gawronski and Strack2012; Simon & Holyoak Reference Simon and Holyoak2002). This form of reasoning does not typically lead to irrational decisions (as in preference reversal); more often, it exaggerates small initial preference differentials into large disparities, thus bestowing a patina of rationality on the eventual choice.

Coherence-based reasoning has been demonstrated across a wide range of decisional domains, including social judgment (Simon et al. Reference Simon, Stenstrom and Read2015), legal reasoning (Holyoak & Simon Reference Holyoak and Simon1999; Simon et al. Reference Simon, Pham, Le and Holyoak2001), moral reasoning (Holyoak & Powell Reference Holyoak and Powell2016), evidence evaluation (Carpenter et al. Reference Carpenter, Yates, Preston and Chen2016; Engel & Glöckner Reference Engel and Glöckner2013; Simon et al. Reference Simon, Snow and Read2004b), probabilistic judgments (Glöckner et al. Reference Glöckner, Betsch and Schindler2010), choice between jobs (Simon et al. Reference Simon, Krawczyk and Holyoak2004a; Simon & Spiller Reference Simon and Spiller2016), consumer choice (Simon & Spiller Reference Simon and Spiller2016), and attitudes toward war (Spellman et al. Reference Spellman, Ullman and Holyoak1993). Similar predecisional distortion of task attributes has been observed in other research traditions, spanning consumer decision making (Chaxel et al. Reference Chaxel, Russo and Kerimi2013; Russo et al. Reference Russo, Medvec and Meloy1996; Reference Russo, Carlson, Meloy and Yong2008), evidence evaluation (Carlson & Russo Reference Carlson and Russo2001), medical decision making (Kostopoulou et al. Reference Kostopoulou, Russo, Keenan, Delaney and Douiri2012), and risky decision making (DeKay et al. Reference DeKay, Patiño-Echeverri and Fischbeck2009).

Second, Cushman's analysis posits that rationalization always follows action (as illustrated in Fig. 2 in the target article). This temporal ordering echoes Festinger's (Reference Festinger1957; Reference Festinger1964) insistence that cognitive dissonance arises only after the person has engaged in counter-attitudinal behavior or has committed to an imperfect choice (see also Brehm Reference Brehm1956). This viewpoint was challenged almost immediately. Bruner (Reference Bruner and Bruner1957) criticized the ex post facto nature of dissonance theory, claiming that “the most interesting aspects of cognition are those that precede the making of decisions rather than those that follow [it],” adding that the theory amounted to a “rather autistic tradition” (p. 152). Abelson (Reference Abelson1983) lamented that cognitive dissonance theory had been reduced to explaining how people “recover from experimentally engineered major embarrassments” (p. 43). Abelson insisted that the behavior of people who have been forced to make “damned fools of themselves” is a distraction from the broader potential applications of structural dynamics (see also Berkowitz & Devine Reference Berkowitz and Devine1989). As Heider (Reference Heider, Holland and Leinhardt1979) argued, the tendency toward consistency should not be interpreted as merely repairing disturbances of balance; rather, it implies reaching out to bring the various pieces of the cognitive field into consonance. Indeed, research findings indicate that coherence-driven distortions largely occur prior to the time at which the decision is made or an action taken (DeKay Reference DeKay2015; Holyoak & Simon Reference Holyoak and Simon1999; Russo et al. Reference Russo, Meloy and Medvec1998; Simon et al. Reference Simon, Pham, Le and Holyoak2001; for reviews, see Brownstein Reference Brownstein2003; Simon & Holyoak Reference Simon and Holyoak2002).

A nice illustration of subtle bidirectional reasoning is provided by the reflective equilibrium model (Daniels Reference Daniels and Zalta2003). As discussed in the target article, when making a considered judgment, the reasoner seeks to harmonize principled rules or reasons with their intuitive judgment about the particular case. When the two do not naturally coincide, the thinker modifies them both to bring them into alignment at a state of equilibrium. Reflective equilibrium is typically considered to be a conscious process, whereas coherence shifts appear to occur mostly beneath conscious awareness (Holyoak & Simon Reference Holyoak and Simon1999); nonetheless, all these processes involve bidirectional reasoning that generates an optimal choice and action given the constraints of the situation.

In sum, by focusing exclusively on the construction of ex post facto rationalizations, Cushman's analysis fails to capture the ex ante impact of bidirectional reasoning on human decision making. We contend that the scope of coherence-based reasoning is far broader, and addresses more important cognitive challenges, than does simple rationalization to justify decisions already reached and actions already taken.