The two perspectives

Once researchers tried to provide a more realistic approach to describe how we decide, it became clear that our rationality was bounded by several factors. This view is shared by all of the above-mentioned authors: Herbert Simon, Amos Tversky, Daniel Kahneman, Gerd Gigerenzer, Richard Thaler and Cass Sunstein. They all claim that our rationality falls short of the ideal prescribed by classical and neoclassical economists. That they all categorize our rationality as bounded does not imply that they all share the same type of perspective on this matter. Some authors divide these perspectives into ‘pessimistic’ and ‘optimistic’, as is the case of Richard Samuels (Samuels et al., Reference Samuels, Stich, Bishop and Elio2002). Following this line of thought, Andrea Polonioli (Reference Roeser2012) posits that while Kahneman, Tversky and nudge theory present a pessimistic view of our rationality, Gigerenzer and some within the school of evolutionary psychology defend a more optimistic approach. The difference is based on how rationality is perceived. Gigerenzer and evolutionary psychologists accuse nudge theorists and those behind the heuristics and biases research program of describing our behavior as merely irrational. This irrationality, they claim, may represent some evolutionary advantage that had to be matched to a specific environment in order to be efficient. Biases do not separate us from effective decisions; they are rational shortcuts that we can apply to our decision processes in a specific environment to increase our efficiency (Polonioli, Reference Polonioli2012, p. 136). From a similar perspective, Grüne-Yanoff and Hertwig embrace the idea that the use of boosts instead of nudges can increase the competences of individuals. While nudges limit themselves to behavioral changes, boosts target competences (Hertwig & Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017, p. 977), skills and knowledge by enlisting human cognition, the environment or both. They are built under the assumption that our rationality, while bounded, is not divided into system 1 (fast, emotional) and system 2 (slow, deliberative, rational). Instead, our cognition is malleable. Boosts, they claim (Hertwig & Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017, p. 974), require the active cooperation of individuals. Accordingly, Hertwig (Reference Hertwig and Grüne-Yanoff2017) does not see boosts as being better than nudges. In a way, boosts are efficient tools that, in some scenarios, are preferable to nudges, but in other scenarios are not. While different, Hertwig defends the use of both boosts and nudges. Those differences are important, and they clarify the view of rationality that all of these approaches have:

The ecological rationality approach, by contrast, criticizes the use of nudges, since they fail to prove that people are impossible to educate (Gigerenzer, Reference Gigerenzer2015). Facing the choice of educating them or nudging, governments and institutions should wish to do the former. Using ecological rationality may lead one to change the environment and how information is presented so that they match the cognition of an individual, or it may address human cognition itself: “The true alternative to nudging is education: making children and adults risk savvy” (Gigerenzer, Reference Gigerenzer2015, p. 375). According to this perspective, nudging without education is a way to infantilize citizens.

Boosts and ecological rationality propose similar approaches, since both highlight the importance of educating citizens so that they can exercise their own agency. Since boosts are an educational resource, they are placed on the side of ecological rationality. Nudges are changes in the choice architecture, and some of them can also be educational. In this paper, following the notion of adaptive rationality and the authors included under that umbrella proposed by Polonioli (Reference Polonioli2016b), Gigerenzer's view and the boosts proposed by Hertwig and Grüne-Yanoff will be considered under the educational perspective:

While those proposing boosts consider them to be an intervention that, despite sharing some similarities with nudges, is essentially different, others (Sims & Müller, Reference Sims and Müller2018) claim that they may be referring to the same types of mechanisms if we look at them under a normative perspective.

A closer analysis of both perspectives (nudges and educational tools) unveils essential similarities: exploiting our rational limitations to increase the efficiency of our decisions by altering the environment is like claiming that determined heuristics are advantageous in certain environments, or that we may boost the competences of people by presenting probabilities and percentages in natural frequencies. The difference, I argue here, is related to a philosophical view of the nature of rationality more than a difference in how rationality works. I share Richard Samuels’ idea that the distinction between both perspectives is only apparent (Samuels et al., Reference Samuels, Stich, Bishop and Elio2002).

While they both are based on a similar approach to bounded rationality, Gigerenzer argues that the heuristics and biases research program poses a vision of human decision-making as irrational. He claims that these irrationalities are advantages instead of errors:

The differences between both views become even clearer when referring to visual illusions. Kahneman divides metaphorically our thinking processes into two systems: one intuitive and automatic, and the other deliberative and rational. Kahneman proposes that these illusions are systematic errors made by system 1, since this system operates automatically and cannot be turned off. These errors are difficult to prevent and biases cannot always be avoided (Kahneman, Reference Kahneman2011, p. 25). These illusions seem to be an example of cognitive illusions. Gigerenzer, on the other hand, argues that those errors are an essential part of our rationality: they are good errors and they cannot be considered as flaws: “Without them we wouldn't recognize the objects around us. If a system does not make errors, it is not intelligent. Visual illusions in fact demonstrate the success rather than the failure of cognition” (Gigerenzer, Reference Gigerenzer2014, p. 47).

Gigerenzer criticizes the division of systems employed by Kahneman, claiming that intuition (system 1 in Kahneman's metaphor) is not opposed to rationality:

Although Kahneman talks about systematic errors, he does not contradict this thesis – both systems must interact for an efficient thinking process. He posits that this collaboration between systems can be very practical since “it minimizes effort and optimizes performance” (Kahneman, Reference Kahneman2011, p. 25). But this type of collaboration is far from the malleability of our cognition necessarily proposed by the proponents of adaptive ecological rationality (Sims & Müller, Reference Sims and Müller2018, p. 21).

Kahneman does not mention that education is impossible. System 2 can provide vigilance regarding the visual illusions of system 1. We can learn about these illusions and follow the more deliberative system 2. We can be educated and resist illusions after recognizing them. “Errors can be prevented only by the enhanced monitoring and effortful activity of system 2. As a way to live your life, however, continuous vigilance is not necessarily good” (Kahneman, Reference Kahneman2011, p. 25). Kahneman believes that although education is possible, it may not be a good thing to try to monitor system 1 errors or cognitive illusions. We can learn how to “recognize situations in which mistakes are likely” (Kahneman, Reference Kahneman2011, p. 25). Those authors who advocate for the use of boosts also criticize the hypothesis of the two systems, arguing that our cognition is malleable (Sims & Müller, Reference Sims and Müller2018, p. 21). The notion of the two different systems is also questioned by Matteo Colombo (Reference Colombo, Sprevak and Colombo2018), who posits that recent research shows that there are neurological pathways that connect both systems.

The libertarian paternalism of nudges uses this view of rationality as a point of departure to develop a theory in order to make our lives easier. After providing some examples of visual and cognitive illusions, Kahneman claims that the “message of those examples is not encouraging” (Kahneman, Reference Kahneman2011, p. 25). Since it is not practical to be constantly vigilant for possible errors made by system 1, nudges provide an efficient mechanism that may increase the efficiency of our decisions.

Gigerenzer proposes that people do not have to be at the mercy of institutions or other people in order to avoid having to monitor system 2 for possible errors. Those institutions and the people in them are also agents with bounded rationality, conflicted interests and illusions that may affect how they nudge. He asserts that people have to learn how to think on their own (agency) and take responsibility for their decisions. People can learn about risks and uncertainty, and governments, instead of investing in nudge units, should invest in people's knowledge and risk literacy.

According to Andrea Polonioli (Reference Polonioli2015), Keith Stanovich criticized Gigerenzer's adaptive rationality, claiming that, according to Gigerenzer's evolutionary approach, if heuristics had the role that he and his group assign to them and they have arisen due to adaptation, then they would be more universal. According to Stanovich, this is not the case: “Stanovich's reported heterogeneity in the use of heuristics seems problematic, as one would expect the use of heuristics to be far closer to universal if adaptationist pressures had led to their use” (Polonioli, Reference Polonioli2015, p. 57). Polonioli asserts that Stanovich's argument is not robust enough. Stanovich's argument rests on the notion that if heuristics arose due to adaptation, they would not be heterogeneous. Polonioli thinks that this is false: “[A]n adaptationist framework might account for heterogeneity in the use of heuristics” (Polonioli, Reference Polonioli2015, p. 57).

There is another criticism that Stanovich presents against Gigerenzer's adaptive rationality that Polonioli addresses: the case that people with higher cognitive ability follow normative rationality. This might compromise Gigerenzer's ecological view and the role of heuristics. Polonioli posits that there is not enough evidence to support this view: “[P]eople who score higher at tests of cognitive abilities achieve better life outcomes because they do not reason heuristically remains at a hand-waving level and is not empirically well supported” (Polonioli, Reference Polonioli2015, p. 56).

Gigerenzer's ecological view, as mentioned above, claims that we cannot measure rationality and its heuristics and biases against the axioms of logic, but instead we must measure them against the environment in which they are deployed. Their success rests on that relationship with the environment. Following the rules of logical rationality could even have negative consequences in real-life situations. What adaptive rationality theorists have omitted, according to Polonioli, is that biases are not homogeneous. The families of biases present themselves in different ways: “What [adaptive rationality] theorists and other scholars have failed to notice or acknowledge is that biases are rather heterogeneous kinds of phenomena” (Polonioli, Reference Polonioli2016b, p. 794).

Nudges and education

Gigerenzer's rejection of nudges and defense of education is sustained by the research he conducts, in which he and his Adaptive Behavior and Cognition (ABC) group question the validity of the arguments provided by the heuristic and biases research program. Some of their claims are based on the possible miscommunication between the experimenters and the subjects when referring to probability, and they argue that some may interpret this as plausibility instead. The criticism centers on the examples of the conjunction fallacy provided by Tversky and Kahneman in some of their experiments (Tversky & Kahneman, Reference Tversky and Kahneman1983) that showed that people cannot work with probabilities very well. For example, people consider the probability of ‘Linda is a bank teller and an active feminist’ higher than ‘Linda is a bank teller’. This is clearly incorrect. Gigerenzer claims that this is due to a miscommunication between the experimenter and the subject, since the subject may be considering not the probability, but the plausibility of such an event (Gigerenzer, Reference Gigerenzer2007). In those experiments, the information is presented without a context. Once content, context and representation are provided, the subjects acquire a different perspective of the event (Hertwig et al., Reference Hertwig1999, p. 276). Subjects answering questions about the probability of Linda are inferring a non-mathematical meaning for the concept ‘probability’ (Hertwig et al., Reference Hertwig1999, p. 293). This is what Polonioli (Reference Polonioli2012) called the ‘interpretation argument’.

The work of Gigerenzer also focuses on the distinction between probabilities and frequencies, trying to make those cognitive illusions disappear: “[W]hen information was presented in frequency formats, statistically naive participants derived up to 50% of all inferences by Bayesian algorithms” (Gigerenzer & Hoffrage, Reference Gigerenzer and Hoffrage1995, p. 684). Gigerenzer's goal, therefore, is to educate people rather than to nudge them, by providing them with an understanding of how to cope with risk and uncertainty. According to this approach, single probability events should be reformulated by using frequency statements; natural frequencies should be used instead of conditional probabilities; and the use of relative risks should be complemented with absolute risks. In a Kantian philosophical approach, Gigerenzer (Reference Gigerenzer2014, p. 15) links knowledge and freedom, arguing that risk-savvy citizens must be the pillars of a free society. He believes that a type of ‘risk school’ would be needed in order to increase risk literacy. Allowing governments to nudge instead of educating would undermine efforts to educate and empower citizens. Nudges will facilitate the rise of citizens who are not going to think on their own, instead waiting to be nudged into doing the right thing (Gigerenzer, Reference Gigerenzer2014, p. 245).

Polonioli (Reference Polonioli2012), in reference to the rationality debate between the pessimistic and the optimistic approaches, concludes that the arguments provided by Gigerenzer are not robust enough and that the momentum gained by field experiments provide stronger arguments to maintain Tversky and Kahneman's view. This reinforces the thesis of some authors (Samuels et al., Reference Samuels, Stich, Bishop and Elio2002) who maintain that the dispute is only apparent and that if we attend to the essences of both perspectives, the similarities are more relevant than the differences.

The epicenter of Gigerenzer's criticism of libertarian paternalism is found in the idea that the use of heuristics leads to biases or mistakes. An example is cited by Thaler and Sunstein (Reference Thaler and Sunstein2009) when referring to representativeness heuristics. They argue that the use of such a heuristic process leads people to confuse random fluctuations with causal patterns. Every heuristic has a bias associated with it. According to Gigerenzer, libertarian paternalists believe that the use of heuristics leads to systematic errors in our rationality. But a deeper look at the texts by libertarian paternalists provides enough evidence to maintain that they do not think that radically; that is, heuristics usually work well for nudge theorists. According to Sunstein (Reference Sunstein2014, p. 33), these shortcuts generally work well, but sometimes they misfire. The differences between Gigerenzer and libertarian paternalists are not as deep as Gigerenzer himself thinks. In a very clarifying statement, Sunstein claims that “heuristics tend to have ‘ecological rationality’, in the sense that they make sense in the settings where they are usually applied” (Sunstein, Reference Sunstein2014, p. 33). Gigerenzer would subscribe to this notion word for word. The differences stay at a policy-making level, since both perspectives would advocate for two different course of actions that respond to two philosophical views of similar events.

For Gigerenzer, nudging people is a way to treat them like children. While nudging does not present any alternatives to its proposals, Gigerenzer defends an educational approach, providing what he calls a sustainable solution. We are already nudged by corporations and governments, and it would be very difficult to compete with those nudges. According to Gigerenzer (Reference Gigerenzer2014, p. 256), the solution has to be different: his proposal is based on the investment in risk-savvy citizens “that can make informed decisions themselves” (Gigerenzer, Reference Gigerenzer2015, p. 380). Between both positions we find what Ralph Hertwig (Reference Hertwig and Grüne-Yanoff2017) proposes with boosts. Although he does not think that boosts are better than nudges, he does claim that boosts are a tool that may be useful in certain scenarios, since they target the competences of the individual, increasing his or her agency: “The objective of boosts is to improve people's competence to make their own choices; the focus is on interventions that make it easier for people to exercise their own agency” (Hertwig, Reference Hertwig2017, p. 144).

The problem with nudges, according to Gigerenzer, is not only that we lack evidence of their efficiency, but they may also pose a conflict of interests for the choice architect, since he or she is the one manipulating the environment that affects the outcomes of our decisions. Gigerenzer argues against nudges per se, against their justification and against the libertarian paternalism program and its pessimistic attempt to show the hopelessness of educating people. That a benevolent choice architect is needed to engineer a nudge to steer behavior is a complicated and idealistic approach.

Instead, Gigerenzer proposes that we should be teaching about risk and uncertainty while educating experts on how to communicate about these topics (Gigerenzer, Reference Gigerenzer2014, p. 14). We also need to teach financial literacy to ensure that people are not exploited (Gigerenzer, Reference Gigerenzer2014, p. 256), as well as health statistics (Gigerenzer, Reference Gigerenzer2014, p. 177), among many other things. It is important to highlight that Gigerenzer, like Hertwig, is not against a little bit of nudging, but as a general policy, he claims that “coercing and nudging people like a herd of sheep instead of making them competent is not a promising vision for democracy” (Gigerenzer, Reference Gigerenzer2014, p. 261). While libertarian paternalism claims that it is very difficult to educate individuals, Gigerenzer posits something different: “the claim that we are hardly educable lacks evidence and forecloses the true alternative to nudging: teaching people to become risk savvy” (Gigerenzer, Reference Gigerenzer2015, p. 361).

Nudges rest on the assumptions that we are not easily educable and that the choice architect who manipulates the environment is a benevolent agent with accurate knowledge about the future outcomes of people's decisions. Not all agents are benevolent; some nudges are evil (Thaler, Reference Thaler2018) and only work for the benefit of the institution or corporation that sets them. In behavioral marketing, we have examples that do not benefit the individual (small print to hide fees is a clear example of an evil nudge, ‘sludge’ as Thaler (Reference Thaler2018) called them). Gigerenzer (Reference Gigerenzer2015) claims that there is not enough evidence regarding the first assumption and that choice architects sometimes are not free of self-interest and/or ignorance (e.g., sometimes they do not understand scientific evidence).

A note about autonomy

Autonomy is the control that agents have over their evaluations and choices. “In valuing the ‘autonomy’ of a person, it is not adequate to be concerned only with whether she receives what she would choose if she had the opportunity to choose; it is also important that she actually gets to choose herself” (Sen, Reference Sen1997, p. 753).

The issue of autonomy in choice theory has been discussed widely by Amartya Sen and other authors. The main benefit of nudges rests on the idea that they are liberty preserving (Thaler & Sunstein, Reference Thaler and Sunstein2009, p. 5): since nudges can be resisted, they do not pose any limitations to our freedom. But others see that nudges are not as libertarian as posited. Nudges may be liberty preserving when considering the amount of choices agents have, but they can be questioned in regard to the way they may affect the autonomy of individuals by shaping those choices (Hausman & Welch, Reference Hausman and Welch2010).

The debate is centered on the type of influence nudges may have in our freedom and/or autonomy as previously defined. Could the manipulation of choice may be an imposition on our autonomy since we lack some of the control over our decisions? When a choice architect manipulates an environment, this is because he or she envisions some type of outcome from the agents. In that sense, nudges are paternalistic: they influence our behavior by organizing the choice environment.

Their perspective rests on the assumption that agents are not able to choose in their own interests all of the time. In order to help them to do so, nudges are used as an efficient, liberty-preserving method. Since governments and institution (choice architects) must start somewhere in organizing the environment, nudges are unavoidable (Sunstein & Thaler, Reference Sunstein and Thaler2003, p. 1165).

Other authors (Barton & Grüne-Yanoff, Reference Barton and Grüne-Yanoff2015) have dismissed that claim, arguing that there are scenarios where neutrality can be achieved in the choice environment with education, training or better design.

Libertarian paternalists’ argument rests on the unavoidable nature of manipulation of the choice environment, instead providing an instrumental approach to the arrangement of choices. They manipulate it to improve the outcome of agents’ decisions as judged by themselves. But this is also problematic, according to some authors, as our act of choice cannot be considered as a statement subject to consistency (Sen, Reference Sen1993). When nudges are introduced, it is assumed that we are being inconsistent with our welfare or with our previous or future selves.

Nudges are efficient because those who implement them understand those limitations and the role of the environment in our decisions. Accordingly, this type of paternalism cannot be avoided, since the environment in which agents decide will always influence their decision. Governments, agencies, institutions, cafeterias, etc., must always present their information and choices in a specific way. Doing so without limiting the choices and with the intention of increasing the chances that agents will act in a way that may be beneficial for them is at the heart of nudges. This impossibility of avoiding paternalism becomes clearer when those institutions have to present some type of default option or rules. In some scenarios, it is even more beneficial for the individual and even more liberty preserving when the choice of not having to choose (a default) is presented.

Nudge theory rejects any type of paternalism that is not libertarian. This rejection rests on the idea that freedom of choice should be a part of the decision-making process. Preserving choices is an essential part of it. But sometimes having more choices does not mean a greater degree of freedom (Sen, Reference Sen2002, p. 606). The objections are based on the argument that if certain choices are objectively bad for agents, why have them? The answer provided by libertarian paternalists rests on the argument that freedom of choice has intrinsic value, which it is better to have.

The other argument against nudges considers them as radically affecting our autonomy if they are not based on rational influence: “Their freedom, in the sense of what alternatives can be chosen, is virtually unaffected, but when this ‘pushing’ does not take the form of rational persuasion, their autonomy – the extent to which they have control over their own evaluations and deliberation – is diminished” (Hausman & Welch, Reference Hausman and Welch2010, p. 128). The argument regarding autonomy essentially posits that nudges represent a challenge to our autonomy because they affect our behavior through bypassing our reason. Some nudges take advantage of non-rational elements in our decision-making processes (Wilkinson, Reference Wilkinson2012), so in order to preserve autonomy, only rational influence should be accepted. Against that position, some (Grüne-Yanoff & Hertwig, Reference Grüne-Yanoff and Hertwig2016) argue that nudges may even allow individuals to fully express their autonomy, since they seek to affect people's behavior in order to make them better off as judged by themselves: “[M]uch nudging is also required on ethical grounds, in part because some nudges actually promote autonomy, in part because some nudges enable people to devote their limited time and attention to their most important concerns” (Sunstein, Reference Sunstein2015a, p. 413).

Nudges per se do not respect autonomy, nor do they not represent a challenge to it – it all depends on the type of nudges and how they are applied. In order to have autonomy, informed choices are needed, and there are nudges that respect and enhance that type of information. This is where the distinctions between boosts and nudges become blurred. It looks like there is an intersection of the two, where the same interventions can be considered nudges or boosts depending on who is proposing them. An educational nudge resembles the characteristics of a boost, or even part of what is proposed by ecological rationality. Nudges are not here to remove autonomy. According to Sunstein (Reference Sunstein2015a, p. 438), if nudges improve biased decisions, they help to promote autonomy.

One of the most important aspects governments and institutions must take into consideration while trying to respect the autonomy of individuals is transparency. These institutions should try to inform those they are trying to nudge that they are in fact seeking to alter their behavior. Libertarian paternalists’ policies have to be open and clear about their intentions: “[P]ublicity is important. One important way to protect against abuse and to respect autonomy is to make sure that the government actually inform people of efforts to shape their choices” (Hausman & Welch, Reference Hausman and Welch2010, p. 135).

The fact of the matter is that there are transparent nudges and non-transparent nudges. A transparent nudge would give information about its intentions and how the intervention would be implemented. A non-transparent nudge hides all of that information from the individual subject to the nudge (Hansen & Jespersen, Reference Hansen and Jespersen2013).

The issue regarding transparency is crucial to understand in relation to the ways in which individuals’ autonomy can be affected by nudges. Most authors claim that transparency should be a requirement for nudges. Individuals should be aware of any type of intervention that is destined to modify the choice environment of their behavior, and through that, their actual behavioral outcomes. According to Luc Bovens (Reference Bovens, Grüne-Yanoff and Hansson2009), although transparency may affect the effectiveness of nudges, this does not mean that nudges are not manipulative: “It should be noticed right away that Bovens’ claim that epistemic transparency is in conflict with the efficacy of nudging is not a direct claim as to nudging being psychological manipulation” (Hansen & Jespersen, Reference Hansen and Jespersen2013, p. 19). On the other hand, there are authors who challenge such a perspective, arguing that transparency may not affect nudges at all: “The empirical part is that nudges often or always work best, or only work at all, when they are covert – i.e. when nudgees do not know they are being nudged. Claims to this effect are typically unsubstantiated” (Grill, Reference Grill2014, p. 155).

According to Till Grüne-Yanoff (Reference Grüne-Yanoff2012), nudges are non-transparent manipulations that limit people's degree of liberty based on non-transparent, supposed biases. Martin Lodge and Kai Wegrich (Reference Lodge and Wegrich2016) also claim that nudges are, in essence, non-transparent tools. But according to Sunstein, some of these objections are not accurate. There are nudges that are clearly transparent, like default rules and disclosure policies (Sunstein, Reference Sunstein2013).

Choice architects should always aim for transparency in order to preserve autonomy and to eliminate any ethical complaints or objections: “Some nudges are objectionable because the choice architect has illicit ends. When the ends are legitimate, and when nudges are fully transparent and subject to public scrutiny, a convincing ethical objection is less likely to be available” (Sunstein, Reference Sunstein2015b, p. 1). Sunstein asserts that transparency should be a safeguard for nudges. If nudges are liberty preserving, this is due to the fact that individuals can choose not to follow them. In order to respect this and to allow nudges to be non-irresistible, they have to be transparent (Barton & Grüne-Yanoff, Reference Barton and Grüne-Yanoff2015).

Nudges and values

Instrumental rationality is the default theory of rationality (Nozick, Reference Nozick1993, p. 133). Although there are different frames that can be used to talk about rationality, the means–end schema is the most widely used in the academic literature. Against that narrow view, Raymond Boudon (Reference Boudon1998a) claimed that we are also rational because we have reasons to believe in a specific thing or to act in a specific manner, sometimes because we have strong reasons to believe it is the right thing to do, regardless of the consequences. Considering rationality merely within an instrumental frame, Boudon posits, will impede our ability to explain certain aspects of our rational decisions that cannot be categorized within a means–end schema. People sometimes behave not according to the consequences of their behavior, but according to the principles or values they hold:

His definition is very similar to the one provided by Amartya Sen: “Rationality of choice, in this view, is primarily a matter of basing our choices – explicitly or by implication – on reasoning that we can reflectively sustain if we subject them to critical scrutiny” (Sen, Reference Sen2009, p. 180).

Some consider that Boudon introduced a notion of rationality that was parallel to the one drawn by Gigerenzer, and that very similar criticisms can be considered: “Whilst, to the best of my knowledge, Boudon and Gigerenzer do not cite each other, their conceptions of human rationality are in fact strikingly similar” (Manzo, Reference Manzo2014, p. 534). According to Boudon, Tversky and Kahneman's description of biases is close to a ‘black box’ (Boudon, Reference Boudon, Hedström and Swedberg1998b), or as Gigerenzer sees them, as mere labels that are too vague to be robust explanations of our decision-making processes (Gigerenzer, Reference Gigerenzer1996, pp. 593–594).

I do consider that Gigerenzer's approach also remains at an instrumental level. His program, while providing normative insights into how to behave in order to be more efficient, fails to explain how rationality works from an evolutionary perspective. According to Gianluca Manzo (Reference Manzo2014), Boudon's ‘ordinary rationality’ is very similar to Gigerenzer's ecological approach. In my view, Manzo sees in Gigerenzer something that is barely there. Gigerenzer's ecological approach is basically an instrumental explanation of our decisions. The role of values is absent in his theories.

Matteo Colombo, although taking a different perspective, also considers that the ecological approach is instrumental and attends to goals. He contraposes it to the constructivist view of rationality, a distinction used by Vernon Smith (Reference Smith2008). According to Colombo, this constructivist view uses norms (axiomatic) as a way to measure its success. The ecological approach, Colombo proceeds, focuses on its connection with the environment, and it attends to goals. Using Dennett's distinction between personal (intentions, beliefs, desires) and subpersonal (networks in the brain) levels of explanation (Dennett, Reference Dennett1969), Colombo claims that ecological rationality is associated with the subpersonal level of explanation, since this connects behavior to the environment (de Pinedo-Garcia & Noble, Reference de Pinedo-Garcia and Noble2008), while the constructivist view is personal. Although both Smith and Gigerenzer use the term ‘ecological’ and they both cite Herbert Simon as the foundation of their theories, they refer to two different models (Dekker & Remic, Reference Dekker and Remic2018). It is important to clarify that although we are dealing with two distinctions (personal–subpersonal/axiomatic–ecological), these pairs are independent of each other. The pair of concepts ‘personal and subpersonal rationality’ does not imply the other pair ‘axiomatic and ecological rationality’.

Our behavior, Boudon explains, is not only instrumental; we also show cognitive and axiological reasons for our behavior. We possess an expressive (axiological) rationality that gives sense to the world – that makes the world intelligible to us (Álvarez, Reference Álvarez1992, p. 82). This type of rational theory might shed some light, for example, on the reasons why we vote when it is not instrumentally rational. Boudon eliminates the ‘black boxes’ that other theories assume when they lack an explanation for what people do. Those black boxes are an example of Tversky and Kahneman's heuristics and biases (verbal labels, according to Gigerenzer). Boudon, critical of the heuristics and biases research program, maintains that we cannot interpret those biases as causes, and nor can we characterize our behavior as irrational when it is biased in Tversky and Kahneman's view. Boudon focuses on the reasons as to why individuals have erroneous beliefs, but this is not irrationality. He defends a rationalistic view of error instead of the irrational (pessimistic) one (Demeulenaere, Reference Demeulenaere2014).

We have beliefs, according to Boudon, and those beliefs are sometimes the reasons behind our decisions. We have ‘personal’ reasons to do what we do. In a way, this notion resembles Dennett's distinction. When we understand the reasons underlying our behavior by attending to beliefs, this is a way to introduce a horizontal explanation (Drayson, Reference Drayson2012). According to this approach, while the personal level of explanation is horizontal, the subpersonal level is vertical. It is this personal view that is similar to Boudon's reasons. Boudon refers to beliefs, for example, as a way to understand the reasons why people act in a certain way:

As mentioned above, Colombo (Reference Colombo, Sprevak and Colombo2018) sees the personal level of explanation as part of the constructivist view of rationality, subject to norms and axioms. The subpersonal level would be closer to an ecological view, following Vernon Smith's distinction (Reference Smith2008), and it would attend to goals.

Reducing all of our decisions to an instrumental perspective produces the appearance of those black boxes Boudon criticizes. Behaving in a purely egotistic and instrumental way would make us the type of rational fool that was described by Sen (Reference Sen1977).

Since we have reasons to do what we do, are nudges a way to manipulate those reasons and to steer our behavior to a mere instrument that would satisfy our goals? Are nudges a way to remove our axiological rationality or are they an instrument to preserve it? These would be difficult questions to answer because of their breadth and because there are many different types of nudge: default rules, information nudges, etc. In my opinion, there is still research to do on the distinction between nudges, boosts and the ecologically rational interventions that aim to educate. The answer provided by libertarian paternalists is that it is impossible to remove nudges since the choice architect will always have to decide upon an environment in which individuals must decide: “[M]anipulation is unavoidable; no matter how options are presented, they frame our choices and behavior” (Roeser, Reference Roeser2012, p. 1037).

Let us consider a hypothetical scenario: recycling. Let us suppose that an individual does not recycle because there is a cost (effort) in separating all of the materials, organizing them and having an extra container to store them in her small kitchen. On the one hand, she has reasons for doing what she does. On the other hand, she also thinks that environmental issues are important and we should limit the damage humans cause to the environment. She has instrumental reasons for not recycling and values that push her to do so. It is obvious that a society would benefit if nudges were used to increase the amount of materials that are recycled, especially if the individuals involved agree that recycling should be done. According to this perspective, a nudge that might facilitate recycling should be desired, since it would reduce the cost of individuals trying to separate materials for recycling.

Some nudges do change our values. Companies that manufacture high-fructose corn syrup have tried to change its name to ‘corn sugar’ in order to manipulate the perspectives that individuals have about their product and its chemistry. ‘Corn sugar’ sounds like a natural product, while ‘high-fructose corn syrup’ sounds like a chemical product. The Food and Drug Administration sent a letter requesting companies to cease their use of the term ‘corn sugar’ in reference to this product. Changing its name could have been considered a nudge that would have altered the values of consumers in regard to this product.

The case is similar with labels referring to trash as ‘landfill’. This label might cause people to change their cognitive rationality since they would be receiving information that also impacts their emotions regarding where their trash will go.

I claim that the issue rests on the relationship between the different aspects of rationality. Sometimes, rationally explaining how things work in an attempt to alter people's cognitive rationality is not enough or does not work at all. This is the premise of nudge theory: education sometimes does not work:

Considering issues related to vaccination, we may find people with reasons to believe that vaccination is dangerous. Following Boudon's argument, if they do not get vaccinated, they have cognitive reasons for this course of action. Their behavior is not irrational; their actions are based on ‘personal’ beliefs. The argument, within the rationality debate, is whether they should be educated, nudged or forced to receive those vaccines. In a sense, while all alternatives would try to change people's actions, not all try to change their cognitive rationality. Rejecting the use of vaccines, from an instrumental perspective, might be viewed as irrational, but not from Boudon's perspective. These people may be ignorant, or maybe they are showing some type of risk aversion that does not allow them to see the benefits of vaccines.

Understanding the reasons why people behave in a specific way, or at least accepting the idea that people may have reasons to do what they do, can shed some light on what type of policy-making technique we should use to change their behavior towards accepting something as objectively beneficial as vaccines without limiting their autonomy. Some have argued that bypassing people's autonomy should be accepted with regard to vaccines:

Even if specific nudges affect the autonomy of individuals, this perspective would defend their use due to the objective benefit of vaccines. That is the argument made by Emily Oster and Geoffrey Kocks (Reference Oster and Kocks2018) in their New York Times article about vaccination and measles. Data show that more people got vaccinated and the risk of measles was lowered in California after an outbreak in 2014 when laws were stricter. But what is controversial is their following idea: “From a policy standpoint, these findings offer a ray of hope for vaccine proponents. Maybe changing minds isn't so important. People may not have altered their attitudes about vaccination, but the fact is that these laws actually changed behavior” (Oster & Kocks, Reference Oster and Kocks2018).

The use of nudges responds to the impossibility of educating people about certain behaviors and the idea that a choice is better than a mandate. A view that includes axiological components might try to help us understand the reasons why some people reject the use of vaccines without relying on the black box of irrational behavior or the dismissal of why people do what they do. Accordingly, from a public policy perspective, the goal should not only be for people to do what is right (instrumental), but also to understand and value what is right (cognitive and axiological rationality). Having a monistic view of rationality that rests just on instrumental principles affects the type of policies to implement.

Nudges should not be rejected, but the theory of rationality on which they rest should include an axiological component. This broader view of our rationality includes not only an instrumental means–end schema, but also an axiological angle, in line with what was proposed by Javier Echeverría and J. Francisco Álvarez (Reference Echeverría, Álvarez and Agazzi2008). Both elements of rationality act in a “single communicative situation” (Echeverría & Álvarez, Reference Echeverría, Álvarez and Agazzi2008, p. 175). Since the instrumental and axiological approaches constitute the fabric of our rational processes, neglecting one or the other would generate problems when trying to account for the way people decide. A narrow version of rationality based merely on instrumental principles would not be able to explain certain phenomena within our decisions. A broader notion that incorporates axiological principles would provide explanations about behavior that otherwise might seem irrational, generating the ‘black boxes’ Boudon was criticizing.

I claim that the libertarian paternalist perspective that advocates for the use of nudges in policy-making, although effective, rests on a narrow version of rationality that cannot account for specific behaviors, which can only be explained by attending to axiological principles. A complete policy-making approach should consider not only the outcome, but also the process itself. Sometimes, when individuals act in a specific way, they have reasons for doing so. It is important that any type of policy-making approach takes people's expressive rationality into consideration when trying to influence their behavior. Nudges, when needed, should tackle both aspects of rationality, not only the instrumental one. If people's cognitive axiological rationality (in Boudon's terms) is based on false predicaments, policy-makers should try to change not only the outcome of people's behavior, but also their axiological rationality.

Nudges are an efficient tool and I do not think they should be rejected. There are educational nudges that help people understand more about their choices. The obnoxious beep that our cars produce when we are not wearing a seatbelt is an example of a nudge that is efficient and serves as a reminder of what to do. Supplementing that with an educational perspective that might help us to understand how seatbelts are beneficial would be a comprehensive approach, since it would consider both aspects of our rational behavior while assuming a realistic view: people are boundedly rational.

Since individuals show two versions of rationality (instrumental and axiological), it is important that both elements are considered by public policy-makers and governing bodies when trying the benefit their citizens. To this end, nudges should have an educative component.

Take, for example, the use of seatbelts. The Centers for Disease Control and Prevention (CDC) of the government of the USA has a campaign to increase the use of seatbelts among drivers. They use the appropriate educational tools in regard to risk and accidents to influence our cognitive axiological rationality. But nudges are also able to alter our instrumental rationality:

• • Cars use beeping sounds to remind us about the use of seatbelts.

• • Advertising campaigns use catchy phrases such as ‘Click it or ticket’.

• • On the roads, there are reminders of the use of seatbelts that are clear and precise: ‘Buckle up, it is the law’.

According to the CDC (n.d.), in 1981, only 11% of drivers used seatbelts in the USA; by 2010, that number had increased to 85% of all drivers. Drivers today are not only changing the outcomes of their decisions; they also understand, from an axiological cognitive perspective, the benefits of wearing a seatbelt in a car.

Regulations, public policies and nudges regarding safety have always had the possibility of backfiring, and sometimes they may cause an opposite effect to the one desired. Regulations proposed in the auto industry regarding seatbelts in the 1960s did not cause any decrease in death rates. Sam Peltzman's (Reference Peltzman1975) research proved that safety regulations provided safety benefits, but they also increased the risks taken by drivers, therefore offsetting the effects of the regulation (Peltzman, Reference Peltzman1975, p. 717). The term used to define this type of case is ‘risk compensation’. It is something that can be seen in the use of helmets while riding bicycles (Adams & Hillman, Reference Adams and Hillman2001).

This introduces a factor to consider when implementing policies (educational, regulatory or in the form of nudges), since there is always the possibility of an offset through a change of behavior of the people targeted by the policies, either by risk compensation or by other unforeseen effects. For example, nudges used to decrease electricity use by comparing usage with neighbors may be effective for those households that are politically liberal, but it may not work for those on the other side of the political spectrum (Costa & Kahn, Reference Costa and Kahn2013). These ‘rebound effects’ happen with different policies. Dual-flushing toilets, for example, may increase water consumption instead of reducing it, since in some cases they may be flushed more times.

The rebound effect may be caused by bad design of the nudge. It is also possible that their effects may be ephemeral. But that policies may not provide the desired consequences is something that policy-makers must consider. Providing educational nudges or boosting (Grüne-Yanoff & Hertwig, Reference Grüne-Yanoff and Hertwig2016; Hertwig & Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017) competences might eliminate the complex ethical issues that might appear if regulation or nudging backfires.

If governments as policy-makers attend only to the instrumental aspect of rationality, they would be neglecting part of their function in a democratic society, which is to educate people in order for them to be autonomous in their decisions. This can only be done by understanding the rationality of their citizens in a comprehensive manner: not only from an instrumental view, but also from a bounded axiological rationality perspective. We are boundedly rational individuals who also exhibit axiological rationality (Echeverría & Álvarez, Reference Echeverría, Álvarez and Agazzi2008, p. 173). It is because we exhibit axiological rationality that educative nudges should complement the instrumental ones.

But there cannot be a liberty-preserving nudge if individuals lack the understanding of and do not share the values (axiological) upon which that decision is based. Values here are an important filter; as Echeverría and Álvarez claim, they “orient our perception … The subject selects what he desires and rejects in the environment, the subject valuates the information” (Reference Echeverría, Álvarez and Agazzi2008, p. 179).

In addition to the above-mentioned claims, not all nudges are for our benefit. In this paper, I have only referred to nudges that try to make life better for individuals, neglecting nudges that may try to do the opposite. Since our world has evil nudges – Thaler's (Reference Thaler2018) ‘sludges’ – and since it is desired to have citizens who are able to escape the effects of sludges, we can only hope for informed and educated (competence-related) citizens capable of recognizing these sludges and escape their effects. Who is providing this information? Who is educating citizens to prevent them from falling into sludges? My approach, stemming from claims that nudges can only be used in tandem with educational interventions, can provide a solution to the problem. Public policy-makers must educate individuals in order to prevent the negative effects of sludges.