2. Situational analysis: an attempt at clarification

The main purpose of this section is to offer an interpretation of SA according to which the latter consists of a methodology whereby social scientists may explain both ‘historical’ and ‘typical’ social events by identifying the deviation of individual actors’ actual behaviour from the behaviour they would exhibit if their view of the P-S coincided with the ‘wider’ or ‘more complete’ view allegedly possessed by the former. We argue that this interpretation implies that the different versions of Popper’s RP identified by his commentators are not only compatible with each other but are also complementary.

3. Do social scientists possess a ‘more complete’ view of the P-S than actors do?

According to the interpretation of SA we have advocated above, the explanation of a social event consists mainly of identifying the deviation of actors’ ‘partially wrong’ or ‘incomplete’ view of the P-S from the social scientists’ alleged ‘wider’ view of it. It turns out that the notion that the latter exhibit a ‘wider’ or ‘more complete’ view of the P-S than actors do is a central premise of SA. As the quotation below reveals, Popper seems to assume that social scientists can indeed reconstruct a ‘more complete’ picture of the P-S:

Now, we have argued above that the notion that historians normally possess a ‘wider’ or ‘more complete’ view of the P-S than actors did at the time the ‘historical’ event in question occurred is a sensible one since the former can observe those factors that were unpredictable when the event in question took place as well as the ‘objective’ mistakes actors made. However, is this premise sensible in the case of the ‘theoretical’ social sciences? Unlike the ‘historical’ social sciences, which deal with unique events, or anthropological studies where scientists study people’s behaviour from the vantage point of a more advanced civilization, the ‘theoretical’ social sciences deal with typical social events. Notably, the latter may recur over time which opens up the possibility that actors ‘learn’ from past mistakes which may, in its turn, jeopardize the social scientist’s ‘privileged’ position, i.e. the alleged possession of a ‘wider’ or ‘more complete’ view of the P-S. Thus, if social scientists are to maintain such (alleged) ‘privileged’ position vis-à-vis the actors, the latter’s ‘learning’ process must be such that it does not enable them to close this ‘knowledge gap’.

Despite the critical importance of this premise, neither Popper nor any of his commentators discuss it. However, we think this central premise accords with Popper’s ‘evolutionary’ theory of knowledge, the cornerstone of his philosophy. To be sure, the presumption that social scientists normally possess a ‘wider’ or ‘more complete’ view of the P-S than actors do accords with Popper’s notion that the objectivity of (natural and social) scientists ultimately stems from the theories they put forward being subject to ‘critical discussion’. In particular, Popper claims that the ‘objectivity’ of science does not really stem from the ‘objectivity of the scientist’, however understood, and that it is ‘erroneous to believe that the attitude of the natural scientist is more objective than that of the social scientist’ (Popper Reference Popper1992: 72). He insists that the ‘objectivity’ of science ‘is not a matter for the individual scientist but rather the social result of mutual criticism’ (Popper Reference Popper1992: 72). Referring to the social sciences, he claims that SA is a method that enables social scientists to subject their theories (i.e. situational models) to ‘critical discussion’:

Therefore, if we follow Popper’s reasoning, the ‘objectivity of science’ must be interpreted in relative rather than absolute terms, namely, the more exposed to ‘critical discussion’ a theory becomes, the more ‘objective’ it is. This implies, in turn, that the ‘objectivity’ of scientific theories is not an ‘immanent’ but a ‘circumstantial’ property: as soon as a theory stops being subject to critical discussion its ‘degree of objectivity’ starts to decline. By the same token, the more exposed to critical discussion is a social scientist’s view of the P-S facing actors as enshrined in her situational model, the more ‘objective’ the latter becomes. To the extent social scientists’ views are normally more exposed to critical discussion – which is what being part of a scientific community is essentially about – than ordinary people’s (i.e. the actors of ‘situational models’), it is sensible to assume the former normally exhibit a ‘wider’ or ‘more complete’ view of the P-S than individual actors do. The upshot is that criticism of theories (as embedded in situational models) by other members of the scientific community is not solely what, in Popper’s view, distinguishes scientific knowledge from other less sophisticated forms of knowledge; it is also a pre-condition of his methodological proposal for the social sciences insofar as it is what underpins his central premise that social scientists possess a ‘wider’ or ‘more complete’ view of the P-S than actors do.

4. Is SA a generalization of the neoclassical theory of individual behaviour?

The basic model of choice implicit in neoclassical microeconomic theory takes an economic agents’ choice or action to be the result of her beliefs and her preferences. Choice is said to be ‘rational’ when it is determined by rational beliefs and preferences. In turn, agents are taken to have rational preferences if the latter are ‘complete’ and ‘transitive’ and to choose rationally if: (i) their preferences are rational and (ii) there is no feasible option that they prefer to the chosen one. ‘Completeness’ requires simply that economic agents can compare all feasible options whereas ‘transitivity’ implies that they exhibit a consistent ranking of preferences in the sense that if A is preferred to B, and B is preferred to C, A must be preferred to C. Lastly, the axioms of ‘completeness’ and ‘transitivity’ are usually coupled to the ‘continuity’ assumption which is necessary for the existence of a well-behaved utility function. Now, although it may not be always explicit, neoclassical theory presupposes that economic agents are normally ‘rational’. Robbins is explicit on this issue in his classic essay on the ‘nature and significance of economic science’ where he makes clear that ‘the main postulate of the theory of value is the fact that individuals can arrange their preferences in an order, and in fact do so’ (Robbins Reference Robbins1932: 78–79). Therefore, it follows that the neoclassical theory of individual behaviour can be interpreted both as a ‘descriptive’ theory and as a ‘normative’ theory. The ‘descriptive’ interpretation of the theory derives from the fact that, once we have defined rationality, we can claim (as Robbins does) that individuals are ‘rational’ in the sense defined above.Footnote ⁷ In turn, this means that when neoclassical economists say that individuals maximize utility they are only saying that the latter do not rank any feasible option above the option they actually choose. Since utility is a measure of the ‘degree of satisfaction of preferences’ we have that an individual who maximizes utility just does what she most prefers. The ‘normative’ interpretation follows from the fact that defining what rational preference and choice are also implies saying how one ought rationally to prefer and to choose.

Some commentators claim that the neoclassical theory of individual behaviour is a variant of SA (Oakley Reference Oakley1999; Caldwell Reference Caldwell2000; Neves Reference Neves2004). By contrast, several other commentators criticize this view (Matzner and Bhaduri Reference Matzner and Bhaduri1998; Notturno Reference Notturno1998; Gorton Reference Gorton2006). Be that as it may, the fact is that Popper explicitly says in different parts of his works that SA is an attempt to generalize the methodology of neoclassical economic theory. For instance, in his intellectual autobiography he explains that SA is an attempt to ‘generalize the method of economic theory (marginal utility theory) so as to become applicable to the other theoretical sciences’ (Popper Reference Popper2002 [1976]: 135). In The Open Society he boldly claims that SA ‘is, in fact, the method of economic analysis’ (Popper Reference Popper1966 [1943]: 97) and, in an essay titled The Logic of the Social Sciences (Popper Reference Popper1992), he writes that ‘the logical investigation of the methods of economics yields a result which can be applied to all social sciences’. Lastly, in a key essay included in The Myth of the Framework (Popper Reference Popper and Notturno1994 a: 154), he notes that ‘my views on the methodology of the social sciences are the result of my admiration for [neoclassical] economic theory’. However, and notwithstanding Popper’s claims to the contrary, we think that SA is not really a ‘generalization’ of the neoclassical theory of individual behaviour. Neoclassical theorists typically assume that economic agents do possess, at least on average, as much knowledge about the P-S they face as (neoclassical) economists do. It follows that the neoclassical theory of individual behaviour presupposes that actors’ view of the P-S does coincide, on average, with the scientist’s and, hence, that there is no difference between RPI and RPs. An implication of this is that individual rationality can be interpreted both in a ‘descriptive’ and a ‘normative’ sense: the neoclassical theory of individual behaviour presupposes that nobody can actually attain a higher degree of satisfaction of individual preferences than any individual does given the knowledge available in the economy. In turn, this premise implies that government interference in the functioning of a market economy is undesirable unless it successfully addresses the ‘market’ failures associated to deviations from the ‘perfect competition’ scenario. By contrast, we argued above that SA presupposes that social scientists possess a ‘wider’ or ‘more complete’ view of the P-S than individual actors do which, as we will see below, opens up the possibility that government policy aimed at steering individual behaviour in a certain direction may bring about an increase in individual welfare. We conclude that, if the interpretation of SA we have presented above is correct, then it follows that SA is not a generalization of the neoclassical theory of individual behaviour.

5. Behavioural economics and SA

Next, although in the basic neoclassical theory of individual behaviour economic agents exhibit ‘perfect knowledge’, the theory has been extended to situations of risk and uncertainty, that is to say, situations where individual actors exhibit imperfect knowledge about the outcome of their actions. The standard approach to modelling such situations, commonly known as ‘subjective expected utility’ (SEU) theory, is to regard individual actors’ actions as gambles with outcomes in the form of prizes and to assume that the probabilities of the different outcomes consist of subjective ‘degrees of belief’ which may or may not arise from knowledge of relative frequencies generated in the past. As in the case of utility theory, SEU theory relates choice to preference by asserting that economic agents never prefer any feasible option to the one they actually choose. However, and notably, SEU theory is not only a theory of rational preference and choice but also a theory of rational belief. This is so because the axioms of rational choice under uncertainty do require that subjective ‘degrees of belief’ fulfil the axioms of probability calculus (i.e. Bayes’ rule). Notably, this implies that, alike ordinal utility theory, SEU theory can be interpreted both as a ‘descriptive’ and a ‘normative’ theory. By contrast, and unlike ordinal utility theory, SEU theory can also be applied to choices that involve risk and uncertainty. It is this feature that allows scientists to study whether individuals really choose the way SEU theory says they do (and should) and, starting with the work of Allais (Reference Allais1953), Ellsberg (Reference Ellsberg1961) and Kahneman and Tversky (Reference Kahneman and Tversky1979), a voluminous literature reveals that individuals do not normally behave as SEU theory says they do.

We have just referred to the seminal work of social and cognitive scientists who criticize the notion that SEU theory is a ‘descriptive’ account of how individuals behave under conditions of uncertainty. It is not our purpose to provide a summary of the ‘state of the art’ in modern behavioural economics. Historical accounts of its development are provided in Angner and Loewenstein (Reference Angner, Loewenstein and Mäki2012), Bruni and Sugden (Reference Bruni and Sugden2007), Lewin (Reference Lewin1996), and Sent (Reference Sent2004). In addition to the studies alluded to above, a classic survey of research at the crossroads between psychology and economics is in Rabin (Reference Rabin1998) and a collection of papers in behavioural economics is in Camerer et al. (Reference Camerer, Lowenstein and Rabin2004). In what follows, we will restrict our discussion to the research denoted as ‘New Behavioural Economics’ (NBE) whose emergence is closely linked to the studies in Tversky and Kahneman (Reference Tversky and Kahneman1974) and Kahneman and Tversky (Reference Kahneman and Tversky1979). Unlike pioneers of ‘behavioural economics’ like Herbert Simon or George Katona who studied the impact of a rationally bounded individual on the predictions of economic and administrative theory, advocates of NBE normally take SEU theory as the source of their ‘null-hypotheses’ and then try to identify systematic deviations of individual actors’ behaviour from (ideal) behaviour associated with the former. As Sent (Reference Sent2004: 742) explains, this means that NBE situates itself ‘squarely within the mainstream’. In particular, by taking SEU theory and Bayesian probability judgements as ‘benchmark’ models for rational behaviour, NBE aims at enhancing the explanatory power of the mainstream model of choice. Specifically, after identifying systematic deviations of individual behaviour from the pattern of behaviour predicted by the standard model of choice, NBE seeks to show how psychology-based theories can explain the apparent ‘anomalies’ (from the standpoint of SEU theory) of economic agents’ observed individual behaviour. As Camerer and Loewenstein (Reference Camerer, Lowenstein, Camerer, Lowenstein and Rabin2004: 3) conclude, NBE ‘increases the explanatory power of economics by providing it with more realistic psychological foundations’.

Now, there are three broad reasons why individual behaviour may systematically deviate from the neoclassical theory of choice under uncertainty (Mullainathan and Thaler Reference Mullainathan and Thaler2000): (i) bounded rationality, (ii) bounded willpower and (iii) bounded self-interest. First, bounded rationality implies that people normally exhibit limited cognitive capabilities that impair their abilities to solve all type of problems. Second, bounded willpower suggests that people’s choices do not always correspond to their long-run interest. Last, bounded self-interest implies that some individuals are willing to sacrifice their own interest to help others. However, not all deviations from the standard model of choice associated with the three types of boundedness listed above imply that the former needs to be modified. For instance, the finding that people’s dislike for losses outweighs their desire for gains can be easily accommodated within SEU theory. By contrast, the presence of systematic biases in judgement under uncertainty and, above all, the finding that some people have self-control problems or problems evaluating their preferences suggests that a profound revision of the standard model may be necessary. Be that as it may, for the vast majority of modern behavioural economists SEU theory provides a ‘normative’ benchmark rather than a ‘descriptive’ theory of how people behave. Yet, and according to us, this is precisely the methodology that Popper proposes for the ‘theoretical’ social sciences: to build an ideal ‘benchmark’ scenario in order to identify the deviation of actors’ actual behaviour from the former. Specifically, the mistakes of economic agents that modern behavioural economists have documented for decades constitute systematic deviations of economic agents’ actual behaviour from the normative benchmark scenario captured by SEU theory. Ironically, it is Savage (Reference Savage1954: 20), one of the founding fathers of SEU theory, who inadvertently best portrays the interpretation of SA we have suggested above when he makes it clear that the axioms of SEU theory are to be interpreted preferably as basic principles that a rational individual ought to follow rather than as a hypothesis about how an individual actually chooses among a set of possible courses of action.

Next, the notion that the neoclassical model of choice under uncertainty (i.e. SEU theory) represents a normative ‘benchmark’ from which systematic deviations of actors’ actual behaviour are to be identified takes us to the role of Welfare Economics. In traditional ‘Paretian Welfare Economics’ (PWE), the focus of researchers is normally on observed deviations of market economies from a hypothetical ‘benchmark’ scenario associated to ‘perfect competition’. In such a scenario, maximizing behaviour by economic agents ensures that the resulting allocation of resources is Pareto-optimal. To the extent that all market participants are utility (or profit) maximizers (which is a basic premise of the neoclassical theory of individual behaviour), all welfare economists can do is identify ‘market failures’ (i.e. apparent deviations of ‘real-world’ economies from the ‘perfect competition’ scenario). Such deviations represent cases for potential Pareto-improving government intervention. Thus PWE presupposes that it is not possible for governments to increase social welfare by influencing individual behaviour other than by transforming economies in order to make them approach the ‘perfect competition’ scenario because it is assumed that nobody (including policymakers) can do ‘better’ than economic agents themselves do. In other words, PWE presupposes that if policymakers ever happened to possess the same preferences as an economic agent and faced the same P-S the latter does, they would achieve, at most, the same welfare she does. Translated into Popper’s parlance PWE presupposes, as the neoclassical theory of individual behaviour does, that scientists (or policymakers) do not exhibit a ‘wider’ or ‘more complete’ knowledge of the P-S than individual actors do, the only difference between PWE and neoclassical theory being that while the alleged goal of ordinary economic agents is to maximize utility or profits, the alleged goal of policymakers is to maximize ‘social’ welfare, a goal they may attain by ameliorating ‘market failures’.

Unlike PWE and alike SA, ‘Behavioural Welfare Economics’ (BWE) is based on the notion that social scientists (and, hence, policymakers) normally possess a ‘wider’ or ‘more complete’ knowledge of the P-S facing actors than the latter do. In particular, BWE presupposes that scientists know the normative ‘benchmark’ scenario (i.e. or how actors ought to behave) and may also get to know, via experiments or observation, how actors actually behave. By contrast, individual actors can only observe the behaviour of other actors who are located in their vicinity. Arguably, the notion that scientists (and policymakers) possess a ‘wider’ or ‘more complete’ view of the P-S actors face than the latter do lies at the core of the policy proposals put forward by behavioural welfare economists such as ‘libertarian’ paternalism (Thaler and Sunstein Reference Thaler and Sunstein2003), ‘asymmetric’ paternalism (Camerer et al. Reference Camerer, Issacharoff, Loewenstein, O’Donoghue and Rabin2003), and ‘light’ paternalism (Loewenstein and Haisley Reference Loewenstein, Haisley, Caplin and Schotter2008). By building on the research carried out by modern behavioural economists in the last few decades and on the extensive empirical evidence that the behaviour of individual actors normally deviates from behaviour predicted by SEU theory, behavioural welfare economists claim that it is possible to devise (weak) ‘paternalistic’ public policies that may help steering individual behaviour in more self-interested directions. For instance, as Loewenstein and Haisley (Reference Loewenstein, Haisley, Caplin and Schotter2008: 212) explain, ‘paternalistic policies have the goal of benefiting people on an individual basis, premised on the idea that people cannot be relied upon to invariably pursue self-interest’. Thus, while the rationale for government intervention in PWE is to limit economic externalities, the rationale for intervention in BWE is instead to limit internalities, that is, to reduce the costs that people impose on themselves when making decisions. Typical examples of ‘strong’ forms of paternalistic public policies include the prohibition of drugs, public health advertising, health and safety regulations, health warnings on cigarettes and alcoholic drinks, or the social security system. By contrast, BWE implies weaker forms of paternalistic public policies aimed at steering individual behaviour in more beneficial directions while, at the same time, maximizing individual autonomy. A common example of a ‘weak’ form of paternalistic policy is the choice by social planners of a ‘default option’ that is known to affect favourably economic agents’ individual welfare in the long run. As Thaler and Sunstein (Reference Thaler and Sunstein2003: 176) explain ‘in a fully rational world such design choices would have little effect (at least in high-stakes situations) because agents would simply choose the best option for them regardless of the default’ yet some studies show that there is a strong ‘status quo’ bias, namely, that people tend to stick to the existing arrangement regardless of its features. Thus, scientists (and policymakers) may take advantage of their alleged ‘wider’ knowledge of the characteristics of different ‘default options’ to influence economic agents’ behaviour in a direction that yields higher ‘social’ welfare by choosing the optimal (from a social point of view) ‘default’ option while minimizing coercion. Loewenstein and Haisley (Reference Loewenstein, Haisley, Caplin and Schotter2008: 215) nicely capture the research agenda of behavioural welfare economists when they write that ‘the economist/therapist endeavors to counteract cognitive and emotional barriers to the pursuit of genuine self-interest’. However, this is possible only if we assume that the ‘economist/therapist’ possesses a ‘more complete’ view of the P-S than ordinary people do. To conclude, behavioural welfare economists implicitly adopt the methodology we have associated to Popper’s SA: (i) they implicitly assume that they exhibit ‘wider’ or ‘more complete’ knowledge than economic agents do; (ii) they treat SEU theory as an ideal ‘benchmark’ scenario; and (iii) they try to identify systematic deviations of actors’ actual individual behaviour from the former.

6. Theoretical models and SA

In the previous section we have argued that our interpretation of Popper’s SA accords with the methodology implicitly adopted by modern behavioural economists. In this section we focus on the role models play in SA and, particularly, we compare the former to the role ascribed to scientific models in recent scholarly debates.Footnote ⁸ Our main claim is that the account of theoretical models that stems from our interpretation of SA implies a way of acquiring knowledge about the ‘real world’ that, to the best of our knowledge, has gone unnoticed in the recent literature on the philosophy of economics. Specifically, as we have explained above, our interpretation of Popper’s SA implies that situational models may allow us to ‘learn’ about the real world by using them as ideal ‘benchmark’ scenarios from which the systematic deviation of actors’ actual individual behaviour can be identified. In other words, according to our interpretation of SA models are heuristic devices that depict individual actors’ behaviour under ‘ideal’ circumstances by which we mean that, when building a situational model, social scientists assume that individual actors possess (at least, on average) as much knowledge about the P-S facing them as the former do. As we argue below, the former implies that situational models are intended to be neither an oversimplified representation of the ‘real world’ nor an ‘artificial world’ that bears some similarity to the former.

In what respect does the account of models associated to our interpretation of SA differ from the accounts we find in recent debates about the nature and role of scientific models? For our purposes, it is convenient to focus on an essay by Donato Rodríguez and Zamora-Bonilla (Reference Donato Rodríguez and Zamora-Bonilla2009) in which the authors claim that the seemingly conflicting accounts of models held by scholars associated to two different camps (i.e. by so-called ‘isolationists’ and ‘constructivists’) may be encompassed within an overarching account of models they denote as the ‘inferential approach’. More specifically, they argue that most participants in current controversies share the view that ‘what allows the making of an inductive inference from the model to the real [or empirical] system is only the belief that the model rightly includes aspects that, in the real world, produce the phenomenon our inferences are about’ (Donato Rodríguez and Zamora-Bonilla Reference Donato Rodríguez and Zamora-Bonilla2009: 115). In turn, this leads them to suggest that theoretical models are better understood as a type of artefact that serve the purpose of surrogate reasoning and whose main use is ‘in helping us to draw inferences from the system they represent’ (Donato Rodríguez and Zamora-Bonilla Reference Donato Rodríguez and Zamora-Bonilla2009: 102).Footnote ⁹ For instance, ‘isolationists’ like Mäki (Reference Mäki2009) and, with some qualifications Cartwright (Reference Cartwright2009), are of the view that the primary purpose of modelling is to ‘isolate’ a causal factor, mechanism or tendency that allegedly produces a given phenomenon. By contrast, ‘constructivists’ like Grüne-Yanoff (Reference Grüne-Yanoff2009), Knuuttila (Reference Knuuttila2009) or Sugden (Reference Sugden2009) argue that models are better understood as ‘artificial worlds’ or ‘parallel realities’ that act ‘as stands-ins for real-world objects or systems’ and which can generate different types of inferences about the ‘real world’. Therefore, neither ‘isolationists’ nor ‘constructivists’ believe that the primary purpose of modelling is to help scientists identify systematic deviations of actors’ actual behaviour from the ‘ideal’ behaviour predicted by the model.Footnote ¹⁰

Next, the thesis in Donato Rodríguez and Zamora-Bonilla (Reference Donato Rodríguez and Zamora-Bonilla2009) that the alleged disagreements among participants in recent debates about the nature and role of models can be, at least partly, reconciled within the ‘inferential approach’ is supported by the fact that participants in those debates like Mäki or Sugden (who advocate seemingly different accounts of the nature of models) nevertheless recognize that an important function of models is to let us acquire knowledge about the ‘real world’ by virtue of inferences that run from the former to the latter. For instance, Mäki (Reference Mäki2009: 38) argues that the primary purpose of modelling is not to explain why a certain phenomenon occurs but to explain how such a phenomenon ‘may have come about’. According to him, the model normally ‘articulates a possible mechanism that could have produced (or more weakly: could have contributed to the production of) the aggregate outcome that is observed’ (Mäki Reference Mäki2009: 38). Notably, Sugden (Reference Sugden2009) expounds a similar view in a discussion of a model of ‘herd’ behaviour in financial markets:

Elsewhere Sugden (Reference Sugden2009: 7), an economic theorist himself, recognizes that modellers often seem to use this type of abductive inference (i.e. inferring causes from effects) to make claims about the ‘real world’ by examining a model even though such claims are rarely explicit. In this respect, his notion of ‘credible worlds’ (Sugden Reference Sugden2000, Reference Sugden2009) is an attempt to identify the requirements that abductive inferences need to satisfy to be legitimate:

By contrast, our interpretation of Popper’s SA does not imply that the ‘artificial world’ represented by the model could be real. In other words, there is no intention by scientists to build an ‘artificial world’ that could hypothetically represent a part of the ‘real world’. Likewise, there is no presumption that there is some kind of ‘similarity’ between the causal factor or mechanism that generates a certain type of behaviour in the ‘model’ and the one that generates a social phenomenon in the ‘real world’. Rather, it is systematic deviations of actors’ observed (or actual) behaviour from the type of behaviour predicted by the model that social scientists seek to identify. It follows that, according to our interpretation of Popper’s SA, models do not let us ‘learn’ by virtue of inferences that run from the former to the ‘real world’: they allow us to learn by helping us ‘detect’ the reasons why actual individual behaviour in the ‘real world’ may not coincide with individual behaviour in the ‘artificial’ world captured by the model.

7. Summary and conclusions

Popper’s ‘Situational Analysis’ (SA) and the principle at its core, known as the ‘Rationality Principle’ (RP), constitute his methodological proposal for the ‘historical’ and the ‘theoretical’ social sciences. We claimed that the two hallmarks of SA are that: (i) social scientists implicitly assume they possess a ‘wider’ or ‘more complete’ view of the ‘problem-situation’ facing actors than the latter do; and (ii) use theoretical models as ideal ‘benchmark’ scenarios in order to identify the deviation of individual actors’ actual behaviour from the former. We also argued that the different versions of RP identified by Popper and several of his commentators are complementary and that the construction of the ideal ‘benchmark’ scenario in the ‘theoretical’ social sciences is grounded upon the ‘intermediate’ version of Popper’s RP (RPI) whereby scientists assume that actors’ individual behaviour is always adequate or appropriate to the ‘problem-situation’ (P-S) ‘as they could see it’ or, perhaps, ‘as they ought to see it’ if they possessed as much knowledge about it as the former do. According to us, the use of RPI enables the study of both ‘single-exit’ and ‘multiple-exit’ P-S in the ‘theoretical’ social sciences. We went on to argue that, notwithstanding Popper’s repeated claims to the contrary, SA is not a generalization of the neoclassical theory of individual behaviour to the rest of the social sciences, the reason being that neoclassical economists assume that economic agents do possess, on average, as much knowledge about the P-S facing them as they themselves do. In other words, unlike in the interpretation of Popper’s SA presented above, in the neoclassical theory of individual behaviour scientists are assumed not to be in a ‘privileged’ position vis-à-vis the actors with respect to the knowledge they possess. By contrast, we argued that such interpretation captures the methodology adopted by modern behavioural economists. More specifically, we showed that the methodology adopted by the latter encapsulates the two key features we associated above to SA. Lastly, we argued that our interpretation of Popper’s SA also highlights a way of learning about the ‘real world’ by building a theoretical model that, to the best of our knowledge, has gone unnoticed in the recent literature on the philosophy of economics.