Constructing Causal Realist Accounts: Causal Republicanism and Causation as Interpretation in International Relations¹

Constructing Causal Realist Accounts: Causal Republicanism and Causation as Interpretation in International Relations

In recent years, the issue of causation has called the attention of several IR scholars who have engaged in discussions of philosophy of science in an effort to analyse the treatment of causation and the provision of causal explanations in the discipline.² In this literature, neopositivist and critical realist accounts of causation advocate for the ontological reality of causation, while interpretivist approaches rather opt for an idealist take, which often derives in criticising or even discarding causal accounts all along. To be sure, in philosophy, the debate on the ontological status of causation has been traditionally divided between causal realism and causal idealism. Causal realism claims that "causal connections are among the real facts of the world independent of our cognition" (Kim, 1981, p. 303) and holds that causal mechanisms exist in the world independently from the human mind (Suganami, 2013, p. 624; Menzies, 2007) and it is the endeavour of science to discover them. In contrast, causal idealism conceives causation as an operation of the mind, a logical process that humans develop based on observation of constant conjunction between events that allow making sense of the world around (Hume, 2007a, 2007b), that is, understanding the world for proceeding in it, which does not imply unravelling its real workings.

Causal realism has been largely assumed as an ontological must for both neopositivists and critical realists. Traditionally, so-called neopositivism (as for the current version of positivism in the social sciences) conceives causation as a constant conjunction between cause an effect, where the verification of its repetition is what counts as causal explanation (Jackson, 2017). Never mind the observability of causes or causal mechanisms, neopositivists find certainty that they really exist by testing their repetition through statistical correlation (Jackson, 2011, 2017; Menzies, 2007).

On their part, critical realists acknowledge the ontological reality of causation but reject its demonstration or inferential approach based on statistical correlation. For critical realists, causal mechanisms exist independently from observation (here thus the realist take), but their scientific study depends crucially on the knowledge tools we use to describe them (and here the interpretive take) (Wight, 2006, p. 32; Suganami, 2013). Though science is the way to understand nature and social life with the greatest certainty possible, science will always be able to provide better accounts of causal mechanisms in so far as it improves its knowledge tools (i. e. concepts, methods) (Bhaskar, 1979). This means that statistical correlation between two or more elements can only open the possibility of a causal relationship (Bhaskar, 1979, p. 11), but explaining causation must consist of the description of causal mechanisms that generate change in the state of things, being those mechanisms usually unobservable, and thus needing to be contextualised, disentangled, classified and ultimately interpreted in order for a causal explanation to prove better than others (Bhaskar, 1979, p. 13; Wendt, 1987, pp. 346-57; Fairclough, Jessop, and Sayer, 2004; Wight, 2006, pp. 29-34; Banta, 2012). For critical realists, causation exists independently of the observer as part of natural and social reality, but its understanding depends on science as an interpretive tool-as will be further argued-to provide better explanations of it.

On the contrary, causation for interpretivists³ is a discursive tool to make sense of reality, thus laying mostly on the idealist side of the philosophical discussion, but crucially very often discarding causation as a relevant object of analysis. For instance, bearing upon the work of various interpretivist philosophers, David Campbell (1992) defines causation as a type of discursive economy where causal relationships are linguistically established between factors to meet concrete aims and anticipate possible consequences. Accordingly, this discursive economy may correspond more or less with particular behaviours observed in nature, but most of the times, and particularly regarding social life, causation helps making sense of motivations and consequences according to what is discursively established as more normalised, that is, more possible or more connected than something else (Campbell, 1992, Ch. 2).⁴ Causation, thus, is a particular form to interpret reality, but it is not something natural existing independently from those who employ it.

As put by Wight (2006), ontological assumptions surely explain these different takes on causation between these main IR epistemological stances. Neopositivists seem to transcend their ontological disparities through methodology, for the way to establish causation's existence is by statistical correlation. In contrast, critical realists claim epistemological relativism to describe the functioning of real causal mechanisms (Wight, 2006. Ch. 1), which logically derives in methodological pluralism in order to search for better descriptions of causation (e. g. Banta, 2012). Opposingly, interpretivists usually neglect the ontological status of causation, rather making causal discourses an object of interpretive (methods) critique.

However, long before causation came in vogue in the IR discipline, and along the unfolding debate, several philosophers of science and linguists have pointed the linguistic and interpretive character of causation, even from both (neo)positivist and critical realist philosophical stances. But more crucially, and as will be further shown, fundamental physics has kept causation under a great mantle of doubt, for it has questioned its existence in the physical reality of the universe. Hence, it is in the interest of this article to inquire into the extent to which a linguistic account of causation may dialogue with those philosophical stances most committed to the ontological reality of causation, and to reflect upon the benefits this might bring to their provision of causal explanations, which should not necessarily relinquish their ontological commitments by acknowledging causation mostly as an heuristic for explaining social and international phenomena, where 'causal factors' are often difficult if not impossible to 'observe' as physical objects. By a 'linguistic account' I mean placing particular emphasis on the linguistic construction of causation, that is, reflecting upon the role of language in understanding causation as a phenomenon that may be part of an independent reality that exists and occurs apart from our thoughts of it, but whose acknowledgement and scientific understanding is subject to our tools for interpreting it.

Yet, and crucially for reasons of space, I here focus on dialoguing with the philosophical position of causal realism in general, and not with any IR theory in particular. In the second part, for illustrative purpose, I focus on the philo-sophical/epistemological stand of critical realism in IR, mainly because its methodological pluralism leaves wider room for accepting the relevance of interpretation for addressing causation as an object of scientific inquire and making causal claims. This implies that in this article I will neither address neopositivist accounts nor interpretive ones (and thus neither any theoretical approach engaging them), despite the surely useful dialogue this might entail. Nevertheless, it is my hope that the purported discussion on the relationship between language and the ontological real status of causation might persuade a good amount of neopositivists to think about it, and may also suggest connections with several interpretivist and idealist takes on causation that should certainly be developed somewhere else.

Hence, in what follows, I will develop my arguments by resorting to the philosophy-physics stance of 'causal republicanism,' which has underscored the heuristic character of causation as a useful interpretive tool to make sense of the physical world, despite the unwarranted status of causation in fundamental physics. Arguably, the fact that the ontological status of causation has no grounds in fundamental physics should at least tempt causal realists, and particularly critical realists, to nuance their conception of causation as something real, existing ontologically independent in natural and social life. Thus, this article will first present the compatibilities of causal republicanism with some interpretive and linguistic views of causation, and will then proceed to examine an example of IR critical realist work that offers an analysis of the terms in which agents understand the causal power of ideas for explaining the unfolding of regionalism in the Andean Community, which I believe facilitates illustrating the dialogue between critical realism and causal republicanism.

It will be argued that for causal realists like critical realists (and surely most neo-positivists), the most relevant contribution of causal republicanism is not just pointing to the unwarranted status of causation in fundamental physics, but acknowledging that causation becomes more ofan interpretive exercise the more unobservable the phenomena intended to be explained are. In these terms, I contend that as the observability of the material existence of objects of study becomes harder-as in the case of ideas-, the real existence of causation becomes more dependent on the social reality of language as the means that bounds together humans' understanding, and thus causation becomes less independent from human descriptions of it. This should not entirely revoke causation's ontological status, but on the contrary, help increasing the usefulness of causal interpretations by providing more accurate descriptions of the alleged causal powers of ideational structures like collective identity and regional institutions, and generally of social structures that are rather difficult to materially observe.

Causal Republicanism and the Linguistic Construction of Causation

This section brings into dialogue the philosophy-physics stance of 'causal republicanism' with the accounts of some philosophers of science and linguists who have highlighted the interpretive character of causation and its linguistic construction. Thus, insights from one perspective and the other will be linked subsequently in order to develop such dialogue. Causal republicanism holds that despite the impossibility to ground it in fundamental physics, causation is a way of observation that is crucial for making sense of the world and proceeding in it in terms of achieving aims through more effective strategies than others.⁵ There are two main reasons why causation does not find support in fundamental physics: first, because it is not easy-and sometimes neither possible-to isolate causes from initial and boundary conditions so their causal power can be warranted by counterfactual thinking or manipulation (Eagle, 2007; Woodward, 2007, pp. 69, 90-3). A main condition for establishing a causal relationship between A (cause) and B (effect) is that in the absence of A, B would have not occurred. Fundamental physics requires that explanatory laws are fulfilled at all levels of analysis, including that of the whole universe. As interventions, choosing the relevant counterfactuals is easier in closed systems or 'small worlds' (i. e. scientific laboratories), where several conditions or 'interventions' may be pointed as 'external' and more easily isolated from the causal relationship intended to be explained (as in the case of a patient taking drug D as the cause for its recovery from a disease). However, if the system is a big and open one like the whole universe (and it will be argued, social life), counterfactuals are practically impossible to choose as they cannot be set as external to the system (Woodward, 2007, p. 93; Eagle, 2007; Kutach, 2007).

The discretional choosing of counterfactuals makes it difficult to ensure which facts could have been of greater influence on the effect through imagining them not having occurred or manipulating them when possible (as in controlled scientific experiments). In order to become a 'causal law' it would be necessary to verify the constant and infallible repetition of the causal power of those factors chosen for counterfactual thinking or experimenting.

Moreover, isolating counterfactuals is harder if the manipulation of one (or some) of them actually affects the influence of other factors in the event taken as the effect (Eagle, 2007, pp. 183, 185; Price, 2007, pp. 280-1). As Eagle (2007, p. 160) points out, it is for explanatory purposes that salience is attributed to certain factors that we define as causes, but such salience "... appears nowhere in its purely physical description, and no appeal to the underlying physics will establish any priority for the genuine causes over the spurious causes."

The previous points can be illustrated with Neeleman and van de Koot's example of John breaking a window with a hammer, which these linguists use to argue that causation is a human intuition, a useful tool to understand why the window broke, but not a reflex of all the factors that explain its breaking, like the man's force, the way he grasped the hammer, his motives for doing it and so forth (Neeleman and van de Koot, 2010, p. 79).⁶ Accordingly, causal explanations seek to highlight a crucial contributing factor (CCF) that mainly caused a particular effect, but the causal counterfactual condition that without A (the cause) B (the effect) would not have occurred is nothing but a heuristic device, because it will never be possible to account for all the facts that made B happen, that is, that caused it (Neeleman and van de Koot, 2010, p. 80). As these authors state it, context is crucial: interpretation takes place where not only observation, but also logical connections are carried out. But even in a highly specific and short-ranged context like a window breaking in a room, any causal narrative will be an abstraction, a synthesis of what happened, which obeys to the observer's interpretation of the context (as pointed also by Lebow, 2014, Ch. 2 and Wendt, 2015, pp. 168, 255, and Ch. 9).

Clearly, choosing a CCF is actually an interpretive move: for one person the CCF could be John, for another the hammer, and for a third one the fact that John was out of control, which does not mean that all those factors shall not be real causal. While identifying the CCF could be a matter of knowledge about the context, note that it does not depend on it (not even on the quality of research): all three people could know the context, even know that John was out of control, and still choose different CCFs.⁷ The point here is not to find out why they choose to utter one CCF or another, but the fact that choosing a CCF is, besides whatever else, a linguistic deed that makes the CCF the (main) cause among many possible others, and so the causal explanation points a cause that is linguistically produced.

To be sure, the former point does not imply that the mere fact of linguistically pointing a cause would turn anything mentioned into a cause-say, the event changing the state of things. In the context of the window breaking, someone pointing to John's shoe as the CCF would just not be taken seriously, unless some evidence could be found that something happened withJohn's shoe that made him slip or stumble, so he fell (with the hammer in hand) over the window. Pointing to a CCF is a linguistic deed that is contingent upon the context towards which the causal account is provided, and this implies ruling out the possibilities that such context prevents observing and interpreting. Context, thus, makes causal factors real, at least more real than those progressively ruled out by context. Causation might be also employed in abstract imaginative scenarios (as someone wonders 'what could happen if...?'), but it is mainly a tool that humans use to make sense of concrete contexts where a set of possible causes is defined and a CCF (or a bunch of them) is linguistically established, crucially among other things because of the concepts (causation included) they (humans) use to name and organise events.

Furthermore, note that the example of the window breaking in the room is merely illustrative, but defining the context of possible CCFs is subject to the observability of factors (thus the italics in defined) and the logics reckoned among them (one knows, for instance, that it is highly improbable that the hammer flies onto the window by its own). If everything in natural and social life were as easy to observe (observable) as John, the hammer, and the broken glass, perhaps causation would depend less on language and more on physical observation. However, following causal republicanism -and as will be seen below-not even the observability of events allows causation to be granted scientific status.

The second reason why causation is not supported by fundamental physics is that temporal asymmetry between cause and effect cannot be ensured due to the relative character of time and our consequent location in it.⁸ In order to warrant temporal asymmetry, we need to locate the cause earlier than the effect, which implies fixing the past in order to establish what is present and future. But this location in time is relative to what us (observers) consider to be the past, more concretely, to what we actually know about the past.⁹ As Price points out, this is relative to each observer, as each may fix the past in different fashion (according to their own knowledge of the past), or in other words, each observer is embedded in her own (conception of) time (Price, 2007, pp. 267-9). Even though they all observe the same fact (effect) in the same present (at the same time), locating what happened in the past may vary according to the way each observer fixes the past (knowing what happened before the effect came about) (Price, 2007, pp. 269-74).

Hence, in order to know that an effect B followed a cause A, we need to fix the past as to know that indeed B was caused by A. Note that fixing the past requires the ordering of past events so we can infer that it was A what caused B. Crucially, this ordering of the past is also a linguistic deed that derives from our interpretation (a more or less accurate 'observation') of what happened. As Price shows, our knowledge of the past will always be limited and perspectival, and therefore identifying a cause as a precedent of an effect will always be dependent on our knowledge, and further, on our linguistic ordering of the past (what happened), or in other words, on our time location (Price, 2007, pp. 280-1).¹⁰ Moreover, depending on different locations in time it is possible to situate the effect as preceding the cause, in a form of backwards causation (Price, 2007; Kutach, 2007). Indeed, it is often the case that only once the 'effect' is located it is possible to locate the cause at some point in the past, thus violating the asymmetry condition, for instead of the effect not having occurred without and after the cause, it is the cause what actually did not exist before the effect was observed.

Resuming our previous example, it is quite straightforward to understand that the window broke (effect) because John hit it with a hammer. But if the cause is located as John being angry (not necessarily 'out of control'), it is very probable that one realised John's anger (the cause) only after observing the window breaking (with the hammer), so the cause only appeared (became 'real') after the effect. In these terms, the more unobservable the cause, the more possible it is that effects establish its existence as 'cause,' because more interpretation is needed to locate it as a (the) cause. Moreover, our actions may change the past if something we do at present not only alters the meaning of what happened to us, but actually to what someone else thought it happened (Wendt, 2015, pp. 194-8; Price, 2007, p. 277). Someone thinking that the cause for the window breaking was John's anger might change their understanding if they later find out that all John wanted was to send a signal to someone outside the house who was expecting the window breaking to proceed with some kind of plan (which does not preclude John's anger as a real event, though not necessarily as a cause). Hence, the fact that humans locate events conceived as causes and effects in a temporal fixed and conjoined dimension is mainly an operation of the mind, derived from the need to make sense of how certain things lead to others, just as causal republicanism purports. This operation of the mind is carried through the linguistic ordering of events, which obeys to the interpretation agents make of what happened. Crucially, ordering events in different fashion may yield different causal accounts, and the unobservability of "real" events will surely make it more difficult to dispute them.

The former arguments plainly coincide with those of philosopher of physics John D. Norton (2003), who bearing upon the work of analytic philosophers like Norwood R. Hanson (1958) and Ernest Nagel (1961) explains that causes and causal principles are "heuristically useful notions" which, in appropriately restricted domains, work in similar fashion as caloric theory or gravitational forces: although in many circumstances heat behaves as if it were a material fluid, and gravity does as if it were a force, science has proven that there is no caloric in the world and heat is not a material substance, and Albert Einstein demonstrated that gravity is not a force (Norton, 2003, pp. 2, 21). This argument is in line with Hanson's, for whom the connection between causes and effects is established by our theories, not by nature (Hanson, 1958, p. 64, quoted in Kim, 1981, p. 293). Arguably, causation is a tool for explanation produced by the human need to explain:

Conceiving causation as a heuristic is founded in what philosopher Ben Gibran (2014) terms 'causal anthropomorphism,' namely the view that humans accommodate causal relationships to their 'need of understanding,' and so causation acquires an anthropomorphic character, which should not necessarily deny its real existence. Of course, such 'accommodation' is based on observation and experience (experiments), and hence it is not entirely arbitrary; but as a tool for explanation, what humans accommodate as causes may not fully correspond to what happens in nature/reality, and thus causation becomes a particular way to understand reality, concretely in causal terms. This latter point is largely developed by Ernest Nagel (1961, p. 324, emphasis original), for whom "the acceptance of the principle of causality ... is an analytic consequence of what is commonly understood as 'theoretical science'", where language plays a crucial role:

According to Kim (1981, pp. 297-298 referring to Hanson, 1954, Ch. 3), Hanson also highlights the importance of concepts for making causal claims, for they derive their explanatory power from the interlocking patterns between them, and a theory's explanatory power is based on its multi-layered system of concepts. The importance of concepts in Hanson's account of causation stresses the importance of language for the provision of causal explanations, for concepts are linguistic constructions referred to 'something' existing somewhere where it can be perceived (be that the 'outside' world or our 'inside' thoughts). In these terms, the alignment of a linguistic conception of causation with causal republicanism is warranted by causal anthropomorphism. From an interpretive position-namely one that vindicates the relevance of interpretation for making sense of reality-,¹¹ humans know the world around them through language, and it is language what makes possible the development and sophistication of human thinking. This stand is sustained by linguist Tanya Reinhart, who holds that:

Causation as derived from 'the input from the world' acquires its real onto-logical status as part of an explanation that is useful to understand reality, but not necessarily as a real phenomenon on its own whose 'nature' or 'ontological existence' is completely independent from scientific interpretation, but instead depends on its linguistic construction in order to become more or less useful for proceeding in the world. In the vein of causal republicanism, philosopher of physics Antony Eagle associates causation-concretely, what he terms 'folk causation,' since causation does not have scientific status-with linguistic practices that he finds useful to distinguish between more effective strategies than others. Accordingly, humans are deliberative agents who need to choose between strategies in order to achieve their aims. Deliberation takes place thanks to language, and so causation is a linguistic practice that allows humans to choose better strategies than others. Even though counterfactual determinism deprives causation of scientific status, counterfactual reasoning is crucial for agents to deliberate on what might be the consequences of their actions in order to meet certain aims (Eagle, 2007, p. 183 passim, substantially building upon Cartwright, 1979).

Causation is an exercise of generalisation, much simpler than the relations found in fundamental physics (Eagle, 2007, p. 187). As Eagle (2007, pp. 184-5) points out, when we demand an explanation about something, we expect to know what was most responsible for the outcome, so we can reproduce it by focusing our efforts on that aspect; thus, without generalisable explanations, reasonable deliberation would be impossible, and humans do not need to counterfactually prove the determinant character of every factor involved in the advent of an event in order to anticipate the most probable consequences of their actions. As a way to know the world, thus, causation is an epistemic tool, very useful to deal with ignorance, time, and probability (Price, 2007, pp. 282-4), without this implying that things in the world are indeed causally connected. Our deliberative nature makes causation inescapable, but as Beebee (2007, p. 232) argues, it is humans' need to make causal inferences what is natural, and not causation as such. In these terms, instead of holding the status of an ontological reality, causation is mostly an interpretive exercise-one of interpretation-, much dependent on the language used for it to understand which strategies may be more effective than others, as will be shown in the following section.

Causal Republicanism and the Causal Power of Ideas in the Unfolding of Regionalism

A previous critical realist work on the Andean Community (AC) regionalist project¹² (Prieto, 2016; see also Prieto, 2015 and 2020) developed an interpretive account focused on the language of causation employed by agents when referring to ideas informing their actions.¹³ Thirty-two national decision makers and regional bureaucrats were interviewed about the role of collective identity and regional institutions as ideational structures that might have explained their actions for fulfilling regional aims, and were chosen according to their salient role as negotiators and decision makers in the three case studies addressed (as further shown). This analysis adopted a critical realist stand that entails acknowledging the independent existence of reality from actors' views, and incorporates Alexander Wendt's (1998, 1999) differentiation between constitutive and causal theorising, and the critical realist conception of causation that considers causes as mechanisms that produce changes in the state of things. Yet, in order to distinguish between constitutive and causal effects of ideas on state action, an interpretive method was used which bases on the linguistic analysis of decision-makers' statements about ideas related to so-called 'Andean' collective identity and regional institutions (the latter being Andean norms and institutional organisations composing the Andean System of Integration). This method consists of identifying transitive verbs that denote transformation,¹⁴ metaphors implying changes in the state of things,¹⁵ and expressions where ideas operate as reasons for action,¹⁶ as cases where ideas play a causal role. In all other cases where interviewees referred to ideas without deploying these types of linguistic forms, ideas were taken as playing only a constitutive role. Assuming that ideas have an independent influence on action accounts for a critical realist ontological and epistemic commitment, and interpretivism operates just as a method. By focusing on language as the main object of study-language is the means to know (observe) the ideas informing state action carried out by decision-makers-, this work shows that causation is essentially a linguistic construction, which far from being arbitrary or subject to each interviewee's perception, and in line with causal republicanism, contributes to understand the causal power of ideas from a critical realist perspective, and to explain the social construction of Andean regionalism.

Some examples of the analysis of the causal role of collective identity and regional institutions in three selected case studies are provided in Table 1.¹⁷ Regarding collective identity, interviewees pointed the political cost the Peruvian government refused to pay for leaving the AC as the cause for remaining in it when Peru did not want to adopt certain commitments to complete a common market, and they associated this political cost with issues of collective identity when asserting that Peru had 'an integrationist identity', and that Peru shared an idea of 'fraternity' with the other Andean countries.¹⁸ Similarly, strengthening the Andean integration process as a motivation (cause) to pursue collective negotiations with the EU denoted strong AC members' identification with the AC regionalist project,¹⁹ and the rupture of the ideological consensus on neoliberalism and free trade was pointed by interviewees as a main cause for the breaking-off of FTA collective negotiations with the EU,²⁰ where the ideological dimension was distinguished as one of the three dimensions composing Andean collective identity, being the other two the cultural and the intergroup dimensions. In case study 3, interviewees made statements that denoted strong identification with the AC based on the benefits derived from regional cooperation and those expected to be provided by the PIDS, ²¹ and the different visions that AC member states had over the conception of development (corresponding to the ideological dimension of Andean collective identity) were pointed by interviewees as causing 'conflicts', 'delays,' and 'slowness' in the PIDS' implementation, thus accounting for causal effects.²²

Regarding the causal role of regional institutions, and basing on Wendt's (1999, pp. 266-278) distinction of degrees of norm internalisation, AC member countries' identification with the regionalist project is explained as a consequence of the benefits provided by Andean norms and regional organisations (second degree of internalisation), and identification as a group in the form of Andean norms being part of Andean collective identity (third degree of internalisation). Regarding the second degree, interviewees pointed out the support that Andean norms offered for member countries' trade, social, and environmental policies,²⁴ the usefulness of Andean norms as a guide to negotiate with third parties,²⁵ and the institutional flexibility provided by Andean norms and regional organisations as a motivation to remain an AC member despite the difficulties in reaching consensus.²⁶ With regard to the third degree of internalisation, interviewees asserted that Andean identity was created by signing the AC's foundational treaty (C3), that the AC's Common Foreign Policy Mechanism mandated AC members to negotiate as a bloc (thus to identify with one another) (C4), and that Andean norms "accounted for member states' political will of acting under binding norms" (A11), thus showing that AC norms were conceived as part of member states' identity.

Through linguistic analysis, this research identifies several cases where interviewees employed transitive verbs denoting transformation and metaphors implying change when they referred to ideas of collective identity and regional institutions, and also on some occasions they put these ideas as reasons for action, thus making possible to distinguish the causal role of ideas. But clearly these views on the role of ideas were the product of interviewees' interpretation of what happened with those ideas, and moreover, locating the causal role of ideas in interviewees' discourse was the product of the researcher's own interpretation of their discourse based on a method of linguistic analysis. The result from such analysis is that the causal power of ideas was somehow created by interviewees' views/discourse on their role, and more crucially, linguistically produced. Furthermore, the interpretive method and the interpretation of interviewees' discourse also contributed significantly to grant such causal power to ideas. Yet, that is not to say this whole analysis is nothing more than a linguistic creation, and that therefore by simply changing the language, the whole assessment of the explanatory role of ideas would crash down like a house of cards. This is precisely where the engagement of critical realism and causal republicanism comes in.

A main ontological tenet of critical realism is that reality is independent from actors' perspectives, but not completely independent (Wight, 2006, pp. 26-9; Lebow, 2014, p. 49). On the one hand, natural reality is not completely independent from actors' minds as long as those things that are part of nature have social meaning (Wendt, 1999, Ch. 3). On the other hand, social meaning, and those shared ideas that compose social reality-like conventions, institutions, and social structures in general-, are the product of social interaction, where agents cannot transform such reality at their will by simply changing their particular views on it. That is precisely why social reality is considered to exist independently from actors' perspectives: because as a social product, reality depends on the others' views to be considered as such.²⁷ This works less for materially observable phenomena like a twister, whose impact on trees and houses is barely dependent on people's shared views. But social structures like identity and institutions (as well as the market, the state and capitalism) are much more dependent on people's shared views for them to be real (Wendt, 1999, pp. 67-77; 2015, pp. 258-60). Consequently, if an individual or a group of agents tries to change reality by changing their perceptions on it, the success of such transforming action will crucially depend on the other actors' acceptance of the soundness and validity of those new or different perceptions. Hence, the margin for change that transforming actors have is limited by what the other actors are ready or capable to accept (Bially Mattern, 2005; Krebs and Jackson, 2007). This social constraint to perceptions is what accounts for the independent dimension of reality held by critical realism, at least applied to the social world (cf. Wight, 2006, pp. 26-9, 31-4).²⁸ In this exercise of social interaction, it can be clearly seen that language plays a crucial role, since language is the main means for constructing, transmitting, and confronting perceptions among agents. Since language is a social construction (with the same limitations for change formerly pointed out), and since it is a crucial means for constructing reality (Epstein, 2013), causation as a way for understanding the world should also be conceived as a social construction, more concretely, a linguistic one.

In these terms, considering causation as a linguistic construction from a critical realist perspective does not clash with the causal realism engaged by critical realism (cf. Suganami, 2013). On the contrary, language is the crucial structure that makes causation not to be completely dependent on human thought, or put in other words, to a certain extent independent from it. It is the social character of language what makes causation 'real' and partly independent from human thought. One cannot say anything about something without running the risk of being simply dismissed if what is said does not make sense to others. Critical realism, and the causal realism it purports, would much benefit from interpretive analysis and from seriously engaging language in order to contend the independent existence of social reality from the minds of the actors involved in it.²⁹ The following quote from Roy Bhaskar (1978, pp. 249-50, brackets mine)-one of the leading figures of critical realism-, is of great explanatory usefulness:

Science, then, is the systematic attempt to express in thought the structures and ways of acting of things that exist and act independently of thought.

If we replace 'thought' with 'language' in the former quote, we have a good basis for clarifying the 'independency puzzle' set by critical realism. Such replacement is warranted by interpretivism's assumption that (human) thought is only possible through language. That said, Bhaskar's posture confirms causation as a thought (linguistic) device to understand the nature of things. If the world can only be expressed in speech, then such expression depends on language. Hence, as a process in thought, science is the linguistic framing of the natures and constitutions of things that exist independently of language (thought). Things exist independently of language (thought), but their acknowledgement through science depends on language (thought). If causation is part of such scientific acknowledgement, so does causation depend on language (thought). What is independent from human thought (language) is the existence and nature of things. But Bhaskar himself stresses that the ways we understand the nature of things (science and causation) are not independent from thought.

The causal realism advanced by critical realism holds that causal powers are part of nature and thus have an independent existence from human thought. This might be true for certain closed systems in which natural scientists often operate (such as labs and controlled experiments). But in open systems such as social life and international politics, the independent existence of social structures, like ideas, is not granted by nature, but by language. It is the social character of language what makes social structures not to depend solely on what agents think or say of them. As the main means for thinking and understanding, language bounds the possibilities of thought and compels agents to employ it in a way they can make themselves to be understood by others. Hence, as pointed by causal republicanism, causation is a very useful way of ordering events so they can be explained to and understood by others.

Coming back to the reviewed piece of critical realist work, it seems quite sound to assess the causal power of ideas on state action through understanding (interpreting) how state actors interpreted such ideas. The number of factors involved in the case studies (norms, interests, identity, aims, costs, benefits, and so forth) is so wide, and their interrelations are so complex (of which the ones present in Table 1 are just some examples) that causal accounts like those posed by interviewees are very useful for understanding state action, just as causal republicanism purports. Moreover, the reality of the interviewed national and regional officials was their interaction in the unfolding of the AC regionalist project, where institutional and social processes such as bargaining, helping each other, and disagreeing on ideological stances were real only inasmuch as the actors involved acknowledged such processes. An alien coming to some Andean country and observing the interaction between national and regional bureaucrats at some meeting room would never understand that regionalism was being there pursued, unless that alien spoke Spanish, or someone else speaking the alien's language explained this to them. Hence, although the language used by interviewees to refer to the role of ideas may vary, it could not be any words they picked up in their minds when doing the interview, because they were socially and linguistically bound to a social reality they not only experienced, but knew that the other actors (and interviewees) shared.³⁰ Therefore, had some interviewees chosen to say things completely unrelated to or just 'crazy' about Andean regionalism (out-of-context crucial causal factors), they would run the risk of being just denied or ruled out by other interviewees, or discarded as valid testimonies by the researcher in case it could be observed that what they were saying about ideas of collective identity and regional institutions was completely awkward in relation to what the others said. There is room for interpretation, yes, but from a critical realist-interpretive perspective such room is socially bounded and limited, and linguistically too (Bhaskar, 1979, p. 39).

It is important to make clear that, in order to avoid bias, interviewees were never asked about the causal role or effects of identity and regional institutions, neither about the extent to which these ideational structures caused state action. Questions were asked in general about identity and institutions, their influence, their relevance (which surely suggested causal language, but not explicitly so). Yet, just in line to what Elder-Vass (2011) explains about the causal power of discourse, and in full alignment with causal republicanism, many interviewees (not all) referred to the role of identity and institutions in causal terms, surely because socially the discourse of causation is so powerful that tends to make agents to think of the relevance of ideas in causal terms. In other words, the (socially prevalent) discourse of causation made interviewees conceive the role/relevance ofregional identity and institutions in causal terms. It could be said that this accounts for interviewees' particular interpretation, but Elder-Vass's account of discourse circles and constructivism's structural conception of ideas, added to the advocated reflection on intertextuality and context regarding the interpretive method employed, should lead to the more sensitive conclusion that positing the causal character of ideas (ideational structures) does not derive from agents' capricious or discretional interpretations, but rather from structural interpretations and discourse circles that make most (not necessarily all) agents acknowledge the causal power of ideas. Acknowledging such causal power does not depend on agents' personal beliefs or interpretations about identity and institutions-thus holding an independent existence form agents' views-, but is also not simply subject to the researcher's interpretation, for its independent character is grounded in the fact that several interviewees shared causal accounts of such ideas.

Undoubtedly, the interpretive approach here advocated is not free from limitations. Besides the fact that the researcher's interpretation is based on interviewees' ones (double hermeneutics), and thus the former's causal accounts are constructed upon the latter's ones, the interpretive method deployed for distinguishing causal from constitutive effects of ideas cannot be assumed to be the best one for seeking such purpose. In other words, the results derived from interpreting agents' causal accounts can be disputed by other causal accounts that use a different language, just in line with causal republicanism's tenets. However, at least from interpretivism's philosophical grounds, it is precisely the engagement of validity criteria for interpretation (intertextuality, transparency, and reflexivity) which confirms the value of interpretive analysis for making knowledge claims (Prieto, 2016).³¹ It was the particular methodological design of the deployed interpretive analysis what allowed assessing the real role of ideas in orienting the unfolding of regionalism in the AC, and in the absence of a better method it could not be possible to understand the constitutive and causal relationships between identity and institutions, neither the relevance of the idea of 'political cost', ideology, and intergroup identity for understanding why regionalism advanced but also receded at times. Besides the fact that basically no other study has addressed the relevance of ideas in orienting Andean regionalism, the interpretive method introduced for distinguishing causal from constitutive effects issues an original contribution to constructivist and critical realist analysis for assessing the explanatory role of ideas.

In sum, this piece of research shows that recognising that the causal power of ideas is linguistically framed does not equate to acknowledging it as unreal. Quite on the contrary, the linguistic analysis of interviewees' views on the role of collective identity and regional institutions allows understanding how socially real is the influence of those ideational structures on the construction of Andean regionalism. This not only helps improving the quality of a constructivist and critical realist explanation, but crucially in accordance with causal republicanism, it shows that it is language and interpretation what make causation real, that is, understandable and useful to others.

Conclusion

In line with the philosophy-physics stand of causal republicanism, this article has vindicated the useful character of causation as a heuristic tool for proceeding in the world and making sense of the relationships between real structures and processes, in the sense of existing independently from agents' views. The analysis of the example of critical realist IR work here developed calls philosophical realists (both critical realists and neopostivists) to acknowledge the importance of language in the framing of their causal accounts, and to understand that the more unobservable social and international issues are, the more causal accounts become interpretations that crucially depend on the language employed for acquiring their usefulness and, to be sure, their real existence, in terms of being influential or even determining for social groups at large.

The causal powers of ideational structures such as collective identity and institutional settings are as real as their interpretation is supported by evidence, and as logical argumentation is based on sound 'causal' language. But those powers are not laying hidden waiting to be discovered. They can only be interpreted and be subjected to discussion and logical tests. Accepting the causal republicanist view of causation as interpretation does question causation's real existence in social (and natural) life, but the real character of causal interpretations is granted by the social character of language and the soundness of scientific language, which substantially prevent its total dependence from the observer's mind, just in line with what many philosophical realists acknowledge about the working of social reality, in which agents' ideas and interpretations are real and independent from their willing choices or views, but not as independent as natural/material phenomena like a window breaking or a twister.