Don’t think it’s a good idea! Four building sites of the ‘ideas school’

Abstract

Ideational explanations of policy change are popular in the fields of political economy, comparative politics and policy analysis. And yet, to make the case for ideational explanations, we must make further progress on the nature of ideas, where they come from, what they consist of, and how they change over time. We highlight four critical building sites concerning the definitional aspects of ideational explanations, micro-foundations, mechanisms and the difference between ideational and cognitive analysis. We make recommendations on how to carry out work in the building sites and describe the range of suggestions and ways forward found in the articles of this Symposium. We also suggest cross-fertilising political science with the findings of neighbouring disciplines that have developed empirically robust models of ideation and cognition.

Keywords:

• Ideas
• explanation
• micro-foundations
• mechanisms
• cognitive analysis
• référentiel

It is somewhat uncontroversial to argue that ideas matter in the explanation of policy change (Cairney 2019). Ideational explanations maintain that alterations in the core dimensions of public policy can be causally generated by ideational elements instead of classic political science variables such as institutions, material forces, and the constellation of preferences and interests. Yet, ideas do not differ from other explanatory variables. Hence, to make a causal argument for ideas as the source of policy change we need, at a minimum, to identify what ideas are and then to put forward and test explanatory theoretically-founded mechanisms where ideational elements can be clearly identified and distinguished from other elements and sources of causation. These are two different constitutive steps, of course. We can indeed reconstruct the debate of the last twenty years or so across different fields of policy analysis, such as comparative public policy and political economy, as an effort to gain more precision on the definitional aspects and on the explanatory mechanisms – for example, by setting conditions under which different types of ideas are suited to carry out the entire explanatory work; or complement institutions and interests; or are used strategically by the policy actors; or, in yet another permutation, constitute and endogenously shape interests.

Entire volumes have been dedicated to the ideational turn (Béland and Cox 2011; for a succinct yet effective overview of ideational explanations in political science until the mid-2000s see Parsons 2007: chapter 4; on ideas and institutions see Béland 2019). Hence, we will not even try to revisit such debate or summarise its results. Instead, we wish to discuss four critical issues that, in our view, the literature has not explicitly and persuasively dealt with so far. We see these critical issues as building sites, meaning that our critical overview is oriented towards the development and refinement of ideational analysis.

The choice to frame a given issue as critical or not depends on one’s lenses, that is research tradition, ontology and epistemology. In particular, we care about definitions, micro-foundations, mechanisms and cognition. Other lenses are definitively justified and feasible and may well lead to the identification of different issues. This is then to say that our approach is one among many that could be applied to sharpen ideational analysis.

But what are these building sites, then? First, an ideational researcher must be clear on what ideas consist of (Berman 2013). What are the properties that ideas possess and what makes them distinguishable from other variables?

Closely related to the definitional problem is the issue of the origin of ideas. An ideational explanation is stronger if it tells us where do ideas come from and how their genesis and foundations interact with cognitive, material and institutional elements. Recently, Jabko (2019: 496) has argued that ‘[…] there is a case for scholars of economic policy making to move away from holistic concepts of ideas and to devote more attention to fine-grained, open-ended discursive practices.’ The literature on the political and epistemic construction of ideas (Anstead 2018; Clift 2018) is yet another signal of attention towards the granular analysis of ideas and their micro-foundations. Recalling what we said about the lenses we adopt to frame an issue as critical, we acknowledge that micro-foundations alone cannot adjudicate on the validity of an approach. And by switching to holistic lenses, one may find that they are not necessary. Yet, even if we do not adopt the language of micro-foundations, we still believe that knowing where do ideas come from, the micro-processes that generate and shape them, and the agents or carriers of ideational elements in policy change is essential – thus some connection between micro and macro must be spelled out. This consideration ties in with the fact that ideational research has started to move away from conceptions of ideas as stable and hardly mutable entities (Anstead 2018; Clift 2018; Eleveld 2016; Franchino 2020; Jabko 2019). A focus on micro-foundations, hence, is also germane to a trend that approaches the empirical dynamism and changing nature of ideas more systematically.

Third comes the causal mechanism(s) that explains policy change in the context of an ideational framework. This is the domain where we search for frameworks that subsume empirically observable ideational categories (like causal beliefs or policy paradigms) and then articulate the explanatory work of such categories through an explicitly theorised mechanism. Macro models of ideational change are more robust if they draw on empirically validated patterns and behavioural mechanisms that characterise individuals - in short, if they are micro-founded.

Once we grasp the mechanism(s), we must answer the question whether ideational explanations are different from cognitive explanations – and how. Cognition is a well-known concept and a much-studied process in all social sciences. By contrast, ideas as such do not seem to have citizenship outside political science, political economy and philosophy.

What is the point of our contribution, then? We discuss, clarify and engage constructively with the building sites. We also show how we can usefully turn to cognitive/social psychology and behavioural sciences to let the pieces of the ‘ideas matter’ puzzle fall into place. Here we join other recent contributions, e.g. Jabko (2019), Jones (2017), Kamkhaji and Radaelli (2017), Van Esch (2014) and Van Esch and Swinkels (2015) - along with the original contributions included in this Symposium. We wish to be clear on one point: our aim is not to attack the work of ideational political scientists. Mindful of the lesson of Craig Parsons (2007: 105 and 131), we know that to exclude or belittle ideational explanations for bad reasons such as ungrounded scepticism and misinterpretation of their claims only leads to less scientific discussions. And we clarify that we are dealing only with one type of ideational approach, the so-called ‘ideas school’ that flourished on the seminal contribution of Peter Hall. The ‘ideas school’ is a precious edifice that we wish to preserve and partly restore, when we ought to, and partly develop, when we can. We conclude that some of the classic ideational models and mechanisms are not corroborated by and based on the extensive, solid findings of other disciplines that study cognition, information processing and learning. It follows that we either we accept these findings and revisit our explanations - or we produce equally solid findings to prove that behavioural sciences, cognitive and social psychology are not suited to provide ideational models with empirically valid assumptions and foundations about cognition and behaviour.

In terms of organisation of our work, the next four Sections are dedicated to the four building sites, before we discuss how to make progress and conclude. We spend more time on the description of the building sites than on what we see as a possible way forward (Section 6) because we leave most of the exemplifications of the solutions to the articles included in this Symposium.

What is an idea made of?

Let us start from the basic question about the nature of ideas. Here the literature has provided a diverse range of suggestions, without converging however (Berman 2013). As a preliminary analytical step, if we argue that ‘ideas matter’ (Swinkels 2020) we should be able to clarify what kind of categories and concepts we associate with our understanding of ideas. This is because it is useful to know how some clearly identifiable operational ideational elements have causal effects on processes of change (here we paraphrase Parsons 2007: 128–129).

Berman (2013) discusses political culture, ideology, beliefs, and norms – observing that sometimes the literature has conflated these concepts with ideas (ibid.: 229). For Parsons, the category of ‘ideational elements’ includes symbols, practices, identities, culture, beliefs and ideas (Parsons 2007: 95). These contributions provide a good starting point for tackling the definitional challenge. However, they do not solve our particular problem. Granted that ideas are one kind of ideational element that is different form the others, what is an idea, exactly? Further, if ideas do manifest in different, distinct varieties why is it the case that we can precisely describe these varieties, but we struggle to define the general concept to which they belong?

Béland and Cox (2011: 3–4) collapse ideas onto causal beliefs and much of the empirical literature seems to agree in that it often employs ideas as, in fact, causal beliefs. This is a concise definition that does not present particular foundational problems, since there is a vast literature on causal beliefs – suffice it to mention the advocacy coalitions framework (Jenkins-Smith and Sabatier 1994). But if we talk about ‘ideas’, at least within the perimeter of the school we are considering, it is because we want to go beyond causal beliefs (that actually, for Parsons, are an ideational element distinct from ideas).

A central claim is that ideas come in different typologies (Campbell 1998; Schmidt 2008). There are of course different types of ideas - but asserting the existence of distinct typologies does not solve the very problem of precisely defining the original concept. Let us take as an example the distinction between normative and cognitive ideas as articulated by Schmidt (2008). This is attractive and justified in the context of the research questions she was addressing, but in our context it does not solve the problem. First, we fail to see where this distinction originates in terms of conceptual and empirical validity. Cognitive ideas (which, we are warned, may also be causal, generating definitional ambiguity – Schmidt 2008: 306) are value-free action-reaction solutions which draw on interests. Instead, normative ideas are about the values actors attach to political action. By dividing ideas in such manner, three immediate questions arise: first, are normative ideas non-cognitive? In other words, is it the case that values and aspirations are non-cognitive products? Second, if cognitive (or causal) ideas are based on interests and interests depend on values, how come cognitive ideas are non-normative? And finally, why labelling policy solutions and value systems as ideas? What is the added value of this rebranding of well-established and fully valid concepts?

Beyond the fact that distinguishing between typologies of ideas complicates the articulation of a clear definition rather than simplifying it, we are left with an overarching problem that we will address more extensively in the following section: what are the empirical foundations of the typological differentiation? Where is the evidence that such categories exist as distinct in the minds of decision-makers and policy actors? In the ‘ideas school’, in fact, the reference is generally to relatively precise, specific policy ideas that emerge in well-defined historical moments. One of the most typical conceptualizations of ideas consists in seeing them as policy paradigms, with their own internal coherence and tightness (Béland and Cox 2011; Parsons 2007: 128). An in-depth analysis of their logic may well expose their internal flaws - nevertheless they can be quite resilient (Schmidt and Thatcher 2013) and enable or constrain policy change. Indeed, paradigms are relatively impermeable to evidence and logical critical analysis. They are supported by their advocates in ways that are empirically so important as to mark the (retrospective) distinction between policy epochs such as Keynesianism, Monetarism, austerity and so on (Blyth 2013). For this reason, an option, a very popular one, is to capture major change-generating policy ideas in terms of policy paradigms – a concept with a macro dimension (Hall 1989, 1993). According to McNamara (1998), economic policy ideas are the outcome of experience, learning and socialisation – but key is the fact that at the end of this process ideas emerge and are institutionalised as paradigms.

In yet another attempt ideas are defined as discourse. What discourse is depends on the variety of discourse analysis we adopt. In the ‘ideas school’, the main author is Peter Hall. In his The Political Power of Economic Ideas, Hall provides a general definition of discourse, close to the main ideas endorsed by the elites in a given society at a given moment (Hall 1989: 383–386). But this does not tell us how to empirically identify ideas: how do we know that we are encountering, observing, documenting and reporting ‘discourse’ and not something else?

Pouliot (2015), Eleveld (2016) and Jabko (2019) point to discursive practices – showing that practice-tracing is empirically manageable, for example via political ethnography. Interestingly, Pouliot is less concerned with ‘ideas’ than with discursive practices, arguably a sign that he is more interested in taken-for-granted ideational elements than in policy paradigms that are shaped and affectively supported by their advocates (recall that this distinction between often unconscious, taken-for-granted absorption of ideas and affective, explicit support is a key discriminant on ideational explanations, Parsons 2007: 121–122).

Similarly, Jabko, when documenting the practices emerged during the euro area crisis (operationalized as discursive repertoires), departs from static, monadic paradigms. For him, monadic interpretations work as post-factum, ‘retrospective interpretations of economic policy patterns’ (Jabko 2019: 506), that do not respect the granular, ‘faithful images of messy decision making processes’ (Jabko 2019: 495). This very problem of explanations as post-hoc rational reconstructions of highly uncertain and ambiguous decisional processes and outcomes – see the evidence in Anstead (2018), Jabko (2019) Franchino (2020), Kamkhaji and Radaelli (2017), and Van Esch and Swinkels (2015). This macro-macro and post-factum approach disregards both cognitive processes and mechanisms which take place at lower levels of analysis and the process of paradigm formation. It is not sufficiently granular and process-based to capture the endogenous change that such frames undergo while they make their way within the decision making practices. We will say more on this in the following sections.

Proceeding with our quest for definitional clarity in the ideas school, Campbell (1998) provides a way forward with his four-fold typology of ideas defined by two variables: the type of idea (cognitive or normative) and where the ideas operate – in the background of policy debates as assumptions or in the foreground as concepts and theories. Paradigms are cognitive and operate in the background, in contrast to ‘programs’ that are as prescriptive as paradigms but operate in the foreground. Frames are normative and active in the foreground. Public sentiments are normative assumptions about the legitimacy of a solution and appear in the background. It is questionable to see ‘programs’ as ideational elements instead of thinking of them as contents of policy. Sentiments constrain the legitimacy of an idea (such as austerity) but this is the same as saying that all policy ideas are taken to the court of public opinion.

Campbell’s ‘frames’ and ‘paradigms’ seem prima facie to connect with other important work done is the same period on the référentiel (Muller 1995). But for Pierre Muller the référentiel is cognitive rather than ideational: ‘The cognitive approach is not an approach based on ideas’ (Muller 2005: 170). Its thrust is not to add the ‘I’ of ideas to institutions and interests. Its aim is to demonstrate that paradigmatic ideas do not exist without interests and that interests exist only because they find expression in cognitive and normative frames of reference (Muller 2005). After more than twenty years, the notion of paradigm has embraced both cognitive and normative aspects, making Cambpell’s typology only a point of departure rather than a firm destination.

Even with these limitations, Campbell’s typology, as well as the suite of ideational elements suggested by Berman and Parsons amongst others, teaches us an important lesson: that there is not a single definition of idea because we have to unpack this construct and be precise on what we are talking about. It also teaches us that we should not make casual references to ideas: if we speak about paradigms we mean one thing. If we are observing policy beliefs, our empirical object changes, and so on. Yet, to identify a typology is not the same thing as defining the underlying concept. On the empirical side, we can identify cognitive and normative elements that historically define broad ideational shifts. However, there are problems when these shifts are reconstructed ex post (Anstead 2018; Jabko 2019; Kamkhaji and Radaelli 2017; see also Jabko and Schmidt 2021).

Where do ideas come from?

Let us start from a simple proposition: policy ideas at a given moment in time can just ‘be there’, solidified and carried forward by institutions, routines, practices. So, why do we need to talk about the individual actor in ideational explanations? Because even in a constellation where actors are Keynesian slaves of some defunct economist or are led by the institutional context, individuals must have gone through a process of learning what the right or bad idea is and does, and how to make the right idea actionable. More to the point, it is through individual agency, advocacy, translation and performance that we observe ideas in processes of policy change. We cannot trace ideas across time without looking at actors and agency. As Jabko (2019) and Anstead (2018) show, the agent-driven, discursive emergence of ideas and their empirically traceable development is part (perhaps the most important) of the explanation. This granular, process and agent-based approach allows to de-reify ideas and refocus our attention not much on how policy outcomes fit static and often artificial paradigms (typically articulated ex post) but on how actors use ideas to frame policy problems and transform them to conceive new solutions.

The closest we get to micro-foundations in the ‘ideas school’ is a reference to individual actors and to the interplay between structure and agency. Ideas constrain and enable actors because they are part of the structural components of the policy interaction. Actors are both limited and enabled by a given language, discourse, metaphors, prevailing categorisation, linguistic rules and so on. However, in public policy, ideas are relevant only if and when an actor deploys them – to produce change, ideas must be enacted, and activated.

There is already a good deal of political economy knowledge that gets us close to this granular, process and agent-based approach. The Road to Mount Pelerin (Mirowski and Plehwe 2009) connects individuals, networks and organisations engaged in the emergence and promotion of a ‘thought collective’. Economists and Societies (Fourcade 2010) provides a forensic account of the economics profession over a century. The National Origins of Policy Ideas (Campbell and Pedersen 2014) tells us that ideas come from ‘knowledge regimes’ – essentially communities of public research organisations, think tanks inside and outside government, special commissions that connect ideas to policy-makers. The Rise of the Conservative Legal Movement (Teles 2008) traces the rise of conservative ideas at the level of individuals, Law Schools, the profession and American institutions. In policy analysis, John Kingdon’s (1984) Agendas, Alternatives and Public Policies focuses on policy entrepreneurs that push their ideas by coupling policy problems, political attention and available solutions.

Where we need to go further is, following Kingdon, in the analysis of the process whereby ideas are enacted, activated, and advocated to produce change. For sure, ideas cannot exert their causal power immanently, just by existing. For instance, austerity, or to put it better expansionary consolidation, has a rather long intellectual history (Blyth 2013). It existed as a reified ‘idea’ for quite some time, but before material pressures and political and epistemic agency activated it, it failed to shape policy or even inform the debate on macroeconomic coordination in the European Union (and some question it did even after – Franchino 2020). In other words, fiscal austerity is an idea which is considered to be an historical driving force of change but it is defined and conceived in contested ways (Franchino 2020), polysemic (Béland and Cox 2016), empirically multifaceted (Anstead 2018; Brunnermeier et al. 2016) and often nationally determined (Ban 2016; Vail 2020).

These remarks point to three suggestions for this building site. First, ideas may well manifest their more palpable effects at the macro level but, conceptually and empirically, they do happen and originate at lower levels of analysis where their cognitive nature is more evident and could be investigated more objectively. Here Muller’s intuition about the interplay between interests and identities takes us (paradoxically perhaps) towards a non-ideational explanation of why individuals embrace ideas. They do so because ideas (a) cognitively organise a relationship between praxis and meaning; (b) forge identities; and (c) build on interests to re-shape the connection between sector (a professional or economic sector) and the global (Muller 2005: see the conversation between the Anglo-Saxon ‘ideas school’ and Pierre Muller’s approach in Boussaguet et al. 2015; Muller 2015).

Second, to understand where do ideas come from, we need to acknowledge that the existence of a dialectic between ideational and material forces (Marsh 2009; Muller 2005) prevents us from reifying ideas and assign them immanent causal influence over outcomes. Connected to the limitations we identified in the first building site, we argue that to assemble ideational elements as heterogeneous as ‘culture’ or ‘policy paradigms’ complicates the endeavour of identifying, tracing the origins and empirically measuring ideas and their shift over time. The boundary between material and ideational factors falters under the thick constructivist assumption that all is ‘irredeemably ideational’. This, as noted also by post-structuralist scholars (see Eleveld 2016: 74), makes empirical analysis more complicated. And most importantly, the very constitutive notion of an idea is paradoxically forgotten. The Encyclopaedia Britannica, in fact, defines an idea as the ‘conscious representation of some object or process of the external world’. If a dialectic between the material and the ideational exists (Marsh 2009) (and it certainly does), then there is limited leverage in denying one side of the dialectic by ignoring that ideas mirror and represent the external world (or a certain intended/desired status of the external world). Even when ideas become systemic elements of the policy process this is because they are reproduced in individual and collective processes and practices, not because they exert some exogenous, immanent power over actors.

And third, going back to the conclusions of the previous section, since there are different types of ideas, the search for their origin is better organised if we define the ideational element we have in mind – beliefs, or paradigms, or frames and so on. Different types of ideational variables are activated by individuals in diverse ways and impact on policy choices via distinctive processes.

In search of a theorised mechanism: learning and paradigms

The mechanism of explanation is our third building site. Broadly speaking, there are three positions. One is to identify power as key mechanism. Schmidt and Thatcher (2013) argue that neo-liberal ideas are not internally coherent, they do not work, and therefore cannot possibly be embraced because societies have learned about the benefits delivered. The core mechanism – it seems – lies in interests and power. This does not strike us as an ideational mechanism.

The second position is popular in policy analysis: Kingdon (1984) has demonstrated empirically the existence of mechanisms that draw on evolutionary theory and biology. These mechanisms of ‘when an idea’s time comes’ apply to populations rather than individuals – hence his mechanisms do not provide ‘point predictions’ but ‘probabilities of outcomes’ (Baumgartner 2016: 59). The most important of the evolutionary mechanisms is the coupling of streams when a window of opportunity appears. In any case, ideas are filtered by social factors (similar to Campbell’s ‘sentiments’) and institutional structures. We suggest ‘ideation’ as mechanism that captures these processes. Ideations takes us in cognitive territory and our proposals about future work in the building sites. Therefore, we will engage with ideation in the penultimate section.

The third option is learning. In her historical account of monetary unification in the EU, McNamara (1998) sets conditions for the decline and emergence of a new set of beliefs, including repeated failure with the existing orthodoxy and the presence of a viable alternative paradigm supported by a coherent and cohesive epistemic community. Under these conditions, a society learns how to dispose of old ideas and embrace new ones. Hence, the mechanism is one of un-learning and learning at the same time. As Berman shows, this is an argument about learning cast in terms of demand and supply of ideas, or emergence, persistence and change. When old ideas are questioned, there is an opportunity for learning. In fact, the demand for new ideas emerges, but then for learning to occur there has to be a supply of new ideas that a society, at the right historical moment, with the right constellation of actors, is ready to ‘buy’ (Berman 2013: 227).

In this view, it is then learning the mechanism we are looking for - following also Hall’s (1993) seminal ‘social learning’ article on the three levels of policy change. Change can occur at the level of the settings of a policy instrument, change of instrument, or switch to new paradigms. The latter (third-level policy change) is caused by macro-level processes of learning that engage policy-makers and societal actors, hence Hall refers to social learning to explain the third-level policy change.

Here we find that the mechanism of learning connects with the concept of policy paradigm. Although the intuitions about social learning are plausible, the foundation in terms of policy paradigms is questionable. Let us see why, starting for the process through which the mechanism of learning should generate paradigmatic change.

Indeed social learning connects with the concept of policy paradigm. Hall (1989; 1993) provides an ad-hoc (yet empirically rich) argument about acceptance and survival of ideas (in their manifestation as paradigms) in two-parts. Ideas are filtered in a three-dimensional ‘space’ - the three dimensions being economic, political and administrative viability. As for the adoption of paradigms, Hall takes into account both political (the orientation of the governing party) and administrative factors (in particular, three features of a state’s structure: permeability of the administration, power of the central bank, and concentration of macroeconomic management) and the structure of the political discourse (Hall 1989: 383-386). This may well be a good retrospective re-construction of the adoption of Keynesianism and its diffusion, but the theoretical base of paradigms remains fragile unless a mechanism is specified. Berman (2013) observes that the institutionalisation of ideas is the product of learning mechanisms involving individuals, such as entrepreneurs: ‘new ideas do not achieve prominence on their own but must by championed by carriers or entrepreneurs, individuals or groups capable of persuading others to reconsider they ways they think and act’ (ibid.: 228).

One problem with this reasoning is that there is no strong theory of paradigms emergence in the sociology of science. The notion of policy paradigms is influenced by the work on scientific paradigms done by Thomas Kuhn. And Kuhn was very much indebted to Ludwig Fleck’s sociology of science, based on the notion of collectives of thought. Fleck’s history of the origin and development of scientific facts was a rhetorical reconstruction to defend scientific positions about immunisation that were later proved wrong. Actually, these positions were already shown to be wrong when Fleck was writing - his mystification of the scientific debate was instrumental to the defence of a position that was becoming weaker. Because neither Kuhn nor the other thinkers of scientific paradigms ever bothered to engage with the substance of the argument made by Fleck, the notion that scientific paradigms are ideational constructs that have nothing or very little to do with naturalism and facts is questionable. Borrowing this notion into political science by transforming scientific paradigms into ‘policy paradigms’ begs the question of providing an endogenous political science theory of paradigms or admitting the serious limitations of Kuhn’s argument. If we characterise actors’ ideas as paradigms we bracket away the syncretic character of practices of institutional change (Jabko 2019).

A recent iteration of Hall’s policy paradigm framework can be found in Matthijs and Blyth (2018) in their study of the ideational determinants of the economic governance reforms adopted by the European Union as a result of the sovereign debt crisis. They adopt a goodness-of-fit approach between the observed macro-level outcomes and a somewhat time-invariant and strictly fenced policy paradigm (austerity/expansionary consolidation). By doing so, they overemphasise the constraining power of ideas while failing to dig into lower levels of analysis and empirical granularity where classical crisis management, solution search and policy learning mechanisms were at play. Decisional arenas characterised by crisis-driven functional pushes and Knightian uncertainty generate political and epistemic ambiguity. And the empirical result of this ambiguity is the permeability and malleability of policy paradigms, both at the individual and collective levels, rather than their intervening, exogenous causal power (Jabko 2019; Van Esch and Swinkels 2015). Ideas, intended as paradigms suited to address uncertainty rather than variable cognitive maps, repertoires or signifiers, lose causal power at early and peak stages of existential crises to regain it after the crisis has been averted. Then, when reflexive sense making and inferential learning step in and the new dominant (and possibly piecemeal) paradigm needs renewed legitimacy, it gets adjusted and/or transformed to embed (sometimes with fatigue) the new normal within its boundaries.

Wrapping up then, the valuable theoretical intuitions of the ‘ideas school’ are about the process described by Kingdon and the observations on learning. At the same time, the question remains whether actors really learn and un-learn paradigms. The main problems we found lie in the concept of policy paradigm. This is mainly because Kuhnian paradigms have been imported into political science without consideration of the fact that they represent an ex-post systematisation or a retrospective sense making exercise of practices and processes that in real time work in a granular, sometimes chaotic and polysemic, fashion.

Ideational or cognitive analysis?

We now address the question whether a consideration of cognition and its processes would make the ideational explanation more robust. We find that explanation may become more robust, but at the cost of dropping ‘ideas’ – which makes this a building site of paradoxes.

But let us go step by step. The backbone of a possible cognitive analysis is the following: Actors first infer from experience and draw lessons, and that when this learning process occurs (‘in their minds’ so to speak), behaviour may change, and in turn different behaviour of policy makers can produce policy change. The learning process ‘in the mind(s)’ can be imperfect, distorted, biased, conditioned by economic resources, interests, institutional constraints or else – in a word: bounded – and affect ideas, their emergence and functioning, thereof. Key is the following logic: first something changes in the world out there (imagine repeated failure of an economic policy paradigm), then lessons are drawn from this experience (learning ‘in the mind’, hence we are exactly in cognitive territory) and finally this causes change through a collective interactive and discursive process where an idea (more precisely, a set of ideas) work as focal point.

A different option, this time for fully cognitive rather than simply learning-informed ideational analysis, is provided by Pierre Muller, in turn often research partner with Bruno Jobert (Jobert and Muller 1987). Whilst the Anglo-Saxon ‘ideas school’ draws on hypothesised (not necessarily proved with experiments and other empirical methods, see below) cognitive processes about learning to support claims about the political power of ideas, Muller contrasts ideational and cognitive explanations, arguing for the latter.

We find two core propositions in Pierre Muller (Muller 2000; 2005). First, the explanation of policy change must combine structures (durable, resilient systems of interaction which exist independently of actors) and actors, who always have a degree of autonomy from constraints in policy-making processes. Second, the cognitive and normative frames are both the empirical manifestation of structural constraints and the outcome of the action of actors on the meanings of empirical facts and events (le travail sur le sens effectuè par les acteurs Muller 2015: 158).

Hence, ideas do not exist independently from the actors that create and carry them through the world with their behaviour and their ‘work on meanings’. Change occurs when the cognitive and normative frames in a sector are no longer in phase with global frames (référentiels). The mediators are actors that re-align the meaning of norms and core beliefs between a policy subsystem and the wider world (Muller 2000). This process can be conflictual – hence mediation may look very different from compromise.

Mediators, and indeed all actors, however, cannot create the world by simply operating on fluid, free-floating meanings. The world, understood as a set of structures, exists independently from actors. But in order to operate on the world, actors need representations of the world. In turn, representations that allow actors to intervene on the world are both representations of ‘what is’ (diagnostics, cause-effect relations, analytical explanations) and representations of ‘what ought to be’ (that is, values and norms). This dual dimension is captured by the key concept of référentiel: the latter provides both knowledge of the world and direction on how to operate in the world.

There is a connection between cognitive and strategic activities. Those actors who work on making sense of the world with cognitive constructs are not doing this independently of the construction of collective identities and interests. The work of the meditators is not exclusively an interpretive work, an effort in understanding the world with the correct ideas so to speak. The référentiel has a cognitive positive dimension (the world as is), a normative dimension (the world as it ought to be) and a third strategic dimension too - because it instantiates or manifests the identities and interests of particular groups of actors.

Concluding on this building site, Muller’s référentiel is intrinsically different from a generic reference to ideas for two reasons. First, it embodies a specific relation between actors and social structures, where autonomy and dependence are intertwined. The référentiel is not true or false in general. It is true in the sense that is in phase with the change that is happening in the world. But the structure (‘reality’ so to speak) exists in social action only in the moment in which it is spoken or narrated by actors. Hence the mediators have a significant degree of autonomy in their work of sense-building and meaning-construction. Second, when actors engage in the process of sense-building (construction du sense), we should not think of this process as ‘ideational’ alone. For Muller, public policy making is essentially about the sense-driven construction of a relationship with the world (Muller 1995). Yet, this process is not separated by the identities and interests of these actors, actually it would not take place if the mediators did not have a strategic dimension in mind – and this is the necessary condition for their autonomy from structures.

Reconstructing ideational analysis

Let us now carry out constructive work inside the four building sites. We discuss a possible conceptual and practical integration to existing scholarship on: definitional and micro-foundational aspects, micro-macro linkages, mechanistic causation and cognitive analysis. Most importantly, we link our proposals for explanatory improvements with existing ideational contributions, both within and outside this Symposium.

The key argument of the integrations and improvements we suggest lies in a ‘cross-fertilisation path’. Broadening the conceptual and analytical toolkit of political science through borrowing (i.e. borrowing concepts from other disciplines) allow us to bring fresh air and possible solutions in all of the four building sites. Given that ideas, even after years of relentless research, have remained broadly an elusive concept in political science we suggest epistemic humility, taking into account the progress made in other neighbouring social sciences.

Obviously, we are not alone. Similar cross-fertilisation paths are exemplified by recent research. Starting with the first building site of definitional aspects, Van Esch (2014) and Van Esch and Swinkels (2015) draw on cognitive mapping, inspired by political psychology. Their answer to the question ‘what are ideas made up of?’ is cognitive, granular, empirically observable, and testable against behaviour; Jabko and Schmidt in this collection (2021) investigate the promise and practice of cultural repertoires to operationalise ideational and discursive elements. Always in this Symposium, Cino Pagliarello (2021) develops Wittengstein’s work. Her approach to polysemy endogenizes the ambiguity of ideas. Ambiguity, we have seen, is a problem of exiting ideational explains. Hence, Cino Pagliarello argues, we must face it up-front and include it in ideational explanations in suitable ways, with the assistance of the insights of sociology of science. Trein and Vagionaki (2021) draw on the behavioural economics and crisis management literature to embed issue salience and polarisation in mechanisms of power and policy-oriented learning.

When it comes to micro-foundations, a dialogue with economics seems promising. Economics is the social science which more heavily employs the concept of micro-foundations and methodically informs macro-level research to micro evidence and models of behaviour. However, we are not a-critical in looking at economics for inspiration. In fact, mainstream economics still struggles to detach itself from an empirically wrong model of behaviour and decision making that assumes full rationality of the individual – with exceptions in the field of experimental economics and mechanism design informed by bounded rationality (Glazer and Rubinstein 2016). Political science and policy analysis, since the 1950s, can profit from much stronger, realistic and empirically valid assumptions about human behaviour. Pioneered by Herbert Simon, the framework of bounded rationality (Simon 1957) is foundational for political science (Jones 2017). Yet, notwithstanding a large consensus on bounded rationality being a highly valid model of human behaviour, micro-foundational thinking is often neglected in ideational research.

Thus, we have a discipline, mainstream economics, that takes micro-foundations seriously but largely draws on an unrealistic model of individual behaviour. On the other hand, we have a discipline that has embraced an empirically valid model of behaviour (bounded rationality) but (in the ideational scholarship we reviewed) disregards micro-foundations. Consequently, it often builds causal arguments by observing aggregate behaviour rather than trying to understand how the parts get together to lead to collective outcomes. To borrow from disciplines like behavioural economics and cognitive psychology seems a reasonable way forward then, most of all if we are interested in micro-founded mechanisms.

As for the building site of mechanisms, in this collection Lesch and Millar (2021) take insights from political psychology and behavioural economics, and investigate across different levels of analysis a mechanism of ‘bounded ideational emulation’ which has been at play in several domains of Canadian tax policy. For Cino Pagliarello (2021), polysemy as cognitive mechanism explains the stability of policy choices and how policy entrepreneurship is enabled in a context of ambiguity.

Our own additional suggestion is to re-construct mechanisms by considering the concept of ‘solution search’. Following Kingdon and his roots in biological selection and randomness (see the compelling narrative of how Agendas, Alternatives and Public Policy came about in Baumgartner 2016), Bryan Jones (2017: 69) argues that the connection between problem definition and solution search in policy making is attention-driven and not sequential. This notion is indeed the bedrock of popular policy process theories, especially the multiple streams framework. The search for solutions at the individual level is informed by the very same heuristic decision rules observed at the aggregate level.

Applying this insight to our topic, we argue that ideas as such are not that different from solutions as theorised by policy process scholars (Kingdon 1984) - above all when ideas take the form of causal beliefs. First of all, both ideas and solutions may seem to exist immanently in the public sphere, i.e. regardless of agents creating and carrying them. But even if an idea or a solution gets bigger than their creators or advocates (Keynesianism being a classic case in point) they could exert little direct influence on outcomes without individual or collective agents activating them in the context of public decision making and problem solving. Second, if solution search at the collective level mimics individual level dynamics (which embed cognitive biases and heuristics – and bounded rationality) then the process of creation and utilisation of ideas cannot be totally detached from what happens at the micro, individual level. Third, policy analysis has detached itself from a sequential model of ‘solution search following problem definition’. This insight, which through individual cognitive biases reproduced at institutional level leads to punctuated policies (Jones 2017), is of great importance also for ideational research. This is because, as far as solutions mimic ideas, an ideational explanation ecologically links exogenous influences arising out of the material to ideational emergence, suggesting that the two dimensions are entangled - if not in a dialectic relationship.

These propositions and empirical examples, we argue, should invite ideational researchers to pause before addressing the macro effects of a given idea and reason more carefully, on a case by case basis, on the generative process of that very idea. In its original, raw form every idea is a cognitive product that is generated at the micro, individual level. In this view, ideas are cognitive constructs that are generated through individual mental processes of representation. Ideas mirror and represent reality in abstract mental terms. Hence, if the generative process of an idea (which we term ‘ideation’, see below) is entangled with sense making practices that take place at the micro- and meso- cognitive levels within the problem space. Ideas will be shaped and redefined by these practices and therefore cannot and should not be considered as static, invariant and impermeable explanatory factors.

And yet, ideas do not come from anywhere. They are the product of a process we call ideation. Although such a process, in policy and political landscapes, can and does scale up at the collective decision making level, its roots are strongly wired at the individual level – which, to repeat, is a plausible behavioural foundation of macro dynamics. This is how we would integrate some of the methodological precision of economics with the empirically robust model of Simon and, today, cognitive-behavioural public policy (John 2018; Jones 2017). As already noted, recent empirical research in political science comes to the rescue in highlighting how ideation takes place and evolves at the individual level before tackling its effects at the macro level and how cognitive models shape ideas rather than being shaped by them. Van Esch and Swinkels (2015) prove that individual political leaders are not simply exposed to ideas intended as immanent, time-invariant elements (somehow as earlier scholarship thought solution search was taking place) but, pushed by exogenous inputs and feedbacks, they go actually through novel processes of ideation where beliefs are re-created and contextually shaped. In Van Esch’s studies this phenomenon is evidenced, for instance, by a counterintuitive mix of Keynesian and ordo-liberal elements emerging at the same in crisis policy responses, something that largely defies ideational expectations about policy change. This is because, exposed to extreme, existential environmental pressures both ordo-liberalism and Keynesianism went through a process of actor-centered re-ideation, affected by cognitive dynamics and biases where ideas, as intended by ideational scholars, are largely epiphenomenal and context-driven.

Moving from individual leaders to the collective decision making setting, Kamkhaji and Radaelli (2017) illustrate a model of contingent learning whereby policy change occurs before sense making. This process of contingent learning, akin to a process of re-ideation, is not influenced by economic paradigms. Instead, it recreates and reshapes paradigms according to cognitive factors and biases triggered by material constraints and feedback effects - while working as generic reference frame to dynamically articulate interests.

Turning to cognition, we draw on two fields to model the generative process of ideas (that is, ideation). The first is information processing, nowadays the dominant framework in cognitive sciences (see Lachman et al. 1979 for a classic treatment). Information processing puts individual actors centre-stage but it does so differently from behavioural psychology whereby agents are often modelled as passive conditioner responders. In information processing individuals (and groups of individuals) are in a dialectic relationship with the environment. Hence, drawing a parallel with ideational research, we argue that information processing captures the key features of processes of ideation. This is an endogenous cognitive process which involves ‘collection, storage, interpretation, understanding, and use of environmental or internal information’ (Lachman et al. 1979: 7). Moreover, information processing is in line with the bounded rationality framework: ‘information-processing psychologists know that they cannot rely on rational and intuitive tools alone, precisely because so many cognitive processes go on outside of people’s awareness’ (Lachman et al. 1979: 9).

This consideration, which points towards cognitive biases, brings us to the second field. Here we echo Bryan Jones and argue that among all the cognitive biases documented in the experimental literature and that can influence collective decision making, attentional and groupthink biases are poised to have more leverage in ideational explanations – as they are wired and reproduced into institutional dynamics. The reason for this claim is that since ‘[a]ttention is selective, hard-wired into human cognitive architecture, and requires serial processing of information’ (Jones 2017: 70) both the generative process of ideas and their institutional embedding, as well as the ability and willingness of an actor to adopt an idea, will be crucially affected by attention and availability biases (more than they are affected by the nature of the idea itself – Berman 2013: 228).

Conclusions

The ‘ideas school’ is a successful research programme. But it should not mean it has to be considered static. Indeed, a constructive approach is to explore where problems lie, and contribute to a stronger research agenda by suggesting ways forward. Our conclusions come with caveats: some of our concerns about the origin of ideas may not be relevant to those who think exclusively in holistic terms. We presented examples and suggestions that are not exhaustive of the options available in the four building sites – another possibility is to borrow from sociology of sciences and professions (Ban 2015 and 2016; Farrell and Quiggin 2017). Another caveat is that the ‘ideas school’ is only one of the existing ways to think about ideational politics, although it is the most prominent approach.

With this caveat, we have exposed the difficulties in taking ideas as the main foundations of an explanatory approach to policy change. We have found complications in pinning down ideas as foundational concept. The way forward is to consider other, more precise ideational elements, like beliefs, and/or to turn to other disciplines for foundational concepts – such as cognitive psychology and sociology. The notion of paradigms, the workhorse of the school, is problematic: if we import it from epistemology and philosophy of science we are taking some serious risks. But, if we search for its proper political science foundation, we find little.

Next are the building sites concerned with the origins of ideas, what we called micro-foundations (although the problem is more general), and with the mechanisms. What is the way forward, then? Either we say that we do not care about micro-foundations, and provide other accounts of where do ideas come from, or, yet again, it may be useful to first distil analytical lessons from the rich historiographic trends in political economy (e.g. Fourcade 2010; Mirowski and Plehwe 2009), and, second, to look at what other disciplines say about the causal link between ideational elements and change.

On the fourth building site, cognitive-political psychology and strands of economics (behavioural and experimental) are ripe for interdisciplinary research. Our own take is more cognitive than ideational. We have suggested the concept of ideation as process. We have illustrated how it connects individual action to higher levels of analysis. Future research should develop this cognitive approach further. The contributions to the Symposium show a range of solutions that complement and add to our propositions in novel and diverse ways.

Acknowledgements

An early draft of this article was delivered to the European Consortium for Political Research, General Conference, Hamburg, 22–25 August 2018, to the LSE workshop on “The EU paradigm in crisis? Ideas, leadership and the search of a new référentiel for Europe” 20 May 2019, organised by Marina Cino Pagliarello, and to a panel on ideational politics at the International Conference on Public Policy (ICPP), Montreal, 25–29 June 2019. We wish to thank the participants to these events – and in particular our ICPP panel co-chair, Femke van Esch. We gratefully acknowledge the detailed comments provided by Vincent Caby, Claire Dunlop, Nicolas Jabko and Pierre Muller. We benefited from the remarks, observations and critiques of two anonymous reviewers. Finally, we wish to say thank you to Klaus Goetz, who has piloted this Symposium at West European Politics. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Jonathan C. Kamkhaji

Jonathan C. Kamkhaji is a Post-Doctoral Research Fellow at the Department of Management, Economics and Industrial Engineering, Politecnico di Milano. His research interests include public policy, regulatory governance, political economy, European integration and the policy and politics of governance and regulatory indicators. He worked also as a policy consultant for the World Bank and has published research in, among others, Journal of European Public Policy, Regulation & Governance and Policy Studies Journal. Email: [email protected]

Claudio M. Radaelli

Claudio M. Radaelli is Professor of Comparative Public Policy at the School of Transnational Governance, EUI, Florence, on long-term leave from the School of Public Policy, University College London. His research interests include the role of expertise in public policy, European integration, policy instruments, regulation, and theories of the policy process. He is PI on the ERC-funded project Procedural Tools for Effective Governance (Protego). Email: [email protected]