Skip to navigation – Site map

HomeIssues6:2Expanding the horizon of policy l...

Expanding the horizon of policy learning

Claudio M. Radaelli and Bishoy L. Zaki

Abstract

Policy learning is an established conceptual and analytical framework to study the policy process. Its progress as a research agenda hinges on solid micro-foundations, its integration with other policy process theories, the consideration of both positive and negative cases (of learning and change) from diverse continents and countries, and the clarification of the causal pathways connecting individual learning to decisions and change. We sketch these problematic areas, showing where policy learning research has encountered problems, and illustrate how the articles of this special issue contribute to their solution.

Top of page

Full text

The early drafts of the papers collected in this Special Issue were discussed at the International Conference on Public Policy 6, Toronto, Canada, 27-29 June 2023, organized by the International Public Policy Association (IPPA). We are grateful to the participants for their comments.

1. Policy learning and policy process research

1Why, when and how do policies change or remain stable? These questions are some of the most fundamental and mesmerizing for public policy scholarship. In the formative years of policy process research, the phenomena of policy change (or the lack thereof) were often approached with the lens of “powering”, where factions in the public sphere overpower one another to impose or enforce their policy preferences. However, over the decades, we, policy scholars, have developed other approaches that deliver more fine-grained insights into the dynamics of policy change and stability. We can think of the Multiple Streams Framework (Kingdon, 1984; Zahariadis, 2014), the Advocacy Coalitions Framework (Sabatier & Jenkins-Smith, 1993; Dudley & Richardson, 1999), and the Institutional Analysis and Development framework, among others (Ostrom, 2007; Heikkila & Andersson, 2018). All these lenses on the policy process expand the analytical horizon beyond powering.

2In this tradition of scholarship, policy learning has evolved from some of the field’s intellectual founders such as Dewey (1946), Lasswell (1951) and Deutsch (1963) (see Dunlop, Radaelli, & Trein, 2018). The learning lens on the policy process posited that, while powering is an essential component of policymaking, policy actors also “puzzle” about policy problems. Encompassing a broad conceptual palette, at its core, policy learning is essentially understood as a process where policy actors seek and process information and knowledge about policy issues, aiming to update their understanding of how they can be addressed. This – the argument goes on - leads to cognitive updates such as the confirmation or revision of policy-related beliefs, or behavioural products, such as changes in policy, or policy advocacy behaviours (Dunlop & Radaelli, 2013; Zaki, Wayenberg, & George, 2022).

3And yet… at the most fundamental level, the premise of the learning approach was not to simply serve an alternative or a ‘rival’ explanation to interpretations taking a lens of power dynamics to policymaking. Rather, at its heart, policy learning argues that even within the omnipresent processes of powering within policymaking, puzzling also exists, not merely as single dimension, but as a fundamental process that drives all policy actors’ behaviour, even their use of power in policymaking. This elementary perspective has positioned policy learning at the nucleus of policymaking – perhaps becoming a coherent ontological approach to our understanding of the policy process (Kamkhaji, 2022; Zaki, 2024). In a sense, this perspective implies that the more we understand when, why, and how policy actors learn about policy problems, the more we can successfully explain, and perhaps to some extent predict policy decisions and outcomes (Zaki, 2023).

4The growth of policy learning literature has been somewhat consistent with the purpose for which it was conceived, that is being a potent explanatory lens on policymaking, particularly focused on analysing and understanding policy change. Accordingly, we have witnessed the development of various learning-based theoretical approaches and descriptors, predominantly with varieties of policy and belief change at their crosshairs. For example, in terms of learning types, we have instrumental learning, focused on changes in policy instruments’ calibration, social learning, centred on updated understandings of the social construction of policy issues and thus changes of policy paradigms, as well as political learning - on updated understandings of political advocacy strategies (e.g., Heclo, 1974; Bennett & Howlett, 1992; May, 1992; Hall, 1993). There is also a set of learning mechanisms, where causal processes underlying learning induced belief and policy change are explored: Inferential, where learning occurs through critical and deliberate reflection on priors, or contingent, where learning occurs through rapidly drawing on cue-outcome associations in a bid to avoid imminent harm (Kamkhaji & Radaelli, 2017). Different modes or “patterns” by which learning as a cybernetic problem-solving process occur were also developed including hierarchal, bargaining-oriented, epistemic, and reflexive (Dunlop & Radaelli, 2013). This in addition to conceptual frameworks addressing how learning aggregates across the individual and collective levels (e.g., Heikkila & Gerlak, 2013), and how learning processes are deliberately strategized and governed towards achieving pre-determined policymaking outcomes (e.g., Zaki, 2024). Consistently, these advances and others were mobilized to advance understandings of policymaking and explain policy change (Bandelow et al., 2019; Dunlop & Radaelli, 2017; Moyson, Scholten, & Weible, 2017; Quaglia & Verdun, 2023).

2. Moving the frontier of learning research

5The fundamental nature of the policy learning lens points to its promise as an inclusive, theoretically and empirically global lens, not only able to strengthen existing theoretical approaches but also to generate its own original insights into the policy process. However, while theoretical and conceptual developments in policy learning literature draw on other literatures (albeit not as frequently), the embedding of policy learning within other approaches in the policy process literature remains somewhat limited, not to mention the relatively scant intersections with disciplines like psychology that have studied cognition and behaviour ever since their creation. Also, the literature maintains a substantive empirical focus on Anglo-Saxon and developed country contexts, where sparing a handful of contributions, very little attention is dedicated to developing countries and other politico-administrative traditions (See Zaki, Wayenberg & George, 2022). Finally, the whole causality of learning, from the individual assimilation of new information to changes in beliefs, the impact on individual behaviour and organizational behaviour, down to the effect of learning processes on policy decisions and change is sketched rather than being forensically theorized and empirically examined. In short, both conceptually and empirically, there are some key steps in moving this literature towards more robust causal conceptualizations and empirical knowledge. This is the background that motivates our Special Issue.

3. The contributions

6In this Special Issue, we shed light on problems that seem to us fundamental for the development of our knowledge about policy learning. Let us examine them one by one, explaining how the individual articles tackle the problems and present significant improvements. At the outset, there is the very ontological status of learning we discussed earlier on. The whole notion of adopting learning as lens on the policy process is to go beyond the narrow vision of power as the only mechanism that drives policy decisions. Although for sure learning involves aspects of powering, most of the studies in our field have embraced an ideational perspective on decisions and change. That is to say that constellations of actors make sense of the world, including making sense of their interests, via their own sets of beliefs. In turn, the specific assemblage of beliefs that characterises an actor can change on the basis of persuasion, evidence, and socialization. This belief-driven ontology, however, does not tell us exactly what beliefs should be looking at when observing processes of learning. Also, the ideational literature (Kamkhaji and Radaelli, 2022 for a critique) has developed around positive cases, that is, empirical examples of why and how ‘ideas matter’. This is somehow physiological – if one wants to explain the role of a variable, one selects cases where the variable is supposed to operate: researchers choose cases where beliefs change, and these changes seem correlated to policy change. And for this reason, we can conclude that the process of learning has taken place. But this also means censoring the dependent variable – in the end, not all ideas and not all changes of beliefs produce change.

7Addressing that, in this special issue, Cino Pagliarello and Tubakovic (2024) help us in dissecting the rather fuzzy notion of ‘ideas’ by exploring the world of polysemic beliefs – those that allow for more than one meaning. When, within a belief system, we observe a proliferation of meanings, it is possible for different actors to embrace the same idea, even if this is done for totally different reasons. The policy world is replete with polysemy. Interestingly, the two authors argue that not all ambiguous ideas are polysemic. Ambiguity can just mean vagueness and fuzziness. A polysemic idea is, yes, ambivalent, but in the sense of hosting two or more meanings that are clearly understood by alternative coalitions. By adopting this polysemic approach, with their research design Cino Pagliarello and Tubakovic create variation in the dependent variable. Their article explains why, in the same context of the European Union, we find a case of agreement and change induced by polysemy as well as a case of failure. This allows them to pin down the conditions for policy learning via polysemy.

8And yet, if we go granular and observe individuals instead of coalitions of actors, what do we see in the psychology of learning? Kamkhaji and Radaelli (2022) note that policy scholars and political scientists have theorized models of learning that are not anchored in the findings of disciplines entirely dedicated to the study of cognition, information processing, and behaviour. Here is where Moyson (2024) comes in, with his analysis of the individual psychology of learning. Namely, Moyson explores the role of the self via egocentrism and self-esteem. Empirically, this article draws on data on the Belgian network industries. Conceptually, the message has a wide reach for scholars of learning: the assimilation of information (key to processes of individual learning) is mediated by psychological variables. Egocentrism, that is, the perception of being unique, hampers information assimilation. Thus, although learning is strongly influenced by information processing, the psychology of individuals can block the learning process at the very beginning.

9Conceptually, these two articles invite us to reflect in depth on the causal relationship between the very origins of learning in individual cognition and behavior, learning as mechanism, and the outcomes of learning (change in policies, or lack of change). How does the causal sequence pan out, exactly? Going back for a minute to Cino Pagliarello and Tubakovic’s negative case, actors can learn by changing their beliefs, but they may as well learn how to reinforce their beliefs - because of motivated reasoning and, more generally, bounded rationality. In these circumstances, the policy preferences of actors do not change, so their cognitive changes do not impact policy decisions.

10Plümer (2024) embarks on the challenging task of explaining the causal sequence by taking on board three dimensions: the cognitive dimension we just mentioned, the behavioural dimension (that is, how actors behave in response to cognitive changes), and the social dimension of learning. The very concept of ‘change’ is decomposed into belief change, policy preference change, and output change. This clarity allows her to highlight four causal trajectories: (a) beliefs change, but preferences and output do not; (b) beliefs and preferences change, but this is not sufficient to change policy – for example because of institutional obstacles to change; (c) beliefs, preferences and policy output change, but not in the same direction, so this is non-congruent change; (d) all three variables change, and in the same direction, leading to policy change congruent with the causal mechanism theorized by learning scholars. Plümer’s contribution sheds light on the dynamics of learning and change. Importantly, it explains why it is so difficult to generate learning mechanisms that produce decisions congruent with what has been learned.

11This difficulty has led governments to deliberately design organizations and governance architectures that facilitate change. On this front, Kim, Heikkila and Wellstead (2024) present their findings on data-based policy innovation labs in the USA – a type of organization that has attracted considerable political attention across the world, for its potential to support innovation and growth. Like Plümer, our three authors distinguish between learning-as-process and the products of learning. Although there is evidence of learning within the labs, often triggered by interesting human-design approaches, yet again the authors find that learning outcomes, or ‘products’, are rarer. Nevertheless, the findings show exactly how learning happens within the labs and the strategies to disseminate new evidence, information and results to the broader policy ecosystem that interacts with the labs. These are very useful findings for governments and international organizations currently experimenting with the idea of data-based policy labs.

12This emphasis on spreading learning is echoed by Zhang’s article (2024) on Ethiopia’s industrial park programme. This time the learning transfer takes place at the international level, between China and Ethiopia. This article fills a gap in the literature on policy transfer – a literature that has not looked enough into cases other than the classic transfer among the so-called Global North or between North and South. Zhang shows how transfer is a bumpy process of learning, with the Ethiopian policymakers initially looking for lessons in the wrong places. The findings also point to the agency of a transnational policy network (well beyond the dyad China-Ethiopia) and the role of a policy entrepreneur in leading by example and building networks and a coalition in support of the park. Networks and policy entrepreneurs are instrumental in the institutionalization of policy ideas – thus connecting the ideational dimension of learning with the agency of actors.

13The reference to entrepreneurs and Zhang’s demolition of the stereotype of learning as linear process invite us into the multiple streams framework territory of Möck and Feindt’s article (2024). There, the authors start by contrasting the image of learning as a linear process led by the production of knowledge/evidence and the Bayesian updating of preferences to the organized anarchy of the multiple streams. They ask provocatively if there is room for learning in organized anarchies: where does learning happen, and can it happen? By empirically observing the decisions around the German wolf-livestock conflict case in Germany, they theorize and trace learning within each stream (policy, problems, and politics) and across the streams. Their contribution has double merit. Once, for the identification of varieties of learning, with a precise conceptualization of how learning pans out in each stream and across streams. Twice, for creating a dialogue between learning theory and one of the most successful theoretical lenses on the policy process. Up until now, the integration of learning with theories of the policy process has been mostly in the region of the advocacy coalitions framework.

14Having sketched the contributions of this special issue, what comes next for future policy learning research? While various avenues for future research exist, perhaps one of the most promising is still to further push forward the frontiers of policy learning research towards more theoretical and empirical diversity. This can proceed in several ways, including carving out the conceptual space for policy learning within and across theories of the policy process, leveraging the powering and puzzling lenses on the policy process as complementary rather than mutually exclusive analytical perspectives, and continuing to deploy policy learning-based analyses in novel and understudied empirical contexts.

Top of page

References

Electronic reference

Claudio M. Radaelli and Bishoy L. Zaki, Expanding the horizon of policy learningInternational Review of Public Policy [Online], 6:2 | 2024, Online since 24 February 2025, connection on 15 March 2025. URL: http://journals.openedition.org/irpp/4875; DOI: https://doi.org/10.4000/13dwd

Top of page

About the authors

Claudio M. Radaelli

School of Transnational Governance, European University Institute, Florence Italy
University College London, School of Public Policy, London, UK
Claudio.radaelli@eui.eu

By this author

Bishoy L. Zaki

Ghent University, Belgium
bishoy.zaki@ugent.be

By this author

Top of page

Copyright

CC-BY-4.0

The text only may be used under licence CC BY 4.0. All other elements (illustrations, imported files) are “All rights reserved”, unless otherwise stated.

Top of page
Search OpenEdition Search

You will be redirected to OpenEdition Search