Story 1: McGarity and the bureaucrat

Radaelli’s first case examined the bureaucrat as policy actor, drawing on Thomas McGarity’s study of regulatory analysis in US federal agencies. As Radaelli noted, traditional models assume bureaucrats simply follow rules and strive for efficiency – what Downs (Reference Downs1967) critiqued as an oversimplification. Bureaucracy, Radaelli emphasised following Downs, is not monolithic but a complex political system populated by engineers, economists, and other experts, each with different priorities and professional commitments.

Reinventing Rationality: The Role of Regulatory Analysis in the Federal Bureaucracy (McGarity Reference McGarity1991) explores how ‘rationality’ in government is supported through the US federal bureaucracy. The book is a direct criticism of the evidence-based policy agenda, which McGarity argues simply assumes that bureaucrats want to learn from science. It examines how US regulatory agencies, including the Environmental Protection Agency (EPA), used quantitative analysis within the institutional rivalries of the Reagan era to arrive at key policy decisions. After presenting the case studies, McGarity concludes that a ‘comprehensive analytical rationality ideal’ could not have been achieved in the real-world decision-making of the agencies examined.

His analysis of regulatory analysis by the federal bureaucracy helps us to understand why. In practice, values conflict. Different types of bureaucrats, such as economists and environmental engineers, have competing views of what constitutes ‘effective’ policy. While the environmental engineer might be concerned with cleaning up the environment, the economist will focus on efficient solutions through market-based mechanisms. Crucially, regulatory analysts come to the enterprise not as objective analysts but equally shaped by the sensibilities and preconceptions of their disciplinary background and training. When seeking to implement ideas around ‘goal-ranking’, they behave as political actors guided by their own intellectual and social heritage.

Advocates of analytical capacity-building, however, argue that institutional design can overcome these tensions, particularly through hybrid or team models that institutionalise deliberation and mutual critique. These models, they contend, enable creative tension and pluralistic input to improve outcomes. Yet McGarity’s findings suggest that such models often struggle against the grain of entrenched professional hierarchies and normative divisions. The question remains: are these models viable reforms, or do they merely layer procedural complexity atop deeper political conflict?

Reinventing Rationality suggests that creative decision-making can harness this tension. McGarity outlines five distinct models for integrating policy analysts into bureaucratic decision-making: hierarchical, team, adversarial, hybrid, and outside adviser. The hierarchical model limits analysts to gathering information and supporting pre-determined decisions, offering little room for shaping outcomes. In contrast, the team model, prevalent in federal agencies, establishes a dedicated analysis office and encourages collaboration between analysts and decision-makers. The outside adviser model brings in external experts who provide independent analysis, though their impact depends on how open agency leadership is to their input. The adversarial model promotes creative tension by pitting analysts against technical experts to provoke informed debate. Finally, the hybrid model, used by agencies like the EPA, combines adversarial and team elements to harness the strengths of both approaches at different stages. McGarity stresses that no single model works universally; each agency’s discretion, policy complexity, and analytical capacity must guide which model is most appropriate. Therefore, the integration of policy analysis must remain adaptive to institutional contexts and democratic values.

Ideology can masquerade as analysis, raising important questions about accountability and the possibility of objective-looking analyses that constrict decision-makers’ effective range of choice. In the context of politically accountable institutions, a ‘comprehensive analytical rationality’ will necessarily always be in some degree of tension with a ‘techno-bureaucratic rationality’. McGarity’s perspective invites us to think critically about the inner workings of bureaucracies as spaces shaped by diverse expectations and contested definitions of what constitutes good policy. Bureaucracies are full of competing values, and we cannot assume any one model will always yield better results. Indeed, even the adversarial model could turn the policy process into a battleground of constant conflict, undermining deliberative problem-solving altogether.

The challenge is to design institutions that balance internal diversity with coherence. However, whether political scientists are making headway remains unclear. The field has produced valuable conceptual frameworks, but empirical progress has been uneven at best. While venues such as European Political Science offer space for conceptual innovation, they are not oriented toward testing institutional reforms. This gap between theoretical development and practical application remains a persistent challenge for the field.

From the perspective of the policy learning literature, McGarity’s account demonstrates the limits of instrumental learning within bureaucracies while revealing the persistence of political learning. The economist and the engineer are both learning, but they are doing so through the lens of their professional identities and disciplinary commitments. What appears as a failure of learning is better understood as learning shaped by professional identities and organisational hierarchies. The institutional design challenge is to create structures that channel this political dimension.

Story 2: Gilardi and the politician

Radaelli’s second case turned to the politician, drawing on Fabrizio Gilardi’s article Who Learns from What in Policy Diffusion Processes? (Reference Gilardi2010). Gilardi’s research focuses on the role of learning in how policies spread across different governments. Gilardi argues that policy diffusion involves selective learning rather than straightforward copying, based on who policy makers observe and what kinds of information they value. Gilardi shows that policy learning is strategic, filtered through political networks, and shaped by electoral and ideological incentives.

Gilardi also points out that not all diffusion is due to learning; other mechanisms like competition, coercion, or emulation may also drive the spread of policies. However, when learning does occur, it is selective and interpretive: governments adapt others’ policies to their own political goals and institutional settings. Gilardi’s article highlights that learning in policy diffusion is politically embedded, challenging simplistic models of policy imitation and emphasising the complexity of how knowledge moves between governments.

We should resist the assumption that politicians are naturally inclined to follow evidence. Rather than starting with the hope that politicians will be moved by scientific rationality, a more strategic and realistic approach is to structure the political environment so that ignoring science becomes politically costly. This reframing shifts focus from persuasion to creating institutional incentives that make adherence to evidence advantageous and deviation from it risky.

Studies of legislative practice show that direct, evidence-based challenges in parliamentary settings are relatively rare (Cairney Reference Cairney2016). This reflects a broader problem: the political cost of ignoring or misusing evidence is often minimal. To address this, institutional mechanisms must be introduced or strengthened to build accountability around evidence use. For example, structured hearings and mandatory publication of cost-benefit analyses can help ensure evidence enters political debate.

Radaelli illustrated this with reference to reforms introduced under Tony Blair’s government aimed at strengthening regulatory scrutiny and evidence use in policy making. UK institutional arrangements for regulatory oversight, including committees and scrutiny mechanisms, were tasked with reviewing whether ministers’ proposals were supported by sufficient evidence and regulatory analysis. The power of this model lies not in coercion, but in reputational pressure: ministers would rather avoid the political fallout of being seen as careless or ideologically rigid.

To embed science meaningfully in policy making, scholars must design political structures that reward evidence use and penalise its absence. This includes promoting transparency and building institutional roles for independent experts. Over time, these mechanisms can shift expectations, encouraging politicians to take science seriously because it is politically advantageous. Whether such a culture can flourish beyond democratic regimes remains contested: authoritarian systems may embed expert authority but rarely sustain the openness that accountability requires.

Gilardi’s account exemplifies political learning in action. Politicians acquire knowledge about policy effectiveness, but this knowledge is filtered through electoral calculations and reputational concerns. They learn strategically, identifying which policies will resonate with voters rather than which solve problems effectively. This result is learning oriented toward political survival. The institutional design challenge becomes how to make ignoring instrumental evidence politically costly, thereby aligning political learning incentives with evidence-informed governance.

Story 3: Zaki & Dupont and the expert

Radaelli’s third case examined the expert as a policy actor, drawing on recent work by Bishoy L. Zaki and Claire Dupont published in the Journal of European Public Policy (Reference Zaki and Dupont2024). Their paper investigates how scientific experts operate within EU climate governance institutions like the European Environment Agency. Contrary to the traditional political science view of scientists as apolitical ‘truth-tellers’, their research complicates this. Scientists who are most effective in influencing climate policy are those who understand and engage with policy politics, a form of politics distinct from party or electoral politics. These experts pursue clear policy goals, such as advancing carbon pricing or strengthening emissions targets. Here, expertise is embedded within ongoing political processes rather than standing apart from them to ‘speak truth to power’.

Zaki and Dupont’s work illuminates political learning by scientific experts who learn how the policy process works and how to position their knowledge strategically within it. These experts strategically align their insights with institutional rhythms, learning iteratively through participation in policy networks. They are not neutral technicians but politically savvy actors working toward normative goals through politically sensitive strategies. Zaki and Dupont also show that these experts are not driven by party loyalty but by a commitment to substantive outcomes, making them key players in shaping long-term policy directions in a system that is politically complex but not electorally volatile.

However, engaging in policy politics does not mean scientists are immune to self-interest or institutional pressures. Zaki and Dupont, drawing on broader sociological insights, show that scientists are often motivated by reputational concerns and the sustainability of their research institutions. Experts often feel compelled to project certainty to maintain authority, even when knowledge remains incomplete. The incentive to appear authoritative can lead to overselling science or masking uncertainty, which contradicts the very foundations of scientific integrity. This observation invites a key question: can scientific actors maintain epistemic humility while competing for influence in politicised settings? Or does influence, by necessity, come at the cost of scientific caution?

This brings us to a broader dilemma: how can we cultivate a scientific culture that remains epistemically humble while still being influential in politically saturated policy environments? Zaki and Dupont’s work suggests that scientists must be trained in strategic communication and political navigation. Simultaneously, publics and political institutions need education about the iterative and probabilistic nature of scientific knowledge. This invites a rethinking of how science interfaces with democratic decision-making. Rather than expecting science to provide definitive answers, we might better aim to build a policy process that takes the conditional nature of evidence seriously. Such a shift would require changes in expert behaviour and political culture.

One way to support this shift is through more effective policy communication. Studies of policy advisory systems show that the ability to influence policy depends on how and when expertise is communicated (Craft and Howlett Reference Craft and Howlett2013). Policy makers operate in fast-moving, attention-scarce environments, where evidence competes with ideology and political pressure. Even the most robust research can be ignored if it arrives too late or fails to speak in the language of policy relevance. Experts must therefore communicate strategically, framing findings to match policy windows and repeating key messages to sustain impact.

Writing better briefs is insufficient on its own. A broader skills gap hampers many researchers, such as a lack of training in storytelling and discomfort with simplification. These barriers are often cultural, rooted in academia’s suspicion of advocacy or fear of reputational risk. Yet as Zaki and Dupont’s politically literate experts show, it is possible to remain committed to evidence and epistemic humility while still making a strong, timely case for policy action. Developing these communication capacities is thus not ancillary but essential to the role of the expert in democratic governance. Strategic policy communication offers a way for experts to engage in politics without being partisan.

The scientists in Zaki and Dupont’s account are engaged in sophisticated political learning. They learn the institutional rhythms of Brussels and the framing strategies that gain traction while pursuing substantive climate goals. This is not instrumental learning (updating technical understanding of climate science) but political learning (understanding how to position that science within governance processes). Yet their learning also reveals something the typology struggles to capture: it is fundamentally relational. These experts learn through embedded participation in policy networks and iterative engagement with decision-makers. Their effectiveness depends on their strategic calculation and on building credibility within a community of practice. This relational dimension, learning as positioned interaction within epistemic and organisational communities, points toward a gap in how we theorise political learning.

Beyond design: The background limits of agency and the learning assumption

The three cases examined above reveal how political learning operates across different actor types, each constrained by professional, electoral, or reputational incentives. Yet even this framing, distinguishing instrumental from political learning, risks overstating actors’ agency. Both types of learning occur within background political conditions that set hard limits on what is possible. What this overlooks is how background political and geopolitical conditions fundamentally shape what counts as learnable or actionable knowledge in the first place. Recent work on policy advisory systems highlights how populism and politicisation are reshaping the conditions under which expertise becomes usable or strategically sidelined (Craft, Head and Howlett Reference Craft, Head and Howlett2024). A compelling illustration of this can be found in the long-standing efforts to promote European rearmament. For years, policy makers, security experts, and bureaucrats within the European Union advocated for stronger defence cooperation and investment (Monaghan Reference Monaghan2023). Despite their technical expertise and strategic reasoning, these efforts failed to gain significant traction. What ultimately shifted the political calculus was not the refinement of a particular policy proposal, but an external shock: the actions and rhetoric of Donald Trump, especially his open questioning of NATO commitments and US security guarantees.

The broader insight is straightforward: the success of a policy often depends less on its internal merits or the competence of its advocates than on whether it aligns with the evolving political background. When that background is inhospitable due to dominant ideologies, institutional inertia, or entrenched public opinion, even the most carefully designed proposals are likely to be marginalised. Conversely, a sudden shift in context can render previously unthinkable options both viable and urgent. In this way, the background functions simultaneously as constraint and as a set of affordances that determine what is politically intelligible and actionable at a given time.

What makes this particularly complex is that many policy actors operate under the implicit assumption that background conditions are stable or, at the very least, can be accommodated through careful policy tailoring. This view suggests that policy proposals should conform to existing parameters of political acceptability; for instance, avoiding ideas like tariffs when prevailing economic orthodoxy deems them inefficient or politically toxic. But background conditions are not static; they can shift rapidly, driven by crises, leadership transitions, or international realignments. When such shifts occur, policy institutions often lack the frameworks or capacity to adapt quickly. Policy makers may scramble to define a ‘new normal’, during which time reactive or suboptimal policies can take root.

This cyclical dynamic reveals a deeper limitation in prevailing models of policy learning, namely a disproportionate emphasis on adaptive technocratic refinement and too little attention to the volatility and contingency of the political context itself. If policy making remains tethered to short-term horizons and current parameters of feasibility, it risks becoming brittle and ill-prepared for the sorts of disruptions that regularly reshape what is politically possible. A more future-oriented approach is required, one that builds anticipatory capacity and attends to shifting political conditions, not only to institutional structures.

Is learning best conceptualised as an institutional process of adaptation, or as a political process of alignment between knowledge and contingent opportunity? The former emphasises control and design; the latter demands responsiveness and timing. Much of the learning literature theorises political learning as individual strategic adaptation; actors learning to manage constraints. But what the European rearmament example reveals is that learning is also fundamentally relational and contextual: it depends on alignment between knowledge, shifting political opportunity structures, and institutional readiness. Actors learn in relation to evolving contexts that may suddenly render their carefully accumulated expertise either urgent or irrelevant. This relational quality of learning – emerging from positioned interactions within contingent political moments – has implications for institutional design that go beyond creating better incentive structures.

This limitation of agency helps explain the growing public perception that meaningful agency lies outside the formal policy apparatus. Figures such as Donald Trump are perceived (by supporters and critics alike) as bypassing the institutional machinery of expertise and bureaucratic process, imposing policy through will rather than deliberation. This sentiment is not confined to populist figures. It reflects a broader critique: that policy institutions have become so entangled in procedures and performance that they struggle to act decisively or imaginatively. Much like critiques of legal systems (where justice is said to be obstructed by its own mechanisms), policy making is increasingly portrayed as a system that must be disrupted in order to regain relevance.

To counter this perception and reclaim agency within policy institutions, there is a need to expand the functional architecture of governance to include spaces explicitly devoted to policy generation, building on a substantial literature in policy design, governance innovation, and participatory planning that has examined such institutional reforms for decades. These might take the form of foresight units, innovation hubs, or independent think tanks (though the historical records show that the latter, far from being purely independent, have often been politicised and their autonomy compromised). Their mandate should be speculative and anticipatory, engaging with long-term, hypothetical, and cross-sectoral challenges, while acknowledging that global commissions of inquiry or ‘blue ribbon’ task forces have frequently been used for symbolic politics, delay, or the quiet shelving of politically intractable issues. Such institutional forms could help prevent policy systems from being continually caught off guard by disruptions, but they must also guard against a drift towards ‘consultocracy’, in which ministries outsource critical thinking and learning to private consultants, potentially hollowing out state capacity.

Moreover, a reimagined model of policy learning must involve a broader range of actors. Instead of concentrating epistemic authority within narrow technocratic circles, policy making should become a distributed cognitive process. Civil society organisations, communities with lived experience, and unconventional thinkers must be treated not merely as recipients of policy but as co-producers of knowledge and strategy, through concrete mechanisms such as citizens’ assemblies, deliberative forums, participatory budgeting, and structured advisory panels that embed non-expert knowledge into formal decision-making. The integration of diverse knowledge forms (scientific, experiential, local, and systemic) can foster policies that are technically sound and publicly credible. Yet this pluralism also carries implications for established institutions: it may disrupt existing university structures and disciplinary boundaries, challenging academia to adapt to more open, transdisciplinary forms of knowledge production.

Ultimately, while this article has critiqued the assumption that policy actors function as unified agents of learning, it also resists the conclusion that meaningful reform is impossible. Instead, it argues that agency in policy making must be reconceived not as a stable possession held by specific actors, but as something contingent and relational. It arises in moments when institutional context, political opportunity, and actor capacity align, often briefly and unpredictably. Recognising and cultivating these moments does not resolve the deeper structural limitations of the policy process. But it does suggest a more realistic and potentially fruitful way forward, one focused less on ideal structures and more on amplifying the conditions under which learning (and political imagination) becomes possible.