Introduction
The roundtable ‘Knowledge and Democratic Accountability in Policy Making’, hosted by the European Consortium for Political Research (ECPR) on 25 November 2024, brought together leading scholars and practitioners to explore the tensions and opportunities at the intersection of expertise, evidence, and democracy. The discussion featured Heather Grabbe and Gaby Umbach, both of whom have long-standing experience in academic research and European policy advisory contexts. This paper extends their insights through the analysis that follows, reflecting on the increasingly urgent challenges of evidence-informed policy making. From the forced relocation of Central European University (CEU) from Hungary (where the government passed legislation specifically targeting the institution, ultimately forcing it to move its operations to Vienna in 2019) to legislative attacks on Diversity, Equity and Inclusion (DEI) research in the United States, expertise is being institutionally repressed, not merely contested. In such contexts, the assumption that evidence-based policy operates on a neutral terrain becomes untenable.
Experts are expected to deliver clarity and neutrality, yet they operate within systems that politicise knowledge and sideline plural forms of understanding. Under what conditions is expertise (mis)used or ignored in policy making? How can its application be made compatible with democratic legitimacy?
The paper analyses the institutional conditions shaping expertise in policy making; identifies three core challenges – literacy, capacity, and politics – and proposes pathways for democratising the knowledge–policy interface without rendering expertise irrelevant. Our contribution lies in analysing evidence use as an increasingly institutionalised site of political repression, beyond the familiar problems of uptake or translation, with implications for how expertise can be democratised without collapsing into relativism.
Motivations and conditions for using expertise in policy making
Why is expertise used in policy making? Conventionally, it is seen as a tool for improving decision-making through greater accuracy or accountability. In complex or contested domains such as climate change, public health, migration, or AI regulation, evidence is often positioned as a neutral, rational foundation for decision-making. This aligns with the normative model of evidence-informed policy making (EIP), which emphasises the use of the ‘best available evidence’ to guide policy choices (Head Reference Head2010; Cairney Reference Cairney2020). Yet this portrayal of evidence as objective and apolitical has been widely challenged.
But what happens when the very invocation of ‘evidence-based policy’ becomes part of a contested political script? Recent developments suggest that we cannot presume evidence carries democratic legitimacy if its production is seen as partisan or exclusionary. Evidence is embedded in normative worldviews and power structures. As Parkhurst (Reference Parkhurst2017) notes, it is essential to distinguish between the quality of evidence and the quality of its use. What counts as legitimate evidence is shaped as much by political context as by empirical robustness. As Grabbe and Lehne (Reference Grabbe and Lehne2015) argue, declining trust in European Union (EU) institutions has become inseparable from broader struggles over democratic legitimacy.
For knowledge to be impactful in policy processes, it must be perceived as credible and legitimate. These attributes are not intrinsic to the evidence itself but are co-produced by the institutions and processes that mediate its use: for example, think tanks, advisory councils, expert commissions, and public consultations. These mediating actors include both boundary intermediaries and specialised analytical bodies embedded within public administrations. At the EU level, organisations have also invested strategically in scientific prestige: the European Central Bank’s process of ‘hyper-scientisation’, developed through its Directorate-General for Research, has functioned as a resource for positioning authority within networks of national central banks (Mudge and Vauchez Reference Mudge and Vauchez2016). For instance, the UK’s Behavioural Insights Team (a quasi-governmental organisation that applies behavioural science to policy design) and Germany’s Expert Council on Integration and Migration (an independent advisory body that provides research-based recommendations on migration policy) exemplify how intermediaries influence what is seen as authoritative knowledge, while raising questions around transparency and technocratic power (Weingart and Lentsch Reference Weingart and Lentsch2011).
Evidence use is rarely neutral. Alongside instrumental goals like cost-effectiveness, impact, or risk minimisation, evidence is often mobilised to advance normative aims such as fairness and democratic accountability. These logics can come into tension, particularly when evidence is used selectively to validate pre-existing ideological preferences or political narratives. During the Eurozone crisis, policy makers invoked economic expertise to justify austerity policies, even when such strategies were contested by alternative schools of thought, and within the institutions involved in crisis management (Pénet Reference Pénet2018; Schmidt Reference Schmidt2020). Even when analytical errors are identified, for example, the IMF’s misestimation of fiscal multipliers, policy orientations can persist while claims to epistemic legitimacy remain intact (Matthijs and Blyth Reference Matthijs and Blyth2018). A key illustration is the concept of ‘expansionary austerity’ – the controversial claim that fiscal retrenchment could stimulate rather than suppress economic growth – later promoted by a network of economists from Bocconi University, proved instrumental in shaping responses to the Great Recession. Despite significant scholarly debate about its empirical validity, this concept served to legitimise politically palatable fiscal retrenchment as economically prudent (Helgadóttir Reference Helgadóttir2016), illustrating how evidence can be mobilised to validate pre-determined policy preferences.
What constitutes the ‘public interest’ is itself contested ground. Competing actors interpret it through divergent normative lenses such as economic growth, social equity, ecological sustainability, or national identity. As Bhuta, Malito and Umbach (Reference Bhuta, Malito, Umbach, Malito, Umbach and Bhuta2018) argue, statistical classifications and indicators function as proxies for deeper normative assumptions. What we choose to measure effectively shapes what we govern.
The questions experts are asked, the data that is collected, and the metrics that are prioritised all reflect underlying social relations. Indeed, evidence is often produced within institutional frameworks that privilege particular paradigms, such as economic rationalism or technocratic managerialism. When evidence is used performatively, it closes down debate rather than opening it. Yet, if evidence is always embedded in power, this raises the question of how we might distinguish between treating all expertise as equally ideological and recognising more reflexive and plural forms of expertise that differ from narrower, partisan ones.
Challenges to uptake and trust in expertise
The uptake of expertise in policy making is shaped by a range of contextual and institutional conditions, including political culture, administrative structure, timing, and the nature of trust relationships. Research suggests that knowledge is more likely to be integrated when it is problem-focused, timely, and aligned with prevailing policy frames (Cairney and Kwiatkowski Reference Cairney and Kwiatkowski2017). Evidence uptake is also relational, as it depends on whether policy makers perceive knowledge producers as credible and trustworthy.
In addition, policy processes often operate under tight deadlines, creating a mismatch between the slow, iterative rhythms of academic research and the rapid cycles of political decision-making (Bojovic and Bayley Reference Bojovic and Bayley2025). This temporal misalignment can disincentivise deep engagement with complex or conflicting evidence, especially when it challenges dominant policy frames. As a result, policy makers may resort to symbolic uses of evidence or cherry-picking to maintain coherence and control. Institutional arrangements can help bridge this credibility gap. For example, in-house parliamentary research services, national statistical offices, and the European Commission’s Expert Groups serve as trusted intermediaries that provide tailored, policy-relevant knowledge.
However, recent developments have significantly complicated this terrain. The politicisation of expertise has taken increasingly institutional forms, ranging from government directives banning specific research topics to budgetary threats aimed at silencing dissenting academics. In the US, the defunding of DEI research and threats against leading universities such as Harvard and Columbia have raised serious questions about academic freedom. In Europe, CEU’s forced departure from Budapest remains a stark example of how expertise can be targeted on ideological grounds. The EU’s limited ability to prevent or reverse CEU’s expulsion illustrates a weakness in its diplomatic leverage and raises questions about the Union’s credibility as a guardian of democratic principles. This episode also sits within the EU’s broader struggle to respond to democratic backsliding, including through the development of a rule-of-law conditionality regime enabling the suspension of EU funds in cases of systemic breaches (Regulation 2020/2092; Coman and Buzogány Reference Coman and Buzogány2024).
Expertise is a flexible status that can be embodied by a variety of actors, ranging from scientists in universities to those in think tanks funded by private consortia of companies (Oreskes and Conway Reference Oreskes and Conway2010). Populist movements often frame experts as disconnected elites, out of touch with the concerns of ‘ordinary people’. This narrative has gained traction, particularly when expert advice is perceived as inconsistent or exclusionary. The Covid-19 pandemic made these tensions visible, as rapid shifts in scientific guidance (such as evolving recommendations on mask-wearing) eroded public trust, even where expert disagreement reflected genuine uncertainly rather than bad faith (Jasanoff, Hilgartner, Hurlbut et al. Reference Jasanoff, Hilgartner, Hurlbut, Özgöde and Rayzberg2021). In some cases, experts became targets of conspiracy theories or state censorship.
These attacks reveal a broader crisis: once universities and research institutions are themselves portrayed as partisan or corrupt, the legitimacy of their outputs suffers. The result is a breakdown of the very conditions under which evidence can be trusted.
Three core challenges: Literacy, capacity, politics
The literacy challenge
The first challenge involves the uneven levels of evidence literacy among policy makers, the public, and even some experts themselves. Policy-making environments often privilege certain disciplines over others, creating a hierarchy of expertise. Disciplines like economics and law tend to dominate advisory roles, partly because they produce quantifiable and seemingly neutral data like statistics and models that align with policy makers’ preference for clarity and certainty. By contrast, social sciences and humanities, which offer critical and contextual insights, are increasingly marginalised despite their capacity to contextualise policy problems and expose underlying assumptions. As Umbach (Reference Umbach2020) demonstrates, disciplinary hierarchies systematically privilege quantifiable data over qualitative, local, or experiential forms of knowledge. This privileging of ‘hard’ data creates significant blind spots in problem definition and policy design, marginalising insights that do not fit easily into metrics or cost-benefit frameworks.
The debate here concerns not the need for evidence literacy, but how to cultivate it fairly and effectively. A variety of answers can be explored, including the training of policy makers in the language of science and the training of scientists in the grammar of politics. Academic research often unfolds slowly, emphasising caution, uncertainty, deliberative processes of peer review and iterative refinement. Policy makers, by contrast, operate under tight timeframes, electoral cycles, and media pressure, demanding rapid, actionable insights. Crises such as the Covid-19 pandemic or energy shocks have highlighted this asymmetry, privileging certain forms of ‘fast knowledge’ such as epidemiology or macroeconomics at the expense of broader contextual understanding. The pandemic exposed the pressure to accelerate knowledge production, leading to both innovation and controversy. In the French context, socio-technical controversies around Covid-19 treatments (most notably heated debates about hydroxychloroquine promoted by some high-profile scientists), amplified by extensive media coverage, increased public awareness of scientific disagreements while inadvertently fostering anti-science attitudes (Schultz, Atlani-Duault, Peretti-Watel et al. Reference Schultz, Atlani-Duault, Peretti-Watel and Ward2022).
There is, therefore, another tension here, this time between fundamental research and usable policy recommendations. To bridge these two domains, effective translation mechanisms are needed, in communication and epistemic terms alike. But this raises further questions. For instance, should we embed more academics directly into policy-making structures? Or is it preferable to rely on dedicated boundary organisations that preserve both independence and relevance?
Policy makers cannot be expected to absorb findings from research directly. This gap can be mediated more effectively through boundary organisations rather than individual actors, such as specialised services within public administrations. Examples like the European Commission’s Joint Research Centre or university-based policy labs, such as the Policy Lab at the Université libre de Bruxelles or the KU Leuven internal think tank, offer institutional models. Their goal is to structure and translate knowledge produced by research, without subjecting all academics to policy-writing imperatives. They seek to mediate between academic research and policy making by framing relevant questions and managing trust. In an era marked by information overload and erosion of public trust in experts, knowledge management becomes crucial, particularly the task of filtering data into actionable intelligence.
However, translation is not a neutral act as it involves interpretation and the exercise of power. Without reflexivity (as both an individual capacity and a collective one), translation risks distorting complex findings or reducing them to politically expedient soundbites. Improving evidence literacy, therefore, requires more than individual capacity-building. It requires structural changes in how knowledge circulates, including questions of who gets to speak, what kinds of evidence are seen as legitimate, and how different knowledges are weighed against each other. Without such changes, the divide between experts and non-experts can deepen, reinforcing technocratic exclusion rather than democratic inclusion.
The capacity challenge
Even where there is political will and conceptual understanding, institutional capacity to meaningfully engage with complex evidence remains limited. Policy makers work under time scarcity and information saturation, which privileges knowledge that is quick and familiar, reinforcing reliance on a small pool of trusted experts and institutions (Head Reference Head2016). While this expedites decision-making, it narrows epistemic diversity and marginalises critical or interdisciplinary perspectives. The internal logics of bureaucratic institutions further entrench these tendencies. Legal-administrative procedures and rigid accountability structures often leave little room for interpretive judgement or open-ended deliberation.
The ‘capacity gap’ is as much cultural and structural as technical, reinforcing incrementalism and limiting timely uptake of evidence (Lindblom Reference Lindblom1959). In multi-level governance systems like the European Union, these issues are exacerbated. Expert groups and advisory mechanisms often serve formal legitimacy functions rather than genuine epistemic engagement (Radaelli Reference Radaelli1999). This fosters an ‘evidence façade’, where consultation is performative rather than integrative.
Bridging the capacity gap requires rethinking institutional design. This includes creating reflexive spaces for critical engagement, aligning timelines for co-production, and reforming accountability systems to reward deliberation as much as decisiveness.
The political challenge
We know that evidence is just one of many inputs, alongside values, political feasibility, and institutional norms. Purely ‘evidence-based’ policy making would resemble technocracy more than democracy. In the EU context, this tension has often been managed through a reliance on output legitimacy and throughput legitimacy (Schmidt Reference Schmidt2020). However, policy making is not algorithmic, and complex social issues cannot be reduced to metrics and models. Evidence is inevitably shaped by the political context in which it is produced and used. The core dilemma is whether evidence should be insulated from politics or openly contested in democratic ways. Policy debate often skirts this dilemma, framing evidence as either technocratic or populist while avoiding the harder question of how to reconcile expertise with accountability. While data can inform policy design, it cannot eliminate the need for ethical judgement and contextual sensitivity. This is especially visible in agenda-setting processes, where expertise can be mobilised to justify decisions already made or to close debate rather than inform it.
Conversely, neglecting or manipulating evidence undermines both policy effectiveness and democratic accountability. The challenge lies in balancing analytical rigour with reflexivity, which means acknowledging that evidence should inform but cannot override the inherently political character of policy making. But how can we conceptualise this balance in times when the value of expertise itself is contested?
Populist rhetoric often portrays experts as disconnected elites who impose decisions without democratic consent. Experts such as scientists and economists are, in this perspective, positioned within ‘the elite’. This anti-elitist rhetoric has been used to delegitimise inconvenient evidence, particularly on issues where the stakes are ideologically charged. For instance, climate change denial often draws on the claim that climate scientists are driven by political agendas rather than empirical observation (Oreskes and Conway Reference Oreskes and Conway2010; Eyal Reference Eyal2019). Similarly, in debates on migration, statistical data and economic analyses are frequently dismissed as biased or manipulated, especially when they counter nationalist or exclusionary narratives (Boswell Reference Boswell2009).
The US case is particularly striking. Throughout Trump’s second term, academics and experts have come under intense pressure and threats to independence: from the freezing of billions in federal grants to ideological ‘keyword’ sweeps aimed at DEI or gender-related research. Trumps’s administration has threatened universities such as Harvard and Columbia with withheld funds unless they conform to politically driven measures, including restrictions on hiring and speech on campus. In response, we have witnessed mobilisation beginning in February 2025 in the form of letters and demonstrations, such as the ‘Stand Up for Science’ events that have been taken up worldwide.Footnote 1
This illustrates a deeper tension in contemporary governments where emergency politics are becoming the norm. While crises are deemed to necessitate greater expert intervention, they also intensify scrutiny of who qualifies as an expert, how expertise is produced, and on whose behalf it is exercised. The growing distance between policy making informed by expertise and political communities is itself a democratic problem.
This observation raises questions about how to democratise expertise itself. Expertise is a socially situated practice whose legitimacy depends on institutional trust and public accountability, not the delivery of facts alone. This entails acknowledging that evidence is always value-laden and politically embedded, while still creating institutional conditions for its transparent and accountable use.
Ways forward: Building a better knowledge–policy interface
The challenges outlined above of literacy, capacity, and politics demand constructive responses, not only critique. While no definitive blueprint exists, a range of tentative strategies have been proposed for improving the relationship between knowledge and policy and sustaining democratic legitimacy. Each reform involves trade-offs: between inclusion and efficiency and between autonomy and accountability. We argue for four interlinked directions.
Broaden the circle
We argue for an expansion of actors and forms of knowledge recognised within policy making. Legitimacy is enhanced when expertise draws on practical, local, and experiential knowledge alongside academic and technical inputs. This involves moving beyond elite expert panels and think tanks to engage civil society actors, community representatives, and frontline practitioners in the policy process. In practice, this requires dedicated resources for meaningful participation – not tokenistic consultation – including funding for community organisations to engage substantively and institutional mechanisms that ensure diverse voices shape problem definition, not merely react to pre-formed solutions.
The notion of ‘knowledge justice’ highlights structural exclusions in whose knowledge is valued (Medina Reference Medina2013). Elevating lived experience can increase the contextual relevance of policy, especially for marginalised groups, but requires participatory infrastructures and processes that allow for sustained, meaningful contributions. This means transforming knowledge from a technocratic input into a co-produced public resource.
This pluralistic approach may introduce friction and complexity, but friction can build deeper trust. Innovative deliberative formats, such as citizens’ assemblies, participatory foresight, and science-policy labs, have been proposed as mechanisms to embed pluralism in the knowledge-policy nexus (OECD 2020). These approaches have achieved some success, for instance the Irish Citizens’ Assembly on abortion (which recommended repealing Ireland’s ban, subsequently passed in referendum) and the EU’s Conference on the Future of Europe (2021–2022). Such initiatives show what structured participation can achieve, though their effects on policy outcomes remain uneven and highly contested. For example, France’s Convention citoyenne pour le climat brought together 150 citizens who deliberated extensively with scientific experts and produced 149 policy recommendations; however, many of these proposals were ultimately diluted or set aside, illustrating how mini-publics are often institutionalised as consultative devices rather than mechanisms with binding decision-making power (Landemore and Fourniau Reference Landemore and Fourniau2022). Co-production means involving publics in formulating questions and interpreting findings rather than presenting them with finished results. Such approaches can demystify evidence and build trust by grounding knowledge generation in real-world concerns. They also require institutional openness to deliberation and methodological diversity.
Inclusion, though, comes with its own difficulties, such as who participates and how to handle divergences between expert consensus and citizen views. Deliberative formats offer partial answers. They foster dialogue across epistemic divides and help counter technocratic drift. Yet they also demand careful design to avoid elite capture or symbolic engagement. This means establishing clear selection criteria and transparent processes for how citizen input influences policy outcomes.
Use crises as learning moments
While emergencies like pandemics or climate disasters increase the demand for expert guidance, they also risk triggering the centralisation of power and sidelining democratic oversight (White Reference White2019). Expert advice, in such contexts, may be treated as unassailable, resulting in diminished public trust when uncertainties or revisions emerge.
Crises can be sites of learning rather than justifications for centralising power. Assessing whether learning occurs evenly requires attention to power asymmetries: some actors are better placed to learn from crises than others, depending on the context (Hamm Reference Hamm2025). Reflexivity and transparency are essential for navigating uncertainty openly rather than suppressing it. Institutions must acknowledge when evidence is incomplete or evolving. Mechanisms such as post-crisis evaluations and policy experimentation can embed learning into the governance cycle. This suggests a broader shift from a logic of prediction and control to one of reflexivity and adaptation. The politics of expertise must be prepared for surprise and disagreement, within science as much as in public life.
Invest in independent infrastructure
Expertise cannot function without the institutions that sustain it. Research funding and academic freedom are foundational, and yet these foundations are increasingly eroded. Performance pressures and shrinking institutional resources leave little time for the slow, reflexive scholarship that independent expertise requires. The project ‘Universities for the Future: Infrastructuring Reflexivity’ at the European University Institute (EUI) explores how universities can foster the reflexivity and collective work that they are uniquely positioned to support, in opposition to the pressures of career acceleration and intellectual isolation. Support for the social sciences and humanities is particularly vital for interpreting social values and institutional dynamics. Political interference and the short-term, deliverable-driven logic of many policy contracts can place pressure on the autonomy and quality of research. Interdisciplinary structures, such as the international Horizon-funded platforms, have shown promise but require long-term commitment. Such commitment means moving beyond project-based funding towards sustained institutional support, protecting academic freedom through legal frameworks, and establishing buffers against political interference.
A related debate is whether expertise should be institutionally ‘embedded’ in governance or remain at arm’s length. Too much proximity may lead to capture or self-censorship, while too much distance may render expertise irrelevant. The challenge is to build hybrid models that sustain both independence and dialogue through boundary organisations, secondments, fellowships, and robust protocols for transparency and conflict of interest.
Accept contestation
Finally, we must embrace the idea that disagreement over evidence is productive, not pathological. In pluralistic democracies, contestation is inevitable and often desirable. What matters is the quality of the forums in which it occurs. Scientific independence calls for an engaged scholarship that is both accessible to policy makers and protected from political capture.
Transparency in the selection and use of evidence is essential to prevent manipulation and build public trust. Evidence is an invitation to informed debate and not a verdict. This requires new skills, technical and analytical, but also relational and political. If we want expertise to play a meaningful role in public life, we must make it more accountable without surrendering authority, and humbler without becoming self-erasing. This is a difficult balance, yet one worth striving for.
Conclusion
In November 2024, when the School of Transnational Governance at the EUI hosted the ECPR roundtable ‘Knowledge and Democratic Accountability in Policy Making’, the new US government had not yet taken office. Since then, we have witnessed the intensification of political pressures on universities and expert institutions. In the United States, academics face growing legislative scrutiny, with funding for entire research areas including climate science and gender studies under threat. In France, Germany, and the UK, researchers are increasingly accused of ideological bias for engaging with questions of race, colonialism, or inequality. These developments form a pattern where the authority of expertise is no longer merely questioned but actively undermined.
If we believe expertise remains a vital component of effective and legitimate policy making, its value can no longer be taken for granted. This paper has outlined three interconnected challenges of literacy, capacity, and politics that shape how evidence is used and contested. These challenges are deeply political, not merely technical. Addressing them requires a more reflexive understanding of what expertise is, who defines it, and how it can serve democracy in a time of polarisation. But the core dilemma remains: how can we democratise how evidence reaches policy without weakening the credibility or coherence of expert advice? This is not a question with a clear answer. While technocratic closure and populist rejection represent opposing risks, both stem from the same breakdown – a failure to sustain trust across institutional and political lines. The question is how to use expertise well, not whether to use it.
We have proposed several tentative steps toward this aim. The proposals share a common thread: expertise must become more accountable without ceasing to be authoritative. What is at stake goes beyond institutional design: it is the preservation of democratic principles in how we know and decide. Reclaiming expertise as a democratic good demands clarity about its values and humility about its limits. The task ahead is to democratise it without diminishing its epistemic power.
If expertise is to retain its public value, it must remain open to challenge without becoming merely instrumental. That balance is challenging, but the only one worth striking.
Data availability statement
Data availability is not applicable to this article as no new data were created or analysed in this study.
Financial support
There was no other funder in the drafting of this paper.
Competing interests
There are no conflicts of interest or competing interests in the development of the article.
Open access
