LIST OF PANELS

The role of public policy evaluation is undergoing a profound transformation in today’s fast-paced and interconnected world. The evaluation is at a crossroads, confronted by significant challenges that compel a rethinking of its role to improve the methods and conditions of public discourse at all levels and stages of policy-making (Majone 1989). Other than the recognized, growing complexity of policy issues (Bundi& Trein, 2021; Stame, 2022; Zimmerman et al., 2023), new challenges stand out: the convergence of generative artificial intelligence (genAI) (Ballestar et al., 2019, Zuiderwijk et al., 2021; Margalit& Raviv, 2023, Reid, 2023) and the rise of populism (Head& Banerjee, 2019; Batory&Svensson 2019; Robert et al. 2021; Mira et al., 2024). The enormous quantity and speed of information circulating through the social media downsize the importance of accuracy and challenge the role and reputation of experts.

This is at odds with the fact that, especially the most complex problems can only be tackled by understanding their fundamentally ambiguous and unpredictable elements. Grasping these forms of complexity in public issues is the core of evaluation and the only way to design actions that are not only accurate but also actionable (Zimmerman et al., 2023).

These are not merely disruptive forces but rather catalysts for change that demand our reflection and drive to innovative approaches. It also means reflecting on the role that evaluation can or should play in identifying and mitigating the negative effects of these dynamics in the policy process, such as over-simplification, intolerance of ambiguity, and reductionism.

In particular:

• genAI offers new capabilities for data analysis, but its introduction into evaluation raises significant concerns and requires careful consideration of the ethical and methodological implications. One of the key challenges is the increasing use of AI by government officials themselves, potentially bypassing professional evaluators altogether. AI-driven tools like machine learning algorithms can process large amounts of data, offering speed and scope that human evaluators cannot match (Ballestar et al., 2019), and challenging traditional ways to analyze causal relations (Cowls and Schroeder 2015). Excessive simplification and misinterpretation of information pose a significant risk in the careless use of AI and the judgements that could derive from it.
• The rise of populism has also significantly influenced policy evaluation. Populist politics often emphasize emotionally resonant narratives and social expectations over in-depth, rigorous, evidence-based decision-making ​(Batory&Svensson 2019; Robert et al. 2021). The trend toward simplified, emotive communication can marginalize the role of in-depth evaluations, particularly in complex areas like the migration policy, social inclusion or (in case of EU) Green Deal strategy. As short-term political gains take precedence, evaluators are left with the challenge of maintaining the integrity of evidence-based findings. 

The panel addresses three key questions:

• How can AI be ethically and effectively integrated into policy evaluation while preserving human oversight and judgment?
• What strategies can evaluators use to ensure that evidence-based findings influence decision-making in populist political environments?
• Which approaches, methods, and tools can be used to bring the complexity of public policies and the inadequacy of simplistic solutions back to the center of public debate?

The proponents of papers are invited to present studies and analyses addressing one or both of the following topics:

genAI: 

genAI can be used to reduce the discretion of Street Level Bureaucracies (SLBs) and avoid very different decisions on the same cases and reduce ‘noise’ (Kahneman et al. 2021), especially for simple and repetitive cases; but there may be barriers in the ability to use it, including the lack of adequate training on how to use the platforms, shortage of staff, professional culture, and user distrust. But it is necessary to emphasise that attention must also be shifted to the level of those who design the algorithms, since we are often faced with bias and the inability of algorithms to capture the nuance of the situations confronted by street-level bureaucrats, whereby the distance between designers and users signals the shift from the discretion of SLBs to ‘digital or system discretion’  (Gillingham et al. 2024; Bovens and Zouridis, 2002; Zouridis et al. 2020).

Moreover, as research shows, AI can be prone to biases, hallucinations, and a lack of transparency in its decision-making processes, leaving us uncertain about the sources of its analysis ​(Reid, 2023; Margalit& Raviv, 2023). The fact that these tools are now in the hands of public administration officers and policymakers raises ethical and practical questions about accountability (Novelli et al. 2024), the quality of insights they generate, and the potential sidelining of evaluators.

Populism:

In a context dominated by populisms, the need to develop strategies for communicating complex evaluation results to both policymakers and the public becomes more urgent than ever. How can we ensure that evidence-based evaluation remains central to policymaking in the face of rising populist sentiment? How can evaluation findings be communicated effectively in a political climate dominated by populist rhetoric? How evaluation can contribute to ‘improve the methods and conditions of public discourse at all levels and stages of policy-making’ (Majone 1989).