Understanding both party positions on issues and the relative importance, or salience, they attach to these is central to the study of party competition. Recent work shows that large language models (LLMs) can estimate issue positions from manifestos. This article asks whether LLMs can capture issue salience. Building on work by Benoit et al., it adapts LLM-based methods to measure relative issue salience in a multilingual corpus of manifestos. Alternative strategies are tested - including unconstrained scoring, ranking, saliency budgets, and pairwise comparisons - assessing output validity against benchmark expert surveys, human-labelling of manifestos. LLMs produce meaningful estimates of relative issue salience, though with lower correspondence to expert judgements than for issue positions. This highlights the promise of LLMs, the need for careful thinking about concepts of issue importance and salience, and about whether 'strategic' party manifestos are the best source of information about the 'true' importance for parties of particular issues.

Artificial intelligence (AI) and data science are reshaping public policy by enabling more data-driven, predictive, and responsive governance, while at the same time producing profound changes in knowledge production and education in the social and policy sciences. These advancements come with ethical and epistemological challenges surrounding issues of bias, transparency, privacy, and accountability. This special issue explores the opportunities and risks of integrating AI into public policy, offering theoretical frameworks and empirical analyses to help policymakers navigate these complexities. The contributions explore how AI can enhance decision-making in areas such as healthcare, justice, and public services, while emphasising the need for fairness, human judgment, and democratic accountability. The issue provides a roadmap for harnessing AI’s potential responsibly, ensuring it serves the public good and upholds democratic values.

Political scientists lack domain-specific measures for the purpose of measuring the sophistication of political communication. We systematically review the shortcomings of existing approaches, before developing a new and better method along with software tools to apply it. We use crowdsourcing to perform thousands of pairwise comparisons of text snippets and incorporate these results into a statistical model of sophistication. This includes previously excluded features such as parts of speech and a measure of word rarity derived from dynamic term frequencies in the Google Books data set. Our technique not only shows which features are appropriate to the political domain and how, but also provides a measure easily applied and rescaled to political texts in a way that facilitates probabilistic comparisons. We reanalyze the State of the Union corpus to demonstrate how conclusions differ when using our improved approach, including the ability to compare complexity as a function of covariates.