Abstract The importance of evidence‐based policymaking is widely recognized, but how science influences policy remains insufficiently explored. This study aims to examine how policy documents cite research articles, thereby tracing the complex impact process of scientific research on policymaking. A conceptual model is proposed to classify four types of citation pathways by distinguishing between direct and indirect impacts and observing whether a reinforcement effect is present. To operationalize this model, we collected nearly 10 thousand policy documents related to artificial intelligence (AI) and over 1.6 million links between these policies and their referenced articles. A large‐scale data analysis and a case study were conducted. Results exhibit distinct citation pathways among specific types of institutions, geopolitical areas, and policy areas. Indirect influences emerge as an important mechanism. Research articles from EU countries primarily serve the policymaking of inter‐governmental organizations (IGOs) and the EU, while research articles from the USA significantly support both domestic and foreign policymaking. Notably, IGOs serve as key intermediaries, facilitating the indirect influence of research on policymaking. In addition, while the knowledge from the social sciences provides substantial support for policies in various areas, an increasing involvement of the natural sciences in the development of AI‐related policies is found.

Abstract The classification of research according to its aims has been a longstanding focus in the fields of quantitative science studies and R&D statistics. Since 1963, the Organization for Economic Co‐operation and Development (OECD) has employed a classical distinction among basic, applied, and experimental research. Building on this framework, our previous work highlighted the utility of differentiating between scientific and societal progress as two primary research objectives. This distinction enabled the quantitative analysis of scientific publication abstracts and the development of an automated method for large‐scale classification. In the current study, we systematically evaluate text classification techniques, including traditional text mining models, classification tools, BERT‐based language models, and decoder‐only large language models (LLMs) such as ChatGPT. Our findings show that the fine‐tuned GPT‐4o‐mini model performs the best among single‐model approaches. However, traditional and BERT‐based models outperform in certain fine‐grained classification tasks. Leveraging majority voting strategies to incorporate their strengths yields performance comparable to closed‐source GPT models. A case study on 10 biomedical journals further validates the method, demonstrating strong alignment between journal scopes, model predictions, and outputs generated by the fine‐tuned GPT‐4o‐mini model. These results highlight the robustness and practical effectiveness of the proposed methodology for nuanced research aim classification.

Abstract The purpose of this study is to investigate the validity of tweets about scientific publications as an indicator of societal impact by measuring the degree to which the publications are tweeted beyond academia. We introduce methods that allow for using a much larger and broader data set than in previous validation studies. It covers all areas of research and includes almost 40 million tweets by 2.5 million unique tweeters mentioning almost 4 million scientific publications. We find that, although half of the tweeters are external to academia, most of the tweets are from within academia, and most of the external tweets are responses to original tweets within academia. Only half of the tweeted publications are tweeted outside of academia. We conclude that, in general, the tweeting of scientific publications is not a valid indicator of the societal impact of research. However, publications that continue being tweeted after a few days represent recent scientific achievements that catch attention in society. These publications occur more often in the health sciences and in the social sciences and humanities.

Abstract This paper investigates different uses of the Journal Impact Factor (JIF) in national journal rankings and discusses the merits of supplementing metrics with expert assessment. Our focus is national journal rankings used as evidence to support decisions about the distribution of institutional funding or career advancement. The seven countries under comparison are China, Denmark, Finland, Italy, Norway, Poland, and Turkey—and the region of Flanders in Belgium. With the exception of Italy, top‐tier journals used in national rankings include those classified at the highest level, or according to tier, or points implemented. A total of 3,565 (75.8%) out of 4,701 unique top‐tier journals were identified as having a JIF, with 55.7% belonging to the first Journal Impact Factor quartile. Journal rankings in China, Flanders, Poland, and Turkey classify journals with a JIF as being top‐tier, but only when they are in the first quartile of the Average Journal Impact Factor Percentile. Journal rankings that result from expert assessment in Denmark, Finland, and Norway regularly classify journals as top‐tier outside the first quartile, particularly in the social sciences and humanities. We conclude that experts, when tasked with metric‐informed journal rankings, take into account quality dimensions that are not covered by JIFs.

We investigate the state of multilingualism across the social sciences and humanities (SSH) using a comprehensive data set of research outputs from seven European countries (Czech Republic, Denmark, Finland, Flanders [Belgium], Norway, Poland, and Slovenia). Although English tends to be the dominant language of science, SSH researchers often produce culturally and societally relevant work in their local languages. We collected and analyzed a set of 164,218 peer-reviewed journal articles (produced by 51,063 researchers from 2013 to 2015) and found that multilingualism is prevalent despite geographical location and field. Among the researchers who published at least three journal articles during this time period, over one-third from the various countries had written their work in at least two languages. The highest share of researchers who published in only one language were from Flanders (80.9%), whereas the lowest shares were from Slovenia (57.2%) and Poland (59.3%). Our findings show that multilingual publishing is an ongoing practice in many SSH research fields regardless of geographical location, political situation, and/or historical heritage. Here we argue that research is international, but multilingual publishing keeps locally relevant research alive with the added potential for creating impact.

Abstract Numerous studies have shown that female scientists tend to publish significantly fewer publications than do their male colleagues. In this study, we have analyzed whether similar differences also can be found in terms of citation rates. Based on a large‐scale study of 8,500 Norwegian researchers and more than 37,000 publications covering all areas of knowledge, we conclude that the publications of female researchers are less cited than are those of men, although the differences are not large. The gender differences in citation rates can be attributed to differences in productivity. There is a cumulative advantage effect of increasing publication output on citation rates. Since the women in our study publish significantly fewer publications than do men, they benefit less from this effect. The study also provides results on how publication and citation rates vary according to scientific position, age, and discipline.