Research

Federal research funding shapes the direction, diversity, and impact of the US scientific enterprise. Large language models (LLMs) are rapidly diffusing into scientific practice, holding substantial promise while raising widespread concerns. Despite growing attention to AI use in scientific writing and evaluation, little is known about how the rise of LLMs is reshaping the public funding landscape. Here, we examine LLM involvement at key stages of the federal funding pipeline by combining two complementary data sources: confidential NSF and NIH proposal submissions from two large US R1 universities, including funded, unfunded, and pending proposals, and the full population of publicly released NSF and NIH awards. We find that LLM use rises sharply beginning in 2023 and exhibits a bimodal distribution, indicating a clear split between minimal and substantive use. Across both private submissions and public awards, higher LLM involvement is consistently associated with lower semantic distinctiveness, positioning projects closer to recently funded work within the same agency. The consequences of this shift are agency-dependent. LLM use is positively associated with proposal success and higher early-stage publication output at NIH, whereas no comparable associations are observed at NSF. Notably, the productivity gains at NIH are concentrated in nonhit papers rather than the most highly cited work. Together, these findings provide large-scale evidence that the rise of LLMs is reshaping how scientific ideas are positioned, selected, and translated into publicly funded research, with implications for portfolio governance, research diversity, and the long-run impact of science.

Understanding how scientific research informs policy is central to the social sciences, yet systematic evidence of this process remains limited. Bearing in mind that uses can take many different forms, we present a large-scale analysis of how psychological research is cited in U.S. policy documents between 2000 and 2022 to examine one pathway of research uptake in the public sphere. Curating a large data pool linking policy documents (N = 175,987) to social science papers they referenced (N = 1,124,466), we first find that psychological research was cited at consistently higher—and rapidly growing—rates compared to other social sciences. Next, a logistic regression comparing psychology papers (N = 61,521) cited by policy documents versus psychology papers not cited by policy documents (N = 1,672,748) identified paper-level features associated with increased odds of being cited by policy documents, such as coverage by mass media (OR = 1.23) and funding from the NIH (OR = 1.14). Additionally, a structural topic model on the cited research reveals that papers focused on applied topics are associated with more citations in policy documents, though not necessarily with more citations within academia. Finally, using network and computational text analysis, we find that policy documents citing psychology most often address education, health, and social justice issues. Together, these results offer an empirical foundation and methodological framework for understanding an important, public, and traceable pathway in which research translates into policy uptake.

Addressing climate change requires political elites to share a basic set of facts about climate science, yet political elites in the United States are divided in their views about climate change. We document this using the first large-scale survey of over 3,500 U.S. political elites—including elected officials, staffers, regulators, lobbyists, and policy professionals—to assess the partisan divide in beliefs about climate change held by political elites. We show near-unanimous agreement among the Democratic elite on the scientific consensus that global warming is occurring, primarily caused by humans, and widely recognized by scientists. In contrast, substantial minorities of Republican elites reject these scientific facts, with fewer than half affirming anthropogenic climate change and nearly one-third endorsing a climate-related conspiracy theory. Comparing elites to the general public, we find that political elites are more aligned with climate science, but partisan gaps among elites are as wide as those observed in mass opinion. Regression analyses show partisan identity explains far more variation in elite climate beliefs than ideology, trust in science, or broader conspiratorial predispositions. These findings suggest that partisan polarization among elites reflects not only strategic electoral behavior but also privately held attitudes.

We propose Sci2Pol-Bench and Sci2Pol-Corpus, the first benchmark and training dataset for evaluating and fine-tuning large language models (LLMs) on policy brief generation from a scientific paper. We build Sci2Pol-Bench on a five-stage taxonomy to mirror the human writing process: (i) Autocompletion, (ii) Understanding, (iii) Summarization, (iv) Generation, and (v) Verification. It features 18 tasks in multiple-choice and open-ended formats. Specifically, for the Generation stage, we show that BERTScore and ROUGE scores fail to capture the quality of brief writing, and introduce a new LLM-based evaluation metric aligned with expert judgement. Using this benchmark, we evaluate 13 leading open-source and commercial LLMs to uncover key limitations. To improve LLM performance on brief writing, we curate the Sci2Pol-Corpus for fine-tuning. We start by linking each cited scientific paper to its corresponding policy document, drawn from 5.6 million policy records. This produces 140,000 candidate pairs. We then employ an LLM-as-a-judge to filter high-quality examples, followed by in-context polishing using three expert-written samples as references. This process yields a final set of 639 new pairs. Finally, we fine-tune three models on Sci2Pol-Corpus: LLaMA-3.1-8B, Gemma-12B, and Gemma-27B. Fine-tuning leads to consistent performance improvements across Sci2Pol-Bench. Notably, after fine-tuning, Gemma-27B surpasses the much larger GPT-4o and DeepSeek-V3 (671B). These demonstrate the effectiveness of our corpus in bridging the gap between science and policy.

Lawmakers vary considerably in how effectively they advance their priorities through Congress. However, the actual proposal-writing strategies undergirding these differences have remained largely unexplored, due to measurement and methodological difficulties. These obstacles have included prohibitively small sample sizes, costly data requirements, and strong theoretical assumptions. In this paper, we address these obstacles and analyze the proposal strategies of effective lawmakers directly, using original measures of the spatial locations of congressional bill proposals and associated status quos generated by jointly scaling cosponsorship, roll-call, and interest group position-taking data for 1,007 bills from the 110th through 114th Congresses. Because interest groups take positions on bills before they receive votes, our measures cover many bills that die in committee, permitting comparisons between successful and unsuccessful bills. We demonstrate that legislative advancement favors moderate proposals over partisan ones, and that effective lawmakers are those who make proposals closer to the median even at the expense of their preferred policy.

This article investigates whether campaign contributions and lobbying are complementary, substitutive, or distinct forms of organizational political engagement. Our study reveals minimal overlap between organizations that engage in lobbying and those that make campaign contributions despite the perception that these activities are interchangeable forms of "money in politics." Using comprehensive contribution and lobbying report data from 1998 to 2018, we find that most politically active organizations focus exclusively on either lobbying or making campaign contributions. Only a small percentage of organizations engage in both activities. This finding challenges the assumption that these forms of political activity are inherently linked. The majority of organizations engaged in political activity do so exclusively through lobbying. However, the top lobbying groups spend the most money and almost always have affiliated political action committees (PACs). Most lobbying money is spent by a small number of big spenders—organizations that also have affiliated PACs. Organizations that both lobby and make campaign contributions tend to be well resourced and rare.

We examine the strategic use of contract and in‐house lobbyists by interest groups in response to shifting policy agendas. The role that lobbyists play in the policy process changes based on the policy agenda. Most of the time, subsystems manage small changes to public policy, rewarding actors with long‐term relationships. Organizations with a deep interest in the issue area maintain permanent lobbying presences, earning some degree of privilege over policymaking. However, when the broader macropolitical agenda lurches toward the issue, new participants become involved. New participants often lack the lobbying expertise of the in‐house lobbyists of established actors. Contract lobbyists play a critical role in providing spare capacity on‐demand, allowing participants not normally involved in subsystems to lobby. They also allow the best‐resourced actors, who may employ a long‐term lobbying presence, to further expand it when necessary. We test this theory using a new dataset of the lobbying content of 1,370,396 bill mentions in U.S. lobbying disclosure reports by 11,842 organizations from 2006 to 2016. We compare their policy agenda to that of the U.S. Congress. We find strong evidence that organizations hire contract lobbyists to respond to brief moments of agenda setting while permanent in‐house lobbyists have a more stable agenda.

This article examines the policy topics and theoretical debates found in Policy Studies Journal (PSJ) articles over the last three decades. PSJ is the premier journal for scholars studying policy processes, seeking to create generalizable theories across the spectrum of specific policy areas. To examine trends in PSJ over time, we collected 1314 abstracts from PSJ articles. We identified abstracts that mention major theories of the policy process and stages of the policy cycle. Next, we measured their policy content using the Comparative Agendas Project codebook, as well as their citations in academic journals and policy documents. We then explore these data, finding that changes in the content of PSJ articles over time correspond with other trends in the policy process field and PSJ's increased impact factor.

Campaign donors and corporate interests have greater access to Congress, and the legislative agenda and policy outcomes reflect their preferences. How this privileged access converts into influence remains unclear because petitioner-legislator interactions are unobserved. In this article, we report the results of an original survey experiment of 436 congressional staffers. The vignette manipulates a petitioner's identity, the substance of the request, and the supporting evidence being offered. We test how likely staff are to take a meeting, to use the information being offered, and to recommend taking a position consistent with the request, as well as whether they perceive the request to be congruent with constituent preferences. Donors and lobbyists are no more likely to be granted access than constituents, but staffers are more likely to use information and to make legislative action recommendations when the information source is an ideologically aligned think tank. Subgroup analysis suggests these effects are particularly strong among ideological extremists and strong partisans. And, information offered by aligned think tanks are thought to be representative of constituent opinion. Our results reveal the partisan and ideological predispositions that motivate legislative action that is more costly than merely granting access.

The COVID-19 pandemic has induced a system-wide economic downturn disrupting virtually every conceivable economic interest. Which interests do legislators publicly champion during such crises? Here, we examine mentions of particular industries across thousands of press releases issued by members of Congress during the onset of the COVID-19 pandemic (January to June 2020). We show that members consistently emphasized interests significant to their constituency and party network, but less so their direct campaign contributors or ideological allies. This suggests that members believe that they must be seen as good district representatives and party stewards even when national crises could justifiably induce them to favor any number of interests.

Think tanks have become central players in the political and policy ecosystem of the United States, yet the communication and coordination strategies and connections between them remain relatively unexamined. This paper begins to remedy that by using IRS 990 data to construct and analyze the largest board interlock network among policy planning organizations to date. An analysis of 277 policy research organizations located in Washington D.C. reveals high levels of resource seeking, issue-area coordination and weak evidence of ideological coordination among these substantively important organizations in American politics.

Attribution of cyberattacks has strategic and technical components. We provide a formal model that incorporates both elements and shows the conditions under which it is rational to tolerate an attack and when it is better to assign blame publicly. The model applies to a wide range of conflicts and provides guidance to policymakers about which parameters must be estimated to make a sound decision about attribution and blame. It also draws some surprising conclusions about the risks of asymmetric technical attribution capabilities.

This study provides evidence that geopolitical considerations systematically shape funding evaluations of international collaboration proposals. We examine this dynamic in the consequential context of U.S.–China collaboration. Across two large-scale randomized experiments with U.S. policymakers and U.S.-based scientists, we find substantial and consistent penalties for proposals involving China-based collaborators. Policymakers express much greater unconditional support for proposals with Germany-based collaborators than for otherwise identical proposals with China-based collaborators (68% vs. 28%). Crucially, this penalty is not confined to policymakers: scientists themselves exhibit a sizeable 18 percentage-point gap (48% vs. 30%), despite professional expectations of merit-based evaluation. Much of the difference reflects a shift from unconditional to conditional approval rather than outright rejection. These penalties are remarkably consistent across scientific fields and respondent characteristics, with little evidence of heterogeneity, indicating that they reflect geopolitical rather than domain-specific concerns. Overall, the findings suggest that geopolitics influences gatekeeping judgments in government funding, with broad implications for peer review, scientific norms, and the future of international collaboration in an era of intensifying geopolitical competition.

Policymaking relies on institutions that translate expertise into politically actionable knowledge. Yet little is known about how such knowledge is created, structured, and politically polarized, or how science shapes these processes. Using 2 million U.S. policy documents and nearly 1 million scientific citations from more than 200 think tanks and 100 government organizations, we analyze the supply side of science in policymaking: the production of policy knowledge. We find that think tanks are the dominant suppliers of science-based policy knowledge to government, but that this production has become increasingly politically polarized, driven primarily by growing insularity among left-leaning institutions. We further find that science is linked to lower polarization, as policy documents grounded in scientific evidence, especially those citing high-impact science, are less ideologically segregated and occupy more central positions in policy knowledge networks. These results reveal the potential of scientific expertise to shape and, at times, bridge the ideological landscape of policy knowledge production. Amid rising political polarization and the growing role of science in policymaking, understanding how policy knowledge is produced and how scientific expertise relates to its ideological structure is essential to strengthening the informational foundations of democratic governance.

Scientific institutions are designed to produce and evaluate claims through shared evidentiary and disciplinary standards, yet they are embedded in an increasingly polarized society. This fundamental tension raises the question of whether scientists’ partisan alignment is associated with the production, circulation, and uptake of scientific knowledge. Here we integrate campaign-finance records, faculty rosters, voter-registration data, an original survey of approximately 7,000 active U.S. scientists, bibliometric data, and measures of uptake in policy documents, patents, and news coverage to map the partisan composition of the scientific workforce, the knowledge it produces, and its uptake within and beyond science. We find that U.S. science is politically asymmetric without being ideologically homogeneous: scientists who make political contributions are overwhelmingly Democratic-aligned, and this asymmetry has widened over time, yet the broader workforce includes many Independents and substantial ideological heterogeneity within partisan groups. Among scientists with strong, behaviorally observed partisan alignments, Democratic- and Republican-authored papers are systematically distinguishable in semantic content across all fields, even after conditioning on fine-grained topic and year. Differences in total scientific citation rates are modest and have narrowed over time, whereas citations occur more often within the same partisan author communities than expected under topic- and year-preserving null models. Uptake beyond science also varies by partisan authorship and domain, with the clearest directional selectivity among ideological think tanks. These findings reveal an uneven profile of partisan association across U.S. science, visible in published content and directional attention, but comparatively limited in aggregate scientific recognition. This profile is consistent with a scientific system in which shared evaluative standards coexist with political structure in the production and circulation of knowledge, with implications for science’s capacity to remain a common evidentiary resource in a polarized society.