Machine learning has exhibited substantial success in the field of natural language processing (NLP). For example, large language models have empirically proven to be capable of producing text of high complexity and cohesion. However, at the same time, they are prone to inaccuracies and hallucinations. As these systems are increasingly integrated into real-world applications, ensuring their safety and reliability becomes a primary concern. There are safety critical contexts where such models must be robust to variability or attack and give guarantees over their output. Computer vision had pioneered the use of formal verification of neural networks for such scenarios and developed common verification standards and pipelines, leveraging precise formal reasoning about geometric properties of data manifolds. In contrast, NLP verification methods have only recently appeared in the literature. While presenting sophisticated algorithms in their own right, these papers have not yet crystallised into a common methodology. They are often light on the pragmatical issues of NLP verification, and the area remains fragmented. In this paper, we attempt to distil and evaluate general components of an NLP verification pipeline that emerges from the progress in the field to date. Our contributions are twofold. First, we propose a general methodology to analyse the effect of the embedding gap – a problem that refers to the discrepancy between verification of geometric subspaces, and the semantic meaning of sentences which the geometric subspaces are supposed to represent. We propose a number of practical NLP methods that can help to quantify the effects of the embedding gap. Second, we give a general method for training and verification of neural networks that leverages a more precise geometric estimation of semantic similarity of sentences in the embedding space and helps to overcome the effects of the embedding gap in practice.

Do ideologically extreme candidates enjoy fundraising advantages over more moderate candidates? Extant work documents a relationship between candidates’ positions and campaign contributions subnationally and in donor surveys, yet identification challenges have hampered investigation in the congressional context. I employ a close primaries regression discontinuity design to examine how “as-if random” nominations of extreme versus moderate House candidates influence general election contributions from individual donors and corporate political action committees (PACs) from 1980 to 2020. Results at both the nominee and contributor levels demonstrate that corporate PACs financially penalize extremists, while individual donors respond similarly to extreme and moderate candidates. These findings contribute to ongoing debates regarding the extent and nature of campaign contributors’ role in congressional polarization.

This article examines the death of Colin Roach in Stoke Newington Police Station, Hackney, in 1983, and explores the emotional politics of the campaigns that followed his death. These campaigns were focused on both determining the circumstances of Roach's death and highlighting tensions between the police and the Black community of Hackney. Using hitherto unpublished archival sources, local newspapers, and visual sources, the article documents racial politics in Hackney in the early 1980s and examines the relationship between race and policing at that time. The article argues that the experience and expression of grief and anger were critical to understanding the political problem of race and policing in London in the 1980s, to forming and mobilizing political communities, and to interrogating the power of the state. The article also argues that a critical element of the emotional economy around race in Hackney in 1983 was the indifference and lack of empathy of the police in Stoke Newington to ethnic minority communities. This lack of empathy not only illustrated the problem of race within the police force at this time but further fueled local campaigns to make the police accountable. This links the Roach case to a later turning point—the 1999 Macpherson inquiry into the murder of Stephen Lawrence, which characterized the Metropolitan Police as institutionally racist.

We investigate the role of visual attention in risky choice in a rich experimental dataset that includes eye-tracking data. We first show that attention is not reducible to individual and contextual variables, which explain only 20% of attentional variation. We then decompose attentional variation into individual average attention and trial-wise deviations of attention to capture different cognitive processes. Individual average attention varies by individual, and can proxy for individual preferences or goals (as in models of “rational inattention” or goal-directed attention). Trial-wise deviations of attention vary within subjects and depend on contextual factors (as in models of “salience” or stimulus-driven attention). We find that both types of attention predict behavior: average individual attention patterns are correlated with individual levels of loss aversion and capture part of this individual heterogeneity. Adding trial-wise deviations of attention further improves model fit. Our results show that a decomposition of attention into individual average attention and trial-wise deviations of attention can capture separable cognitive components of decision making and provides a useful tool for economists and researchers from related fields interested in decision-making and attention.

Emotion recognition, the ability to interpret others’ emotional expressions and infer mental states, is crucial for caregiver–child interactions. The ability to accurately recognize infant emotions may facilitate attuned and responsive caregiving. Across two studies, we validate a novel measure to assess the recognition of infants’ emotions (Reading the Mind in Infant Eyes Test [RMIET]) and investigate how this ability relates to observed caregiving. Study 1 examined item-level performance in 55 infant mental health experts and 100 undergraduate students. Study 2 examined RMIET scores in 133 pregnant people and their later caregiving when their children were 18-month-old. In Study 1, agreement was high among both mental health experts (ICC = .82) and undergraduate students (ICC = .93), providing evidence of the content validity of the RMIET. In Study 2, scores assessing the recognition of adult and infant emotions were positively correlated (r = .22, p = .012). After accounting for covariates, RMIET scores were statistically significantly associated with higher sensitivity and warmth and lower negative regard. Taken together, these studies provide preliminary evidence of content and predictive validity for the RMIET.

Classroom discussions of current events and controversial topics can devolve into unproductive and highly charged debates. This article describes an in-class exercise used to foster respect during difficult conversations by encouraging students to design rules for discussions and guidelines to create a safe space for dialogue. This activity relies on three underlying principles: trust, empowerment, and empathy. These principles can be integrated into a broader pedagogical approach that emphasizes a democratic classroom and active learning. Student feedback shows that the intervention can be useful for promoting respectful and engaging discussions during moments of tension and polarization. However, an emphasis on civility also may undermine the diversity of opinions and require respecting students’ silences.

President Theodore Roosevelt once called the Chautauqua movement “typical of America at its best.” Arguably, no public humanities effort in the United States has been as popular or influential as the Chautauqua movement. Those seeking educational, religious, and cultural fulfillment have flocked to the lakeside retreat in western New York since 1874. From lectures and book clubs to theater and debates, Chautauqua has provided visitors with endless opportunities to engage with the arts, humanities, and cultural studies to help them better understand their world, themselves, and each other. Efforts to expand the experience beyond the summer include the adult education movement the Chautauqua Literary and Scientific Circle, one of the nation’s longest-running book clubs. Shrewd businessmen, and communities who did not see themselves represented in the Mother Chautauqua, also opened their own copycat offshoots across the country. As the pillar of “edutainment” and tourist destination wraps its 150th season, are the utopian visions of the founders enough to sustain it for another 150 or is a new model needed for a nationwide public humanities movement?

This article draws a comparison between US border policies in the 2020s and the policies implemented by the British colonial regime in 1940s ‘Aden to dissuade Jewish immigration. It makes an original argument, based on documents from the British colonial archives and Jewish philanthropic sources, that the plunder of Jewish migrants was a consequence of British policy, and not, as scholars have sometimes assumed, a vaguely-defined “anarchy” in the Aden Protectorate sultanates (today, southern, and eastern Yemen). The history of British immigration policy – and the unofficial incorporation of both environmental and human forces into the project of dissuading Jewish migration – bears a striking resemblance to American policies in recent years. The perils of the Darién Gap and other deadly routes and the concentration of migrants in dangerous conditions on the US–Mexico border de facto incorporate the jungle, the desert, and criminal syndicates into the border regime’s efforts to disincentivize migration. A look at the archival record of a parallel story in 1940s Yemen/‘Aden allows us to glimpse the construction of policies that utilize unofficial actors and factors (from bandits to the hot desert sun) in a border regime’s campaign of terror against (potential) migrants. The article demonstrates the value of historical comparative cases for understanding the policies of governments today. Scholars of current events lack access to the intelligence reports, correspondence, and other once-classified documents available to historians, which allow for a fuller understanding of the ways in which similar policies have been developed and implemented.

The development of new papaya cultivars with high genetic potential for production, combined with quality traits that meet the demands of emerging markets, facilitates the expansion of the genetic base, reduces production costs, and broadens papaya cultivation. This study continues the largest Brazilian papaya breeding program, a partnership of over 25 years between UENF and Caliman Agrícola S.A. The objective was to evaluate the ability of inbred lines to generate hybrids with market potential in Brazil and for export. A total of 62 hybrids were obtained through topcross strategy. The lines were evaluated based on their specific combining ability (SCA), and the hybrids were analyzed through estimates of functional and varietal heterosis (VH) using three widely cultivated commercial varieties in Brazil: ‘UENF/CALIMAN 01’, ‘Tainung 01’, and ‘UC10’. Promising lines were identified for both hybrid creation and use as commercial varieties, exhibiting desirable traits for domestic and international markets, such as high fruit firmness and elevated soluble solids content (the lines UCLA08-053 and UCLA08-087 with the ‘Intermediate’ pattern and the lines UCLA08-066, UCLA08-122, and UCLA08-080 with the ‘Formosa’ pattern). Several hybrids, including H23, H26, H51, and H89 from the ‘Intermediate’ type and H4, H9, H19, and H68 from the ‘Formosa’ type, outperformed their parental lines and commercial varieties. These genotypes demonstrated superior SCA and VH compared to commercial controls, highlighting their strong genetic potential for the production market, increasing the shelf life of the fruits during storage and transportation, and allowing them to travel long distances without compromising fruit quality.

Healthcare systems are increasingly exploiting the advantages of Internet of Things technologies: cloud-connected devices with perceptive sensors can gather very accurate health data from people even if they do not get to the hospital or private clinics. For potential innovators of new medical IoT devices, the legal framework applicable was until now limited to the application of the General Data Protection Regulation and the Medical Devices Regulation.

This paper will investigate what will happen when medical IoT-generated data are shared to create new products or services according to the framework now depicted by the Data Act and the European Health Data Space.

Given that the EHDS and the Data Act are both aimed at facilitating the secondary use of (health) data, the contribution will compare the two processes set up to establish a roadmap to solve health-data sharing theoretical and practical queries.

This article seeks to cast a critical eye on musical modernism through the experiences of its percussionist practitioners. It charts the origins and accepted truisms of percussion ontology as it is understood through the modernist sensibility, and demonstrates how certain modernist assumptions have been inherited by many contemporary practitioners. Some of these individuals’ resulting expressions of grief, anger, and sadness in the wake of modernism's waning are presented, and a reparative reading of modernist percussion that seeks to make the repertory inhabitable and sustaining is instead offered. This practice is illustrated through a feminist and performer-led analysis of Karlheinz Stockhausen's Kontakte (1958–60), for piano, percussion, and tape. It is ultimately argued that performer knowledge and affective attachment is essential to understanding modernism's history and aesthetics, as well as its place in the contemporary moment.

English kings from the late ninth century onwards were sometimes presented as kings of Britain. While this is widely attested, less studied is how the English positioned themselves in relation to Britain – or rather, how English they considered Britain to be. A number of late Old English poems engage deeply with the problem of Britain, but have so far escaped attention in this broader context. In multiple cases, late Old English poems elide non-English peoples and present English dominion as normative and unmarked throughout Britain. This is particularly prominent in the Old English poetic Menologium, which is studied here in detail. The treatment of Britain in late Old English poetry also leads to a reinterpretation of one of the most well-known Old English historical poems, The Battle of Brunanburh. The battle becomes a defence of a particularly English version of Britain, with the invading enemies being successfully driven out.

This work investigates the online machine learning problem of prediction with expert advice in an adversarial setting through numerical analysis of, and experiments with, a related partial differential equation. The problem is a repeated two-person game involving decision-making at each step informed by $n$ experts in an adversarial environment. The continuum limit of this game over a large number of steps is a degenerate elliptic equation whose solution encodes the optimal strategies for both players. We develop numerical methods for approximating the solution of this equation in relatively high dimensions ($n\leq 10$) by exploiting symmetries in the equation and the solution to drastically reduce the size of the computational domain. Based on our numerical results we make a number of conjectures about the optimality of various adversarial strategies, in particular about the non-optimality of the COMB strategy.

In this short report, the challenges and lessons learned from implementing scientific research in primary care are discussed. It highlights the complexities of conducting studies in primary care, where ‘Lasagna’s Law’ rules too often. Using the CONCRETE trial – a pragmatic multicenter implementation trial – as an example, eight key elements are identified as important factors for successfully conducting scientific research in primary care, such as optimizing digital processes and improving engagement.

Understanding and tracking societal discourse around essential governance challenges of our times is crucial. One possible heuristic is to conceptualize discourse as a network of actors and policy beliefs.

Here, we present an exemplary and widely applicable automated approach to extract discourse networks from large volumes of media data, as a bipartite graph of organizations and beliefs connected by stance edges. Our approach leverages various natural language processing techniques, alongside qualitative content analysis. We combine named entity recognition, named entity linking, supervised text classification informed by close reading, and a novel stance detection procedure based on large language models.

We demonstrate our approach in an empirical application tracing urban sustainable transport discourse networks in the Swiss urban area of Zürich over 12 years, based on more than one million paragraphs extracted from slightly less than two million newspaper articles.

We test the internal validity of our approach. Based on evaluations against manually automated data, we find support for what we call the window validity hypothesis of automated discourse network data gathering. The internal validity of automated discourse network data gathering increases if inferences are combined over sliding time windows.

Our results show that when leveraging data redundancy and stance inertia through windowed aggregation, automated methods can recover basic structure and higher-level structurally descriptive metrics of discourse networks well. Our results also demonstrate the necessity of creating high-quality test sets and close reading and that efforts invested in automation should be carefully considered.

The historiography of liberalism has taken a theological turn. Many scholars now trace the origins of liberal thought to Christian orthodoxy, with its emphasis on the radical equality of humanity under the absolute sovereignty of God. Others trace it to the heresy of Pelagianism, with its emphasis on the radical freedom of humans to choose between good and evil under the rationalistic judgment of God. Focusing on a classic expression of early-modern liberalism, Benjamin Franklin’s Autobiography, this article questions the theological turn: Franklin’s thought, I argue, rejects both Augustinianism and Pelagianism, along with their underlying metaphysical presuppositions concerning human liberty.