Suicide is a significant global public health concern, and conventional clinical risk assessment is constrained by workforce availability, scalability, and clinician variability. The classic suicidality risk evaluation is largely dependent on clinical judgment, which, although helpful, can demonstrate a ceiling effect in its predictive validity. AI-driven chatbots, conversational systems that engage users in real-time natural language dialog, have emerged as candidate tools for augmenting suicide risk detection and prevention. This narrative review aimed to: evaluate the performance of AI-driven chatbots in detecting and assessing suicidal ideation relative to clinical benchmarks; examine the effectiveness of chatbot-based interventions for suicide prevention; and identify ethical, cultural, and implementation challenges limiting clinical translation. Six electronic databases were searched, with the initial search conducted in 2024 and updated in 2026 through targeted monitoring, with no upper cutoff date applied. A thematic narrative synthesis approach was applied across five domains. Eleven primary studies met eligibility criteria and were included in the synthesis. Chatbot-based risk assessment showed adequate response alignment with expert judgment at the extremes of suicide risk, but consistently failed to distinguish intermediate risk levels across multiple model families. Across 29 tested commercial chatbot agents, none met the criteria for an adequate suicidal crisis response. A clinically designed, framework-anchored chatbot achieved high efficacy across six outcome domains. Three percent of social chatbot users reported halted suicidal ideation, and a purpose-built clinical chatbot in emergency department settings significantly improved evidence-based care delivery. Systematic risk severity underestimation and the absence of cross-cultural evaluation were identified across all studies. AI-driven chatbots show potential as adjunctive tools across the suicide care continuum. Clinically designed, evidence-anchored chatbots demonstrate feasibility and meaningful benefit; commercially deployed chatbots without clinical validation demonstrate near-universal crisis response inadequacy. Mandatory clinical validation prior to public release, clinician oversight, and crisis system integration are prerequisites for responsible deployment.

Gauging the extent of public acceptability of reforms is an important concern for policymakers. Timely insights into public perceptions can illuminate how reforms are received and how attitudes evolve over time. In this study, we build on the OECD’s Public Acceptability Tool, a framework encompassing four key dimensions of reform acceptability—Economic, Fairness, Behavioural, and Process—to evaluate the public acceptability of policy reforms. We take the 2023 French pension reform as a relevant case study, using online media articles and parliamentary speeches as indicators of discourse surrounding the reform. Using word embeddings, we classify these texts according to the four dimensions and apply matrix factorisation topic algorithms to uncover the latent themes within each. Our analysis shows that the Process dimension dominated media coverage during the discussion and legislative phases of the reform, consistent with previous literature on pension reforms. In contrast, no particular dimension was predominant in parliamentary speeches, suggesting a mismatch between policy and public debates. Finally, we identify the main topics driving public discussion within each dimension, highlighting notable differences between media narratives and parliamentary discourse that offer further insight into the dynamics of public acceptability.

Printed media can shape its audience’s political preferences. Research on governments influencing their narrative focuses on censorship of political scandals but largely ignores day-to-day political content. We analyze opinion columns from Mexico, a country with a democratic culture where outright censorship is unfeasible, and yet there is journalistic evidence on the government “sponsoring” the media to communicate messages with a less negative tone. We estimate the negative tone of over 200,000 opinion pieces from Mexico’s eight most prominent newspapers using sentiment analysis. We propose a novel metric of media capture that conveys the over-favoring of publicity spending by the government in news outlets. We find that captured media are less negative about the incumbent president and more negative about their main political rival, suggesting media organizations adapt their content to their advertisers’ preferences. The results are validated by rare qualitative evidence that uncovers the blacklisting of media deemed too negative of the incumbent.

Natural language processing (NLP) technologies increasingly shape public life, yet their deployment for social good remains unevenly distributed across domains, languages, and geographies. This piece inaugurates the NLP for Social Good column in this journal. In this piece, I map the current state of NLP for Social Good (NLP4SG) across nine application domains. The picture that emerges is one of striking imbalance: AI harms, inclusion, and digital violence attract the bulk of research attention, while poverty, peacebuilding, and environmental protection remain critically underexplored. I argue that the field must address three structural gaps, domain coverage, linguistic diversity, and evaluation methodology, if NLP is to fulfil its potential as a force for equitable social progress. The piece concludes with five directions that I believe will define the next chapter of NLP4SG research.

This study addresses the challenges of performing narratological analysis on low-resource languages, with a focus on Old Church Slavonic. Understanding the roles, interactions, and networks of persons is central to narrative analysis, yet such investigation is hindered by the scarcity of experts and the limited availability of annotated resources. We explore both established natural language processing (NLP) methods and large language models (LLMs) for analyzing pre-modern Slavic Lives of Saints, including several Slavic versions, the Greek original, and an English translation. Pre-modern Slavic texts pose particular difficulties due to rich morphology, orthographic variation, and limited standardization, which complicate the application of both traditional NLP tools and off-the-shelf LLMs. Through experiments using annotated and non-annotated ground truth data, we demonstrate that while conventional NLP methods often reach their limits on such low-resource, highly variable texts, LLMs provide complementary capabilities that can support narratological insights, especially in tracking persons and their interactions, albeit with important caveats regarding accuracy and coverage.

Democratic resilience is as much about the narratives of our nation we affirm, as the institutions that enshrine our values and laws, a fact re-affirmed by scholarship across many branches of social science in recent decades. For policymakers and quantitative social scientists, analysing or tracking public discourse through the lens of narrative and framing has historically involved the annotation of texts by hand, placing severe limitations on the scale and modality of discourse under inquiry. Yet, a revolution is at hand—a transformer revolution: first arising in computer science, and now enabling a new kind of automated narrative analysis at scale, transformers are opening up new horizons for the tracking of public narratives of democratic resilience. Here, we: formulate a conceptual framework linking computational language methods to democratic resilience analysis; introduce transformer-based artificial intelligence (AI) methods as a third wave in natural language processing technology; and demonstrate two practical applications of transformer methods to democratic narrative analysis. Finally, we conclude by contributing data and research recommendations which flow naturally from the opportunities unlocked by transformer methods for public stakeholders who wish to see these opportunities realised. Together, we suggest that, perhaps for the first time, the “holy grail” of the quantitative social scientist is near: the ability to identify, accurately, and efficiently, nuanced narratives in text, at scale.

Women politicians report that social media abuse harms their personal and professional lives. However, prior text-based research finds that men receive more general online hostility than women – except among the most visible politicians. I hypothesize that backlash to perceived gender-role violations – such as public visibility – will include distinctly gendered content, such as slurs and references to appearance. Using a novel and replicable method, I analyze hostile and gendered language in three million social media mentions of US state representatives. I find that hostility towards visible women differs from men in content, not volume. Visible women face similar volumes of generic hostility but twice as much gender-specific abuse as men. This pattern holds across two alternate measures of perceived conformity to traditional gender roles: legislator tone and the presence of women in the chamber. Incorporating gendered content into text-based analyses reconciles discrepancies between observational and self-reported data and validates women politicians’ reports.

The rise of health care AI raises concerns over whether patent disclosure supports reproducibility and legal validity. This study analyzes 865 granted medical AI patents (2015–2025) from the US, China, and the EU using a five-dimensional framework (algorithm transparency, training data accessibility, model reproducibility, result verifiability, and mathematical support) implemented through NLP-assisted expert scoring. Results suggest limited technical transparency; approximately 40% of patents score zero in at least two dimensions. Performance varies significantly: algorithm transparency is relatively strong (>60% score 2), while training data accessibility is less prevalent (4.6% score 2) and mathematical support is frequently omitted (39.4% score 0). Statistical testing indicates US patents significantly outperform Chinese patents (p < 0.001), while EU results remain exploratory (N = 31, mean 6.2). These patterns appear associated with institutional factors, strategic applicant behaviour, and technical complexity. Such limitations may pose risks to enforceability and market development, highlighting the need for targeted disclosure improvements. This study contributes a replicable framework for translating legal standards into measurable indicators, providing cross-jurisdictional evidence to guide examination, litigation, and policy refinement in medical AI governance.

Construction safety inspections typically involve a human inspector identifying safety concerns on-site. With the rise of powerful vision language models (VLMs), researchers are exploring their use for tasks such as detecting safety rule violations from on-site images. However, there is a lack of open datasets to comprehensively evaluate and further fine-tune VLMs in construction safety inspection. Current applications of VLMs use small, supervised datasets, limiting their applicability in tasks they are not directly trained for. In this article, we propose the ConstructionSite 10 k, featuring 10,000 construction site images with annotations for three inter-connected tasks, including image captioning, safety rule violation visual question answering (VQA), and construction element visual grounding. Our subsequent evaluation of current state-of-the-art large pre-trained VLMs shows notable generalization abilities in zero-shot and few-shot settings, while additional training is needed to make them applicable to actual construction sites. This dataset allows researchers to train and evaluate their own VLMs with new architectures and techniques, providing a valuable benchmark for construction safety inspection.

This article examines the use of neural networks in electromechanical sound art and music, where sound is materially enacted through physical means such as motors, solenoids, and physical resonators. It begins with a survey of documented works, outlining a range of current strategies and discussing how technical, material, and performative factors influence their design. Identifying natural language processing as underexplored in this domain, a practice-based case study, Seven Studies for Electric Motors, develops one such language-based approach. The project embeds a small language model for real-time sentence generation, extracts syntax structures, and translates these into patterns of motor-driven sound. Taken together, the survey and case study offer a picture of how machine learning has been integrated into electromechanical practices over the past decade and point to possible directions for further work.