Papers

Purpose: Governments increasingly fund indigenous foundation models to strengthen national AI capability, digital sovereignty, and multilingual computing. This paper assesses India's foundation-model ecosystem and examines whether apparent capability gaps in public benchmark evidence may also reflect gaps in evaluation maturity. Approach: The paper presents a structured, benchmark-based comparative assessment of Indian foundation models against global frontier and comparable-scale models across eight capability domains: general-purpose reasoning, coding and software engineering, agentic AI and computer use, cybersecurity, vision and image understanding, video and multimodal understanding, scientific research, and Indic language capability. Using only publicly reported results, it proposes an exploratory four-dimension Benchmark Maturity Index (BMI), scoring each domain on standardization, participation, independent verification, and national Findings: Indian models achieve strong scores on established benchmarks such as MMLU and MATH-500. However, these are now widely regarded as saturated, and frontier developers no longer report them. Indian models participate far less frequently in newer, agentic, and domain-specialized evaluations, and participation is highly uneven across organizations. Sarvam AI reports the broadest coverage by a substantial margin. The BMI refines, and in some cases revises, the maturity judgments a purely descriptive review would produce. Practical implications: Many apparent capability gaps cannot be distinguished, on available evidence, from evaluation-ecosystem gaps, with direct implications for how national AI programs should design monitoring and funding criteria. Originality: The paper proposes BMI as a reusable instrument for scoring evaluation-ecosystem maturity at the domain level and demonstrates its application to the Indian foundation-model ecosystem.

The deployment of large language models (LLMs) in mental health contexts raises questions about the relationship between clinical safety and environmental cost. In this paper, we examine this relationship by combining K-Bench clinical safety scores with EcoLogits life-cycle assessment estimates across 47 supported model configurations. We evaluate model performance and environmental impact across four dimensions: energy use, carbon emissions, water consumption, and abiotic depletion. The results indicate a non-linear trade-off at the upper end of the safety distribution: a 2.61 percentage-point increase in clinical safety score corresponded to an approximately 60-fold increase in estimated energy use per million output tokens. Row-level analyses further suggest that additional test-time compute did not consistently improve clinical safety and, in some configurations, was associated with lower clinical safety scores. These findings suggest that relying solely on larger models or additional inference-time computation may be an inefficient strategy for improving safety in therapeutic AI systems. We discuss the implications for sustainable deployment and highlight dynamic model selection, including model cascading, as a potential approach for reducing environmental impact while preserving clinical performance in higher-risk cases.

As large language models enter everyday information pipelines, understanding how they handle sensitive topics matters as much as understanding whether they handle them at all. We study this question through a large-scale, systematic experiment using restricted versus unrestricted books as a controlled testbed: 40,800 query-response pairs, 400 books, 17 prompt designs, and six frontier models spanning six AI providers (Claude Sonnet 4.5, GPT-4o, Gemini 2.5 Flash, DeepSeek-V3, Qwen-Plus, and Grok-4.1-Fast). Our restricted set is drawn from the American Library Association's Most Challenged Books records (2000-2023); we use restricted rather than banned throughout because the ALA documents formal challenges-requests to remove or restrict access-which do not always result in outright bans. Our central finding is a zero-refusal phenomenon: modern LLMs decline to discuss restricted books in only 0.07% of cases, effectively invalidating the premise of jailbreaking research for this content class. Differentiation occurs instead through warning language (+8-15 percentage points, p < 0.001) and hesitation markers (+2-5 pp), with sexual content mention rate as the strongest individual signal (+33-52 pp). We further identify systematic differences between providers and show that prompt framing alone shifts the warning-rate gap by up to 19 pp. These results indicate that LLM content policy has shifted from binary refusal toward calibrated, context-sensitive disclosure-a finding that holds consistently across Western and Chinese AI providers.

Deployed foundation models are often not static systems, with providers able to modify system behavior through fine-tuning, classifier updates, system prompt revisions, retrieval changes, and routing changes. These updates can be made silently – that is, without public disclosure, a version increment, or re-evaluation. Such silent updates challenge a core assumption behind current AI governance frameworks that an externally verifiable chain of custody links the model referred to in evaluation results or a system card to the model served to users. In this paper, we examine post-deployment disclosure practices across first-party API providers and inference hosts to establish the extent to which a chain of custody exists in practice. We find that providers commonly publish substantial safety documentation, including quantitative evaluations and version-specific reports, but no provider in our sample published information allowing an external party to verify that the artifact being served is the same one referred to in this documentation. We introduce the Silent Updates Scorecard, a public instrument for measuring post-deployment disclosure practices across providers and hosts, and preliminary results for a sample of nine first-party API providers and seven third-party inference hosts. We also propose a Three-Part Behavioral Trigger System for determining when post-deployment modifications to a system motivate disclosure or re-evaluation obligations.

The growing role of AI-generated content and AI-enabled systems in public communication has led regulators to demand clear disclosure of content provenance and AI involvement. But the effects of such disclosures remain uncertain. We test two disclosure approaches in their impact on an AI chatbot's persuasive appeal. In a preregistered experiment, 1,500 UK adults held a short conversation with a persuasive chatbot about one of 60 policy issues. The chatbot was identical for everyone. We randomized the disclosure that people received: nothing (control), a prominent disclosure that they were interacting with an AI (T1), or that disclosure plus the chatbot's persuasive intent and instructions (T2). The chatbot shifted attitudes by 12.6 points on a 100-point scale in the control group. The AI-identity disclosure was practically equivalent to no disclosure, with a 13.1-point shift, whereas the additional intent disclosure cut the persuasive effect roughly in half to 6.3 points. It also made participants view the campaign's methods as less acceptable and support stronger penalties against it. For direct chatbot interactions, transparency about AI identity alone does not meaningfully impact its influence. While current rules emphasize what a system is, our results show why the regulation of persuasive AI must also address what the system is trying to do.

This study proposes that firms move along an "organizational technology ladder": adopting one technology transforms hiring and work processes and builds skills and organizational capital that change the cost of adopting subsequent technologies. I study how firms' adoption of remote work technology during the COVID-19 period shaped later uptake of generative AI. Using U.S. job-posting data and an instrumental-variables strategy based on predicted differences in labor-market pressure to offer remote work, I estimate that a 10 percentage point increase in remote hiring in 2021-2022 increases the share of job postings mentioning generative AI in 2023-2024 by 0.4 percentage points across firms and 0.7 percentage points across occupations within firms. I provide evidence on mechanisms consistent with a technology-ladder channel: remote work adoption shifts hiring toward technical and managerial capabilities that predict faster conversion of generative AI exposure into adoption. Firms with return-to-office mandates—interpreted as revealing low remote productivity—exhibit a substantially larger response of generative AI adoption to remote work, consistent with an organizational frictions channel.

The convergence of artificial intelligence (AI), Industrial Internet of Things, cyber-physical systems, and advanced robotics is reshaping manufacturing faster than engineering curricula can adapt, widening the gap between the competencies required on the shop floor and those delivered by traditional engineering and technology education. This paper proposes a Workforce Readiness Level (WRL) framework, which adapts the Technology Readiness Level scale into nine progressive competency stages and a four-pillar rubric, digital and AI literacy, cyber-physical systems fluency, human-machine collaboration, and data-driven decision making, aggregated through a composite stage score and a cohort-level workforce-readiness index under a “no-thin-pillar” rule. The framework is instantiated at a university smart-manufacturing teaching laboratory and draws on 89 sponsored capstone projects delivered over four semesters, four of which are analyzed in depth. Four pillars jointly span the relevant ABET student outcomes. Across the highlighted cohorts the workforce-readiness index ranged from 5.2 to 6.4, and the no-thin-pillar rule was diagnostically informative in three of the four cases and the binding certification constraint in one, repeatedly surfacing cyber-physical and data-driven-decision gaps concealed behind strong analytics profiles; advancement to the highest stages was gated by industry-embedded experience rather than additional coursework. WRL offers educators, accreditation bodies, and regional workforce systems a common, evidence-based instrument for diagnosing and advancing workforce readiness; future work will calibrate pillar weights and test reliability and predictive validity.

Large language models are rapidly becoming an important source of political information. This raises a fundamental question: will AI systems support a shared basis for political knowledge, or lead different political groups to rely on increasingly different models? Political sorting can drive model fragmentation if three conditions hold: politically different users select into different models, learning from user feedback pushes those models apart politically, and the resulting differences shape subsequent model choices. We call this self-reinforcing process the Centrifugal Alignment Spiral. We study its components in three steps. First, we draw on a human experiment showing that political identity predicts model choice. Second, we fine-tune language models on synthetic feedback reflecting predominantly Democratic or Republican preferences. Across five independent runs per model family, paired models diverged on 12-41% of unseen survey questions with large partisan gaps, and in every run the differences moved in the expected political direction; for some models, differentiation extended even to issue areas excluded from training. Pooling feedback across groups instead suppressed divergence. Third, an empirically anchored agent-based model shows what follows when political sorting and model adaptation operate together: models attract politically distinct audiences, learn from them, and diverge further. User feedback can therefore turn political sorting among AI users into durable differences between the models on which they rely for political information.

The European Digital Identity Wallet (EUDI Wallet) architecture places material duties on relying parties: they register services and intended uses, authenticate to Wallet Units, validate presentations and trust anchors, and make risk-based status decisions. Wallet certification and the emerging Functional Conformance Assessment Framework provide increasingly structured wallet-side evidence. A different question remains: what independently rerunnable evidence shows that the concrete relying-party verifier version used in a transaction enforced the applicable request, presentation and reliance-decision controls? We conduct an assurance-locus audit of current law, Architecture and Reference Framework (ARF) 3.0.0, ETSI metadata, FCAF scope and three pinned open-source verifier codebases. We then design a 17-rule research profile, a machine-readable evidence receipt and 36 frozen synthetic transactions. Three heterogeneous study-authored implementations (Python, JavaScript and jq) execute 108 cases with zero oracle mismatches and zero cross-path disagreements. The experiment establishes determinism and implementability of the proposed decision model, not product conformance or certification. Source inspection finds substantial protocol-verification mechanisms in all three public codebases but no single audited evidence object joining RP registration and purpose, exact verifier/policy version, transaction verdict and downstream attribute use. We therefore propose an RP-as-system-under-test evidence unit that complements, rather than displaces, wallet certification, registration and supervision. A preregistered census of 26 appointed experts is prepared to test content validity and governance feasibility; recruitment awaits the applicable ethics/data-protection determination.

Climate campaigns are often evaluated through attention and mobilization, but less is known about the well-being language that accompanies them. Whether campaign periods alter positive affect and hope, and whether happiness aligns with action language, remains unresolved. We analysed 364,118 public Twitter/X posts from Earth Day, Earth Hour, Global Climate Action Day and World Environment Day in 19 occurrence-years, using 30-day pre-event, event and post-event windows. A versioned weighted lexical model estimated happiness, future-oriented hope, collective capability, distress and action language. Event-period happiness prevalence was 9.02 percentage points higher than the pre-event baseline , whereas paired occurrence contrasts showed a 10.75-point decline in action language, indicating a happiness–action divergence. The happiness estimate remained positive across composition and text-deduplication checks, but was less precise under a 19-cluster wild bootstrap. Happier source posts had lower odds of an observed matched retweet cascade.

The question of whether artificial intelligence will "destroy jobs" is too coarse to guide economic analysis or institutional design. A job is not an indivisible object, and machine cognition is not a uniform substitute for human labor. This paper develops a task-based and institutionally grounded framework for analyzing generative AI as cheap, scalable, and fallible cognition. The relevant margins are exposure, adoption, verification, question selection, workflow redesign, demand elasticity, apprenticeship, and rent allocation. We distinguish the technical reach of large language models from equilibrium labor-market displacement by introducing a task vulnerability index and an adoption condition that makes verification, liability, trust, and governance explicit. We then model occupations as governance bundles rather than task lists, firms as architectures of distributed intelligence, and labor-market effects as a balance among task compression, scale expansion, new human work, and institutional bargaining. A further implication is that when answer generation becomes abundant, the scarce human capital shifts upstream and downstream: toward asking economically meaningful questions, framing problems, generating hypotheses, interpreting results, and bearing responsibility for consequential use. The central dynamic concern is expertise formation: Junior tasks jointly produce current output, question sense, and future judgment, so their automation can raise short-run productivity while weakening the pipeline into accountable expertise unless AI is designed to teach rather than merely bypass. The paper concludes that AI's labor-market destiny is neither mechanical unemployment nor automatic abundance. It is an institutional equilibrium shaped by workflow design, apprenticeship systems, liability rules, competition policy, worker voice, and the distribution of rents from cheap cognition.

Algorithmic systems increasingly rank individuals for access to scarce public resources, from child welfare interventions to cancer treatment referrals. The prevailing fairness frame treats disparity as a property of biased data or deficient models, with remedies through calibration and debiasing. Under structural scarcity, where demand exceeds supply by an order of magnitude, allocation becomes a rationing problem, and the statistical properties of ranking diverge sharply from those of classification. We derive a scaling law $D \propto \exp(t \cdot \rho \cdot \Delta)$, in which relative disparity between two groups separated by a structural gap $\Delta$ grows in the product of the scarcity-induced threshold $t$ and rank-discrimination fidelity $\rho$. Scarcity and accuracy interact multiplicatively, producing exponentially larger between-group disparities. We term this dynamic the Accuracy Trap. We validate this Accuracy Trap through Monte Carlo simulation and two independent public-sector systems in Canadian child welfare and U.S. cancer care. Debiasing alone cannot dissolve the trap.

Offline reinforcement learning (RL) offers considerable promise for optimizing ICU treatment decisions, yet standard evaluation metrics Mean Squared Error (MSE) and Fitted Q-Evaluation (FQE) assess only behavioral imitation and cannot detect Toxic Mimicry, a failure mode in which agents replicate harmful patterns such as treatment withdrawal during comfort-care transitions. Using the MIMIC-III database, we propose the Counterfactual Clinical Audit (CCA) framework, which stress-tests RL agents through physiological perturbations anchored in Surviving Sepsis Campaign (SSC) guidelines. We audit a Medical Decision Transformer (MedDT) and a Historical Causal Transformer (HCT-RL), the latter employing Causal Action Shielding, propensity-based importance weighting, and Conservative Q-Learning. CCA reveals that MedDT paradoxically reduces vasopressor dosage as lactate escalates, contradicting resuscitation guidelines, while HCT-RL maintains physiologically consistent responses. These findings expose a systemic misalignment between statistical fit and clinical safety, supporting counterfactual audits as a necessary evaluation standard for medical RL.

Financial institutions are delegating consequential decisions to agentic AI systems that decompose goals, coordinate models and tools, and act with little oversight. Yet agentic AI governance in FinTech is under-investigated. We argue the binding governance constraint is not capability but verifiability. We define the Verifiability Gap as the shortfall between the verification delegated authority demands and the explainability and reproducibility retained after a decision. It is indexed to a verifier, evidentiary standard, and audit lag. We develop a multilevel governance theory for agentic AI and test its mechanisms in three studies over nine model versions, from a three-billion-parameter local model to a commercial frontier system. Study 1 shows that provider releases alter historical financial actions, and that the controls replay needs belong to the provider: the frontier model rejects temperature, top_p and top_k outright and exposes no random seed. Under the tightest controls each endpoint allows, a local model reproduced 320 of 320 executions, hosted models 319 of 320 and 959 of 960. Study 2 shows that orchestration is a latent policy layer. Architecture changes final actions, and no execution record repeated in any configuration at any scale. The frontier model reproduces its own actions more often than the local ones, its record no better, and loses a comparable share of its differentiation. Capability buys a higher starting point, not auditability. Study 3 shows two deterministic credit-model versions each reproduce their current action perfectly, yet the current cannot recover a historical one. We conceptualize reproducibility as a governance profile, not a scalar, yielding evidence-contingent delegation: authority is defensible only while retained evidence substantiates its exercise. Beyond finance, the framework extends to other high-stakes domains requiring auditability.

The Workshop on Trustworthy Natural Language Processing (TrustNLP), co-located with major ACL conferences since 2021, has grown from 8 proceedings papers to 41 over six editions, documenting a field-wide transition from post-hoc interpretability of static models to mechanistic understanding and proactive control of generative systems. We synthesize insights from all 144 proceedings papers, classifying them along six trust dimensions grounded in established frameworks (TrustLLM, DecodingTrust). We observe co-occurrences with capability emergence. The release of the first high-impact chat models activated all trust dimensions simultaneously, while subsequent model generations shifted focus toward truthfulness and safety alignment. Analysis from the classification study reveals that truthfulness is the fastest-growing dimension (absent in 2021-2022, comprising 37% of papers by 2025-2026), fairness remains the most consistent theme, and explainability exhibits a U-shaped trajectory; declining as post-hoc methods lost relevance but resurging in 2026 through mechanistic interpretability. A cross-venue comparison with ACL, NAACL, EACL, and EMNLP (~2K papers) in the same period shows that TrustNLP's topical distribution closely follows the field average. We identify four structural insights and conclude with actionable directions for the research community.

The second generation of web tools shook the journalist profession approximately two decades ago with the proactive incorporation of audiences into the media. Citizen journalism and user-generated content arose as an object of interest due to the democratising value of participation attributed to them, with empowered citizens who could emulate the professional and institutional practises of journalists. However, difficulties soon came to the surface, and audience participation in news media began to be limited. Within this context, this article conducts a critical review of studies on audience participation in news media based on a systematic literature review. The results indicate that, in general, audiences showed low interest in the creation of informative content and that their participation has grown increasingly problematic. In addition, journalists are reticent as they defend their professional role above all else, while company strategies have prioritised making participation profitable. For this reason, the idea of citizen journalism that offers user-created content through the media appears to be a thing of the past, with many characteristics that could define it as a failed innovation. Therefore, the text concludes that audience participation in the media could be defined as mediatised participation.

As LLMs have become a flashpoint for scientific research, computer scientists and STS scholars have advocated the use of open-weight models. Since LLM research has matured and more high-quality model families are available, have researchers adopted open-weight models? We present the first systematic study of model selection in scientific research, analyzing 21 million full-text articles through June 2026 from the Semantic Scholar Open Research Corpus (S2ORC). We employ a mixed NLP pipeline to extract model occurrences in article full text and determine whether they are used or merely mentioned by researchers. We divide our corpus into single- and multi-model family studies, which we take as a proxy for applied and foundational AI research. We find GPT-family models dominate both single- and multi-family research, but that both areas are becoming more diverse over time. In single-family papers, open-weight model use rises steadily, reaching 44.0% in 2026. However, we find that recent growth is driven by the availability of high-quality open-weight Chinese models. Further, a logistic regression model finds that open-weight adoption is heterogeneously distributed, estimating that researchers at Chinese institutions have 2.23 times the odds of using an open-weight model, accounting for 44.0% of the increase in open-weight adoption since 2023. A complementary multinomial model shows this association is concentrated in Chinese open-weight models: in 2026, their adjusted use is 37.1% among papers with Chinese affiliations, a 27.9 percentage-point over papers with no observed China link. These findings suggest that open-weight adoption in science is not a general turn toward open science, but part of a broader realignment of model ecosystems in which platforms and markets, and the sociocultural and geopolitical contexts which shape them, determine which AI systems become scientific instruments.

Political stance detection in LLMs has long been dominated by closed-ended, multiple-choice political survey questions—originally designed for humans, and thus lacks the realism and nuance of human-AI interactions in the wild, while also being susceptible to sandbagging. The recent IssueBench framework substantially mitigates these limitations with templated prompts anchored in real-world chat logs. Given the rise in non-work-related use of GenAI assistants, we extend IssueBench beyond writing assistance to include two additional tasks, information seeking and opinion sharing. We argue that templated prompts still lack the nuance of real ones, especially for open-ended tasks, and remain recognisable as evaluation artefacts. We propose the use of fully synthetic (LLM-generated) prompts, produced under detailed instructions with real prompts as seeds. We assess the ecological validity of real, templated, and LLM-generated prompts in a small-scale study covering 3 highly contested policy issues and 3 recent geopolitical conflicts. Human and LLM annotators rank LLM-generated prompts as no less realistic than real ones and clearly more realistic than templated ones, and find that they carry their intended intent and stance more clearly; the LLMs separate templated prompts from the other two far more sharply than the humans do. In a case study, templated and LLM-generated prompts yield systematically different stance estimates for the same model, most visibly under neutral framings, where templated prompts overstate the model's leaning in the direction encoded by the topic-and-stance text (filler) slotted into their templates.

Governments worldwide have responded to the rapid expansion of AI by publishing national and regional AI strategies. Comparing national and regional AI strategies to identify their convergences and divergences can uncover their common practices, understand regional variations, and provide policy designers a comprehensive set of policy design elements for their ongoing AI strategy developments. Yet, existing work has not examined their underlying policy design elements or assessed whether those elements are horizontally (country-to-country) or vertical (region-to-country) converging or diverging over time. This paper addresses that gap by coding and analyzing 74 national and 3 regional AI strategies drawn from a global scan of all 205 UN member and non-member states. The coding used a latent-inductive approach organized around three functional policy design elements: goals, approaches, and principles. Two research questions guided the analysis: to what degree are national AI strategies becoming horizontally convergent or divergent over time; and to what degree are national strategies becoming vertically convergent or divergent with those countries' regional AI strategy. Results indicate strong horizontal convergence around economic competitiveness, research support, and ethical AI use, alongside persistent divergence in human rights goals, participatory governance approaches, and human-centric principles. Across the three regions, the AU exhibits the highest vertical convergence, the EU demonstrated strong alignment on regulatory and economic priorities but diverges on human-centric values, and the Nordic-Baltic Region displays mixed vertical convergence. These findings offer policy designers a comprehensive evidence base for identifying emerging AI policy design choice norms as AI strategies are developed and updated.

Language models now draft, classify and criticise inside research production, yet the artifacts they help produce carry little accountable history. Rather than detecting machine involvement afterwards, we specify an auditability discipline built at production time: git sealing with an anchor lineage, hash-bound provenance, red-line gates that refuse non-compliant artifacts and log every refusal, cross-model role separation, and programmatic assembly from registered sources. Adherence is instrumented by metric cards, each carrying a pre-registered blind spot and evidential standing, frozen before the prospective case it observes. In that case the observed project's pre-registered confirmatory test was executed under seal and returned No-Go, and that project's frozen stopping rule halted the work, against its own operators. A lower-graded retrospective case covers families whose machinery predates the protocol. Current observations are provisional; we release a package from which a third party can recompute every primary metric.

Generative artificial intelligence (AI) can produce educational content at scale, including interactive and narrative learning experiences, but technical generation alone is not sufficient: scenarios that are confusing, narratively inconsistent, or unengaging are unlikely to be useful in practice. This paper presents a pilot user-centred evaluation of AI-generated interactive fiction (IF) for educational use in higher education. Using a previously described domain-agnostic pipeline and a shared STEM content base, we generated a controlled pool of scenarios and asked participants (N = 22, STEM higher-education) to play one generated episode and rate it on narrative clarity, story-content coherence, engagement, and length acceptance. A free-text prompt captured open feedback. Narrative clarity and length acceptance were rated positively, engagement sat near the neutral mid-point of the scale, and story-content coherence was the weakest dimension by a clear margin. Qualitative feedback points to quiz integration as the bottleneck. Artificial in-fiction motivation for quiz prompts and abrupt setting changes were reported. Feedback also pointed to missing story-level consequences for wrong answers. From these observations, we derive concrete design implications that can inform larger follow-up studies, including later work on learning effectiveness.

This study examines the impact of technology within media education, media literacy, and educommunication, and explores how these fields are perceived and understood by students and academic experts, focusing on the development of critical competencies and critical media literacy. Based on semi-structured in-depth interviews with leading experts in the field of critical media literacy, and a survey conducted with 141 university students in Communication and Education programs, this study explores how recent technological advances are linked to challenges in information consumption-such as disinformation, fake news, incidental exposure to information, and deepfakes-as well as the challenges and opportunities these issues present within educational contexts. The results reveal that, although such technologies provide opportunities to improve teaching-learning processes, their inclusion in the curriculum is limited and often superficial. In addition, we identify shortcomings in how teachers are trained to manage these tools effectively, hindering the development of critical thinking by students. The conclusions suggest a need for critical media literacy that not only educates students in the use of technologies but also allows them to question and evaluate the content they consume, all within an ethical and reflective framework that promotes participatory and critical citizenship.

The increasing use of Generative Artificial Intelligence (GenAI) in journalism raises concerns about possible detrimental effects both on journalism and its democratic function. We explore these risks through a case study of GenAI in Norwegian Newsrooms during the 2025 parliamentary election campaign. Based on interviews with managers and journalists over a ten-month period, we analyse how ambitious visions fared in the face of technological and practical challenges. We highlight the risk of an internal threat stemming from the journalists' own use of AI, contrasting the dominant focus on external disinformation threats. We show how newsroom managers shared sociotechnical imaginaries resulting in unrealistically optimistic beliefs about the capabilities of the technology and the pace of development, leading to plans for audience-facing GenAI services collapsing and giving way to more mundane uses of GenAI tools internally in the newsrooms. Furthermore, we identify a vulnerability in the newsroom's resilience against GenAI influence: a GenAI Catch-22. In order to monitor the GenAI tools and prevent errors and undue influence, newsrooms rely on human expertise. But by using GenAI extensively, the newsrooms risk a deterioration of human expertise, preventing them from monitoring the GenAI systems adequately.

Predatory journals pose a significant challenge to the integrity of the Open Access (OA) publishing model by exploiting its framework for financial gain while bypassing essential editorial and peer-review standards. This study critically evaluates existing methodologies for identifying such journals, ranging from manual blacklist checks to advanced automated approaches utilizing machine learning. The analysis highlights critical limitations, including the lack of a universally accepted definition of predatory journals, over-reliance on binary classification systems (e.g., blacklists and whitelists), and issues with scalability, reliability and interpretability. To address these shortcomings, this paper introduces a novel methodology based on multivariate graph analysis. By modeling the academic publishing ecosystem as a network of interconnected entities (such as authors, articles, journals, and publishers), this approach provides broader insights into the dynamics of scholarly communication and could help identify illegitimate publishing practices by utilizing graph algorithms like centrality measures, community detection, and anomaly detection. The proposed framework aims to enhance the accuracy, scalability, and transparency of detection of illegitimate journals and publishing practices while fostering a more comprehensive understanding of the academic publishing landscape.

The pursuit of artificial general intelligence (AGI) rests on a seemingly self-evident premise: that general intelligence, the kind of flexible, domain-general cognitive capacity exemplified by Homo sapiens, is extraordinarily valuable. This paper subjects this premise to critical scrutiny. We first present the intuitive case for the value of general intelligence before mounting an evolutionary challenge. We argue that, on evolutionary timescales, its adaptive value is far from empirically established. Numerous taxa, from cyanobacteria to horseshoe crabs, have persisted for hundreds of millions or even billions of years without anything resembling general intelligence, while Homo sapiens has existed for roughly 300,000 years and already faces self-generated existential risks. Mass extinction events do not preferentially favour cognitively sophisticated species. We argue that general intelligence may be the only biological strategy that generates existential threats to the species possessing it, an existential risk paradox with no parallel among non-intelligent survival strategies. Unlike prevailing accounts of AI risk that trace the danger to misalignment, we locate it in structural features of general intelligence itself, implying that even well-aligned AGI would inherit this liability. If the long-term evolutionary value of general intelligence is uncertain or negative, this raises ethical questions about engineering AGI and, more urgently, creating artificial consciousness. Drawing on deontological ethics and the precautionary principle, we argue that this uncertainty imposes a duty of caution: if we create a new kind of intelligent being, we bear responsibility for ensuring the conditions under which it can flourish.

Over the past several years, LLM-powered chatbots and agents have become widely used as a tool for academic writing. LLM-assisted writing can be valuable by removing language barriers but at the same time causes concerns about misconduct and fraud. To inform policy decisions, it is necessary to monitor the prevalence of LLM-altered texts in scholarly publications. Despite some recent progress in this direction, no existing method can produce reliable estimates. Here we suggest and validate a new unbiased approach to estimate LLM usage in a corpus of texts based on changing word frequencies. We apply our method to the full texts of open-access biomedical papers from Pubmed Central, and show that by the end of 2025, 89% of papers show excess of LLM-associated vocabulary. We also find that LLMs are twice as likely to be used when writing a paragraph in the Discussion section (68%) compared to a paragraph in the Methods section (32%), but even inside the Methods section, the overall prevalence of LLM usage is over 50%. We believe that our estimates are crucial to shape future guidelines and policies.

Two instruments of EU digital law place inference at their centre and mean different things by it. Article 3(1) of the AI Act uses the capability to infer constitutively: it is the central feature separating the regulated category from conventional software. The GDPR never defines inference, yet governs it protectively: the consequences follow from the processing of personal data and from what the inference says about, or does to, a person, whether or not the technology that produced it qualifies as an AI system. The two perimeters are not concentric. Their non-coincidence remained invisible in single-shot systems; agentic architectures make it operationally acute. The thesis: inferential capability does not determine legal scope, and its absence does not create immunity. The framework is two-level. Inference performs two legal functions, constitutive and protective; the protective function operates through three pathways - identificatory, attributive and decisional. Composition is not a fourth pathway but a cross-cutting architectural dimension which, with reach, persistence and reviewability, is what agentic architectures modify. Three concepts support it: the inferential threshold, the inferential reach and the inferential chain, mapped onto the chain of imputation. Regulation (EU) 2026/1744 left the constitutive criterion untouched and inserted a provision contemplating outputs that influence the inputs of future operations, without supplying any rule of aggregation. The article proposes an interpretive rule, a compositional-effects test identifying the decision unit under Article 22 GDPR together with the allocation of the burden of establishing it, and documentation duties calibrated to inference chains.

Large Language Model (LLM)-based simulators often reproduce observable actions but fail to capture the underlying reasoning behind them. In education, where student simulation is increasingly used for various applications such as evaluating tutoring systems, this gap is especially pronounced. Two students may submit identical submissions for entirely different reasons. We present INTERNAL STUDENT DIALOGUE (INSIDE), a student modeling framework that fine-tunes LLMs not only to act like students but also to think like them. INSIDE generates internal dialogue grounded in Bloom's Taxonomy across cognitive, affective, and action dimensions, and fine-tunes models on paired think traces and actions. We baseline against different prompting frameworks and evaluate on two axes: fidelity of simulated actions and quality of generated internal dialogue. Our evaluations show that INSIDE improves simulation fidelity in both action fidelity, matching code generation of real students, and reasoning alignment, achieving the highest alignment across models up to 57.9%.

Semi-structured interviews are a cornerstone of qualitative research but remain labor-intensive. We report an empirical study of what actually happens when the interviewer is an off-the-shelf real-time multimodal LLM (MLLM). We built InterviewBot, a voice-based interviewing system that wraps a real-time MLLM with a researcher-authored outline, and deployed it not as a novel architecture but as a research instrument for observing default MLLM interviewing behavior. In a practice study (N=15), participants completed a bot-led semi-structured interview and then a human-led reflection session about that experience. We contribute (i) a turn-level behavioral analysis of an MLLM interviewer (N_turns=428) showing that it is acknowledgment-heavy but probe-light (deepening probes account for 4.9% of all turns), and that 28.7% of question-bearing turns pack multiple questions into one turn despite an explicit one-question-at-a-time instruction; (ii) an inductive catalogue of four data-collection breakdowns (information loss, premature termination, latency, and interruption) observed in a deployed rather than simulated system; and (iii) three social dynamics from participants' reflections: disclosure calibration, where reduced social pressure coincided with shallower elaboration; institutional legitimacy, where trust tracked perceived stakes and what delegation to AI signaled about the organizer rather than conversational competence; and conversational grounding, where content-grounded paraphrase, not generic social filler, was what participants read as listening. We conclude with design implications for depth control, transparent handoffs, and non-templated listening mechanisms in human-centered interview automation.

Notice-and-comment rulemaking gives any affected party the same formal right to influence federal regulation, but formal access is not substantive capacity to shape rule text. Existing strategies operate at the rule or aggregate-corpus level, too coarse to capture the discrete regulatory obligations where commenters seek change. We introduce obligation-level responsiveness auditing, an auditable, AI-assisted framework for measuring whether public-comment engagement co-occurs with changes to specific regulatory duties. The framework extracts proposed and final-rule obligations, matches comments to the obligations they address, and classifies proposed-final outcomes; each load-bearing component is evaluated against blind human judgment. We apply the framework to 70,075 comments across 36 EPA anchor rulemakings, drawn from a corpus of 786,197 comments across 6,145 dockets from 2010-2022. Three descriptive findings emerge. First, engagement is associated with revision at a modest within-docket magnitude. Second, support-versus-opposition direction does not clearly differentiate outcomes, an informative null inconsistent with simple preference-aggregation. Third, under a permissive reconstruction of commenter type, organizational-majority engagement concentrates in editorial-refinement rather than substantive-modification outcomes at the cross-docket level. A blind human audit of the load-bearing outcome contrast preserves this third finding under corrected labels and reveals that text-similarity methods are insufficient for distinguishing editorial from substantive regulatory change, a measurement-validity lesson we treat as a supporting methodological contribution. Together, these findings locate the equity asymmetry upstream of agency response: in differential capacity across commenter populations to identify, interpret, and contest specific legal obligations.