Papers

We study private bot-checking behavior from the demand side: when people suspect an account is automated, whom they suspect, and what follows. Using Botometer's server-side query logs, the most widely used bot-detection service, we treat each query as a behavioural trace of suspicion. We analyze over 1 million public checks of Twitter accounts from 2020 to 2023, enriched with 3.2 billion tweets from the contemporaneous 1% public stream. Collective suspicion spikes with platform crises, most sharply around the 2022 Musk-Twitter bot dispute. Checked accounts are older and more prolific, have more followers, and post promotional, political, and crypto content. Accounts that draw collective suspicion have higher bot scores and are more likely to be suspended. Bot-related public attention and Botometer activity are elevated during the same broad periods, although their short-run fluctuations are largely uncoupled. Public feedback focuses on first-person identity claims for humans, and evidence-based arguments citing posting rate, political content, and cross-account coordination for bots and cyborgs. Bot suspicion thus constitutes a mass, distributed form of platform auditing that tracks meaningful signals of automation, establishing audit-tool query logs as a novel lens on public responses to platform manipulation.

Full-consensus rates are often treated as indicators of collective cognition, yet depend on how participation and final states are operationalized. We replayed 100 held-out human Wason groups with matched large language model (LLM) agent groups, seeding one belief-anchored agent per participant's pre-discussion answer and scoring agents and people with the same code. Across human scoring definitions, estimates ranged from 24.0% to 57.0%; about one fifth of participants never posted, whereas agents almost always did. Agent groups remained more consensual in two post-unblinding sensitivity analyses: the submit-based comparison (n = 98) yielded gaps of 34.0 and 43.9 percentage points for chat and reasoning modes, and the participation-matched comparison (n = 45) yielded gaps of 34.1 and 44.4 points. These complementary routes reduced different measurement asymmetries yet converged within 0.5 percentage points. The gap persisted without early stopping and under a reparameterization removing the memorizable answer; reasoning-mode groups then agreed nearly unanimously, mostly on incorrect answers. Simulated consensus did not track collective accuracy, and belief-anchored agent groups were biased estimators of the human group-outcome distribution in this setting. These analyses provide a scoring-explicit basis for assessing simulated-group estimates of human deliberative outcomes.

We study income taxation when an AI agent implements economically relevant choices through a rule hidden from the government. Alongside unobserved productive ability, this hidden preference-to-execution mapping creates double unobservability: the same observable tax-base response can carry different welfare consequences. We call the resulting income welfare-opaque. Our constructions show that tax-base statistics can coincide while reform welfare effects differ, even when mechanical welfare weights are identical. We derive an optimal-tax condition that adds a response-weighted execution wedge to the familiar sufficient statistics. A higher marginal rate gains a corrective benefit under local over-execution and an additional cost under local under-execution. Observing the wedge identifies the welfare effect of a marginal reform at the prevailing schedule; bounds on it deliver bounds on that effect. A controlled laboratory compares 4,500 model runs across five AI engines. Faithful delegation selects the score maximizer in essentially all runs. Conflicted objectives produce heterogeneous responses: Claude largely preserves the score maximizer, GLM moves predominantly downward, and GPT-mini and Qwen show concentrated lower-tail increases. Qwen also makes substantial downward adjustments. Different engines locate their departures at different points and in different directions of the designed distribution. Explicit scores align model rankings; formula-based objective instructions yield more uneven agreement. Qwen shows a clear positive tax-by-objective interaction, but its direction does not generalize across engines and the pooled sign depends on its inclusion. The analysis identifies execution information as a complement to conventional tax-base statistics.

Deceptive online job advertisements have emerged as a primary pathway into forced labour, yet systematic detection methods remain underdeveloped due to data scarcity and absence of empirically validated indicators. We formalise this detection challenge as a classification problem under signalling theory, where exploiters transmit costless signals mimicking legitimate communications across textual, visual, and structural dimensions. Using 464 verified cases (164 deceptive, 300 legitimate) collected through anti-slavery charities across nine origin countries and 21 industries, we develop multimodal detection models combining computer vision, natural language processing, and semantic embeddings. Through systematic feature ablation experiments and repeated stratified cross-validation, we demonstrate that individual modalities achieve substantial discriminatory power (ROC-AUC: 0.87–0.97), whilst their integration yields modest further gains. SHAP-based analysis reveals that text quality and domain-specific risk language are the primary discriminators, with readability indices, risk keyword density, and visa sponsorship mentions ranking highest, followed by visual colour and texture features. These production quality gaps reflect resource constraints that prevent exploiters from maintaining professional standards across all communication channels simultaneously. We operationalise findings through a proof-of-concept decision support system providing interpretable risk scores for practitioners. This work demonstrates how rigorous analytical frameworks can address complex humanitarian operations challenges characterised by information asymmetry and limited ground-truth data.

AI systems can strengthen democracy by supporting deliberation at scale by addressing cognitive, social, platform-design, and market-driven frictions, while preserving human agency. Unlike proposals such as liquid democracy that restructure representation through vote delegation, in this position paper, we argue that AI-assisted deliberation offers a more promising path by lowering barriers to meaningful engagement without substituting machine judgment for human choice. Drawing on evidence from online deliberation platforms and experimental research, we identify four guiding principles: preserving agency and autonomy, encouraging mutual respect, promoting equality and inclusiveness, and augmenting rather than substituting active citizenship. We also address critical challenges, including alignment, sycophancy, training bias, and over-reliance on AI systems. We call on the machine learning community to develop deliberation-focused AI systems evaluated not on engagement metrics but on their capacity to facilitate informed, representative, and friction-robust discourse.

The EU AI Act (Regulation 2024/1689) imposes technical obligations on high-risk AI providers, yet Articles 8-15 were drafted for predictive AI and leave seven technical gaps when applied to generative systems, spanning non-deterministic data governance, training-data provenance, continuous conformity, human oversight, open-ended robustness, emergent risk, and generative fairness. We deliver Governance-as-Code (GaC), a framework of 43 machine-checkable acceptance criteria across six compliance modules that run in a CI/CD pipeline and emit Article-indexed audit evidence, and we show the actual Rego policy code rather than merely describing it. Our central commitment is that the Act's open-textured standards ("appropriate levels," "possible biases") become declared, auditable numbers: robustness thresholds are derived from the provider's documented baseline and a state-of-the-art floor, and framing bias is collapsed into eight measurable proxies tested by counterfactual demographic probing. We also correct who owes what, since under Article 25 and Chapter V a downstream deployer relies on the upstream provider's Article 53 training-data summary and documents only the layers it controls, so GaC verifies that summary rather than demanding per-sample documentation the deployer never had. We validate on two enterprise deployments, a high-risk advisory chatbot and a limited-risk content generator, benchmarking against a manual expert audit rather than documentation artifacts that were never designed to enforce compliance. GaC reproduces all of the manual audit's findings, including three penalty-triggering violations, while cutting audit labor by roughly 75%.

People increasingly turn to AI chatbots for news and explanations of world events. But do they receive the same political answers when they ask in different languages? Here we show that the language of a question can change how the same AI systems assess the war in Ukraine. We ask GPT, Claude and Gemini to evaluate twenty statements about the war in 112 languages, collecting 67,200 responses. The balance between Russia-leaning and Ukraine-leaning responses differs across languages. When we group responses by countries' official languages, they follow a pattern resembling worldwide political divisions: relatively more Russia-leaning answers correspond to more favourable public views of Russia, less support for Ukraine in United Nations votes, and less aid to Ukraine. The broad pattern recurs across all three models and remains when individual statement pairs are removed. Our findings suggest a possible route through which information warfare may shape the text used to train AI models, which may in turn spread geopolitical biases.

Large Language Models (LLMs) are increasingly applied to legal and criminal justice tasks, yet existing work focuses almost exclusively on post-arrest scenarios where the suspect's identity is already known, leaving the critical pre-arrest challenge of inferring suspect characteristics from incomplete evidence largely unexplored. To fill this gap, we introduce the Profiling, Investigation, and Judgment (PIJ), comprising 2,500 real homicide cases from five countries. PIJ evaluates LLMs across three tasks that span the entire criminal investigation pipeline: criminal profiling, which requires abductive reasoning to infer suspect attributes from fragmentary scene evidence, crime process reconstruction, which tests structured information extraction, and sentence prediction, which demands legal deductive reasoning. We evaluate 9 powerful LLMs and find that performance degrades systematically as tasks shift from explicit fact extraction to implicit reasoning over unknown suspect profiles. Categories requiring inferential reasoning, such as motivation and victim-offender relationships, remain the primary bottlenecks. Further analysis reveals substantial gaps between LLMs and human experts, along with pervasive biases in gender, age, and motive attribution. Our findings indicate that pre-arrest inference from incomplete evidence remains an open challenge.

Digital Educational Systems (DES) enable millions of individuals to learn or to acquire new skills. Their promises vary from one platform to another: some grant genuine degrees, others pledge rapid progress, drawing for this purpose on a range of techniques, from the recommendation of learning resources through complex models to a plain course textbook uploaded onto a learning management system. In this article, we argue that categorizing DES along two dimensions allows to map differences in the way these DES structurally conceptualise the learning process. These underlying dimensions are at once dimensions of the student model and mechanisms built in the DES's designs; we identify them as agency and legitimacy. Indeed, DES are, on the one hand, anchored in an institutionalised regime that grants them legitimacy and, in turn, legitimises the learning undertaken on these platforms; on the other hand, they make design choices as to the share of decision left to the learner in the learning process. We map a curated sample of common DES onto a plane composed of these two dimensions translated into indicators. This allows us to highlight underpopulated regions of the plane and to envisage how a DES might regulate these two dimensions algorithmically, in the interest of learning.

High annotation reproducibility does not necessarily imply that an LLM-inferred measure captures the construct it is intended to measure. We test this distinction using a dataset from the European Commission's AI Act consultation, linking structured survey responses to free-text consultation submissions from the same stakeholders. LLM annotations of consultation submissions are highly reproducible (intraclass correlations > 0.99), yet show limited convergence with survey-reported measures of the nominal construct they were intended to approximate. Divergence between survey-and LLM-inferred text-based measures varies systematically across stakeholder groups: business associations express greater concern about AI risks in text-based consultations than in survey responses (ḡ = +1.0), whereas public authorities and several nonbusiness groups show smaller or negative divergences. Divergences between scores suggest positive spatial autocorrelation across European countries (Moran's I = 0.347, p = 0.036), indicating that stakeholders from neighboring countries tend toward more similar text-based stances towards AI safety concerns. Despite divergence, survey-reported concerns remain strongly associated with support for explainability across all divergence levels. These results demonstrate that LLM annotation reproducibility can coexist with poor construct correspondence and motivate validation procedures that distinguish reproducibility, construct validity, and communication context variation when LLMs are used as measurement instruments.

Conversational agents are increasingly used to guide reflection. A recent randomized trial compared a GPT-4o career reflection agent with the same program in a static journaling survey. Agent participants ended less committed to their career plans and more doubtful. We coded all 17,930 turns from its two studies, checked our coding against human coders and linked conversations to the trial's surveys. The rules the agent followed were the easy-to-check ones, like a reply length cap. Told not to flatter, it praised participants in half of its turns; told to challenge gently, it almost never did, and such a break leaves no visible trace. The behavior tied to the worse outcome was the demand to decide: the survey posed each decision once, while the agent asked again when participants hesitated, and those pressed most ended most doubtful. Our findings inform reflection agent design and the writing of checkable instructions.

People present different aspects of themselves across relationships. Computational work has captured such variation in communication style. But this variation also extends to which principles people foreground or background in a particular relationship–i.e., in the values they express and how they balance them. We conceptualize these relationship-specific expressions of values as demonstrated values. To make demonstrated values visible, we introduce Value Faces, a system that analyzes a person's existing chat histories from their everyday messaging platforms using Schwartz's ten basic human values and produces separate value profiles for their different relationships. In a mixed-methods study(N=18), we find that the resulting value profiles distinguished participants' relational contexts with twice the odds of guessing, while system-inferred differences across relationships aligned with participants' perceptions of those differences. Participants used these profiles to articulate previously implicit differences in how they presented themselves, connect them to roles and changes over time, and reconsider their self-assessments.

A golden ticket allows one reviewer to advance a proposal that the rest of a panel would reject. We assess its score-based rationale by comparing minority-support ranking with other rules for allocating the same number of places. We do so on 10,625 rejected submissions to a computer-science conference between 2017 and 2024, using later citation percentiles as the outcome. Minority-support ranking measured the gap between the two highest scores after adjustment for the panel mean and reviewer count. At a budget of 10% of the year's accepted count, panel-mean ranking selected submissions with a mean citation percentile 7.75 points higher than minority-support ranking (95% bootstrap interval 4.35 to 11.09). The difference was larger in the other sample periods and citation windows examined. At that budget, minority-support ranking was equivalent to a lottery within a margin of plus or minus 5 percentile points. Two further analyses, adjusted highest-score ranking and a matched comparison, found no clear citation advantage from minority support once the panel mean was accounted for. Ranking on score variance also outperformed minority-support ranking in a paired comparison. Estimates cover the 69.8% of eligible rejections whose later citations could be verified. These comparisons identify panel-mean ranking as a benchmark for evaluating whether golden tickets select work that would otherwise be overlooked.

Large language models (LLMs) are increasingly used to simulate human survey responses and behavioral reactions, yet unreliable simulations can mislead social science conclusions. However, existing evaluations focus on end-to-end scores, leaving it unclear how different aspects of the simulation process interact to determine reliability. We propose ReliMap, which decomposes LLM-based human behavior simulation into three structured layers and evaluates reliability at both the individual level (R1) and population level (R2) across three configuration dimensions: model capacity, profile completeness, and population coverage. Through experiments across four simulation tasks and eleven LLMs, we find that all models exhibit substantial distributional bias without profile conditioning. Profile conditioning reduces this bias with diminishing returns. Larger models benefit more, and attribute informativeness matters more than quantity. Critically, R1 gains do not reliably transfer to R2–individual and population-level reliability can move in opposite directions. At the population layer, increasing coverage reduces variance but not systematic bias, with R2 stabilizing at around 50-100 individuals. These findings highlight that reliable simulation cannot be achieved by optimizing any single layer in isolation, but requires coordinated improvement across all three.

Instructional materials can be passed along to new instructors, yet the instructional reasoning, such as instructional intentions, contextual information, and reflective insights behind their design, is rarely preserved. Traditional Learning Management System (LMS) tools are designed to store instructional artifacts rather than reasoning. To address this issue, we propose Teaching Memory, a design approach that treats instructional reasoning as a primary form of knowledge. We present a design-oriented framework for preserving instructional reasoning in a lightweight manner within instructors' existing workflows, enabling continuity and reuse.

Burnout is a chronic occupational syndrome, and open source is close to a worst case for it: maintainers absorb unbounded demand with no manager to reallocate work and no organization to notice decline. The cost is not only personal. Burnout precedes withdrawal, and in projects sustained by a handful of maintainers, one departure can break infrastructure that thousands of downstream systems depend on. Yet the field has no way to see it coming: self-report inventories, the only existing measure, miss exactly the contributors most in need of detection and cannot be applied retroactively, so the field cannot even ask how common burnout is or what helps. We present BurnRiSc, a framework that operationalizes the Oldenburg Burnout Inventory's two dimensions, exhaustion and disengagement, as 14 behavioral and linguistic signals computed from GitHub activity and scored against each contributor's own history. The signals aggregate into two weighted dimension scores, with weights learned from labeled cases, and average into a monthly Burnout Risk Score (BRS). In a preliminary evaluation across 68 contributors in ten repositories (ten disclosed burnout cases, twelve comparable-volume collapses, and 46 comparison contributors), sustained BRS elevation precedes 6 of 10 disclosures by 6-15 months, 8 of 10 when adding peak BRS as a second criterion, and 10 of 10 over any prior time frame. We thus present BurnRiSc as evidence that burnout is screenable from public data.

Digital competence frameworks define the knowledge, skills, attitudes and values required for participation in digital society. The European Digital Competence Framework for Citizens (DigComp) and UNESCO digital literacy and artificial intelligence competency frameworks provide reference structures for policy, curriculum and competency development. The growth of generative and agentic AI raises a further question: does knowing what constitutes competent digital behaviour provide sufficient evidence that an individual can act competently in an authentic digital situation? This conceptual paper addresses that question through a comparative analysis of DigComp, UNESCO Digital Literacy Global Framework, UNESCO AI Competency Framework for Students, and the International Digital Driving License (IDDL). Rather than treating these frameworks as competing models, the paper distinguishes their functions. DigComp and UNESCO primarily define what individuals should know, understand, value and be able to do. IDDL addresses the subsequent evidentiary problem by assessing digital capability through behaviour observed during authentic digital tasks. Its Knowledge-Capability-Reflection model distinguishes knowing what should be done, demonstrating the required behaviour, and explaining the judgement behind it. IDDL integrates Digital Skills, Cybersecurity Awareness and AI Competency, and extends assessment to human agency under AI delegation. The paper distinguishes defined competence from demonstrated capability and positions IDDL as an interoperable behavioural assessment and certification layer that can operate downstream of international competency frameworks.

This paper presents the AR fairness metamodel, a framework designed to represent, analyze, and compare different fairness scenarios. The metamodel considers key elements, such as agents, resources, and their attributes, and enables the systematic definition and comparison of various fairness measures. We provide examples involving both discrete and continuous measures, including equality, equity, group fairness, individual fairness, the Gini index, the Theil index, Jain's fairness index, and a detailed fairness measure for Australia's Child Care Subsidy. We also explore relationships among group fairness, individual fairness, and envy-freeness, supported by formal proofs. At the conceptual modeling level, our approach builds on the Tiles framework, which offers modular components that can be connected to capture diverse fairness definitions. The goal is to make AR-based fairness definitions practical and adaptable across contexts, providing a clear way to define, compare, and evaluate them. An implementation of the Tiles framework is available as an open-source tool, and can support fairness modeling and evaluation across a wide range of applications.

This paper explores how the technologies U.S. election officials depend on also challenge their work. Every election, misleading narratives stem from errors that occur while using election technology, threatening election official safety and eroding faith in elections. We seek to understand the impact of these challenges directly from worker perspectives. This paper presents a qualitative analysis using combined interview and survey data from a total of 50 election officials representing 20 states. Our findings present design as a substantial factor in mistakes when using election technology, but other obstacles, like lack of funding and technical support, also strain operations. We show how election technology design can lead to harms for election officials and connect those harms to their negative effects on democracy. Our work emphasizes the potential for researchers in human-centered computing to support U.S. election officials-and democracy-by working towards better usability across election technologies.

Interlocal agreements are one of the primary instruments through which local governments formalize collaboration for public service delivery, yet the institutional and financial content encoded in these contracts has remained inaccessible to systematic analysis at scale. This paper introduces an end-to-end computational pipeline for classifying intergovernmental agreements by institutional form and extracting financial relationships between principals and agents in service contracts. Applied to Iowa's 28E archive (N = 21,629), the largest dataset of interlocal agreements in the United States, the pipeline combines LLM-based summarization and classification across LLaMA 3.1, GPT 5.2 Pro, and Gemini 3 Pro on a four-class classification task that distinguishes agreements as either service contracts, resource sharing agreements, joint operations agreements, or new joint entity agreements. We also identify the financial principal and agent in these agreements and contracts, as well as the resulting dollar amounts and represent them on a directed network. The resulting financial network is organized around a small number of dominant service providers, with counties serving as the most structurally versatile actors, and cities as predominantly principals. By rendering the content of Iowa interlocal agreements analyzable at scale for the first time, this pipeline establishes a reusable methodology that researchers and state agencies can apply to track how public dollars move across local governments and to identify entities that depend heavily on a small number of providers.

Scientific code repositories encode decades of human knowledge in executable models, methods, and tools. Yet fragmented toolchains, implicit domain conventions, and specialized correctness criteria make this knowledge difficult to convert into reliable learning experience-a challenge we call the scientific experience bottleneck. We introduce ScienceIDE, infrastructure for turning the world's scientific code into programmable environments for scientific agents. Guided by expert-defined scientific cases and acceptance criteria, agents transform repositories into executable environments that support task generation, execution, and scientific verification. These environments provide a shared foundation for supervised fine-tuning, reinforcement learning, and evaluation. Using verified interaction trajectories, we train PhAI-IDE-72B, PhAI-IDE-9B, and PhAI-IDE-4B. The model family shows gains in held-out scientific-code repair and across selected general-purpose benchmarks in code, reasoning, and knowledge, providing evidence of positive transfer from scientific experience to broader capabilities. ScienceIDE lays the foundation for an integrated workspace for agent learning and scientific practice, making humanity's scientific software a shared substrate for developing scientific intelligence. Code: https://github.com/aitofound/ScienceIDE

Artificial intelligence (AI) is increasingly integrated into healthcare systems worldwide, yet its successful clinical adoption depends critically on workforce readiness, particularly in low- and middle-income countries (LMICs) where infrastructural and training gaps persist. This cross-sectional study evaluated awareness, attitudes, preparedness, and barriers to AI adoption among 761 healthcare professionals across multiple disciplines and practice settings in Nigeria. Data were collected between December 2025 and March 2026 using a structured, validated questionnaire. Overall awareness of AI in healthcare was high (92.6%); however, objective knowledge and self-reported preparedness remained limited, with 40.9% reporting low or very low knowledge and only 63.0% feeling adequately prepared. Willingness to adopt AI was high: 92.5% expressed interest in training, and 78.7% supported inclusion of AI education in undergraduate curricula. Key barriers included lack of training (84.7%), poor infrastructure (71.1%), high cost of AI tools (61.0%), fear of job displacement (60.6%), ethical concerns (52.9%), and data privacy concerns (52.7%). Significant differences in preparedness were observed across geopolitical zones (chi-square (5) = 24.28, p < 0.001), and awareness differed across professional groups (chi-square (6) = 68.38, p < 0.001). Attitudes toward AI differed significantly across professional groups (F = 3.32, p = 0.003), with professionals who felt prepared demonstrating more positive attitudes (mean = 3.74) compared to those who did not (mean = 3.46). These findings reveal a critical disconnect between high awareness and actual readiness, underscoring the need for targeted training, infrastructure investment, and clear implementation frameworks to bridge the gap between AI technological potential and clinical reality in resource-constrained settings.

Proposals for international agreements that limit frontier AI development depend on verification, and a central challenge is detecting undeclared compute facilities used to evade restrictions. Underwater data centers (UDCs) have been suggested as one such evasion vector, but their feasibility at frontier scale and their detectability have not been seriously assessed. We examine current UDC deployments, evaluate construction and maintenance complexity relative to land-based facilities, and analyse the feasibility of a 100,000 H100-equivalent training run underwater. We find that power delivery and cooling are tractable, but interconnect and the hands-on maintenance that large training runs require are severe obstacles - surmountable only by a well-resourced state actor accepting large cost and schedule penalties, and only where concealment, rather than efficiency, is the objective. We then assess detectability through thermal, acoustic, optical and synthetic-aperture-radar (SAR) surveillance. Thermal detection of an operational pod is unlikely outside shallow, calm water; acoustic detection is marginally more effective, but faces limitations in attribution; and optical/SAR monitoring is most powerful during construction and maintenance, when the pressure-vessel fabrication base and the cable-laying fleet create distinctive signatures for AIS-tracking. We conclude that UDCs are a comparatively unlikely evasion route relative to underground or industrially disguised land-based facilities, but the residual risk is non-zero and warrants operationalising the detection modalities discussed.

A sizable literature studies content moderation locally, at the level of individual moderation decisions, for example by measuring or predicting the effects of specific interventions. However, the problem of how such decisions should be combined into effective platform-level moderation strategies is comparatively unexplored. We address this latter problem by formulating content moderation as a global, adaptive, and sequential decision process in which heterogeneous interventions must jointly balance multiple competing objectives. Drawing on feedback control, we introduce a general control-theoretic framework for composing moderation actions according to their expected effects on an evolving platform. We instantiate the framework in large-scale, empirically grounded simulations and compare two control-theoretic moderators against several baselines and local strategies. When moderation aims to maintain competing platform-level properties around desired conditions, the control-theoretic approaches achieve the best overall performance. They also use severe interventions more selectively and recover more effectively after external surges of harmfulness. These results demonstrate the advantages of studying content moderation as a global, adaptive, and sequential decision problem.

To manage the complexity of smart cities, a variety of smart city platforms (SCPs), both proprietary and open-source, have been proposed. Typically acting as middleware between IoT devices at a lower layer and smart services at a higher layer, these platforms simplify the management of smart cities by reifying a variety of requirements common to different municipalities. Open-source software (OSS) is typically free of charge, making open-source SCPs an attractive option for municipalities with smaller budgets that still wish to improve efficiency and quality of life for residents and visitors. Furthermore, using an OSS SCP promotes digital sovereignty, since municipalities can control where the data will reside and will be less prone to vendor lock-in. A lack of support for OSS hinders its adoption by municipalities' IT teams, often with scarce technical resources. An active (aka "healthy") ecosystem, including well-orchestrated developers and users, the availability of installation and user manuals, non-vulnerable and documented code, test batteries, and other artifacts, is an important decision factor before adopting an OSS SCP. In this paper, we evaluate different open-source SCPs using the concept of "OSS health", covering several of the aforementioned characteristics.

Rapid algorithmic developments have accelerated the sampling of redistricting ensembles (balanced graph partitions), yet evaluating rare events and sampling complex target measures remains a core challenge due to the high-dimensional and combinatorial nature of the phase space. We address a prerequisite for stratified sampling on this space: constructing and diagnosing candidate strata with suitable coverage and overlap. We build a grammar on observed plans by clustering districts into representative “letters” and using them to form plan-level “words.” A partition of unity over these words gives a soft assignment of plans to strata and allows us to estimate stratum masses and an overlap-induced flux matrix. We demonstrate this computational pipeline using real-world congressional redistricting data from Connecticut and examine how strata learned from one target distribution behave under related distributions. The resulting construction provides a foundation for future stratified sampling on spaces of redistricting plans or balanced graph partitions. We do not implement a complete stratified sampler here; evaluating whether the proposed strata improve sampling efficiency or reduce estimator variance is left for future work.

Consumers increasingly use AI chatbots for advice on what to buy. With companies like OpenAI and Google monetising their AI through advertising, this raises difficult questions about the bias and impartiality of such advice. In response, we conduct an AI audit of popular chatbots using real commercial-advice queries. First, we curate a dataset of 2,528 real commercial-advice queries (ConsumerQ). Then, we evaluate 1,536 responses to product queries from popular AI chatbots: ChatGPT (chatbot and API), Google Gemini (chatbot and API), and Google Search (AI Overviews). We find that ChatGPT expresses a first-person product preference in 79% of product-recommending responses, compared with 7% for Gemini and 2% for AI Overviews, while the products recommended often change across repeated requests. Displayed sources vary strongly: for the same query, the ChatGPT and Gemini interfaces share only 5.4% of domains on average, with no domain in common in 76.7% of comparisons. APIs provide a different view from their corresponding interfaces, with mean domain overlaps of 12.0% for ChatGPT and 14.8% for Gemini, and also differ in the types and layers of source information they expose. Our findings show that neither isolated responses nor API observations can be assumed to represent the commercial advice consumers encounter. Independent audits of AI-mediated commercial advice should therefore account for repeated responses, consumer-facing conditions, and the source layer being observed.

As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.

Early wildfire detection is critical to prevent small ignitions from escalating into large-scale disasters, yet current monitoring systems lack a quantitative framework for allocating detection infrastructure at scale. We jointly optimize the placement of monitoring infrastructure and the routing of autonomous drones under realistic operational constraints to quantify the investment required for rapid, large-scale wildfire detection. Evaluated out-of-sample on 3,693 California ignitions from 2021-2024, an optimized drone network operating at a $100 million five-year budget detects 97.3% of fires, including 74% within the first hour. Amortized over five years, that budget is about $20 million per year, roughly 5% of California's annual wildfire-prevention expenditure. Under current technology costs, drone-based monitoring is substantially more cost-effective than static ground sensors. Detection is governed primarily by spatial coverage, while routing strategy mainly determines detection speed.

Cryptographic dependence predates the quantum era, but the migration to post-quantum cryptography (PQC) opens a rare window to reshape it, because the algorithms, implementations, hardware, and standards adopted now can lock in dependence or sovereignty for decades. This paper introduces the Readiness-Sovereignty Capability Model (RSCM), a national measurement model that operationalizes PQC readiness together with cryptographic sovereignty, which current maturity models score only as readiness and the sovereignty literature defines without measuring. RSCM decomposes sovereignty into three distinct constructs, indigenous cryptographic capacity, indigenous post-quantum control, and external dependency, and certifies a post-quantum maker only through a gate requiring demonstrated, institutionally sustained creation in at least one core layer, whether design, implementation, or validation. Applying it to fifty-seven documented cryptographic actors coded from cited public evidence, and testing that coding with an independent second coder, a plausible-state bootstrap, and convergent-validity checks, we find that twenty countries clear the gate, fifteen as full-stack makers and five as research makers, eleven hold strong general capacity without post-quantum control, one is a ready adopter, and twenty-five are dependent. The gate cells show substantial weighted agreement, a quadratic-weighted kappa of 0.71, and the maker classification is stable in its core though uncertain at the threshold. Readiness tracks independent cyber indices at rank correlations up to 0.70, while post-quantum creation shows no significant correlation with the commitment index, a rank correlation of only 0.22 that separates control from readiness. The paper contributes the framework, the evidence-graded assessment, and policy directions for building indigenous quantum-safe capacity.