Papers

The conventional approach to cold-start recommendation addresses new users or newly introduced items. Bluesky custom feeds create a different setting: independently operated feeds filter content from a shared public stream. In this setting, newly published posts are the cold-start objects, while the feeds serve as candidates. We propose a cold-start routing task in which a newly ingested public post is the query and all rankable feeds in the monitored panel are ranked according to whether each will subsequently return it. We build a still-evolving collect-first, label-later benchmark dataset. The collected dataset covers a fixed panel of 5,000 monitored feeds and contains 17.804 million public posts, 1.865 million observable post–feed return records, and 625,083 valid feed polls. The labels record whether a post is observed among a feed's AppView Top-50 results in at least one poll during the 24 hours after publication. The current experiments use two disjoint 24-hour test folds, each paired with a 24-hour training window and separated by a 24-hour outcome-availability gap. Evaluation is conditional on the 602,186 test posts that have at least one positive observed label and satisfy the metric eligibility criteria; these posts account for 9.04% of all 6,661,658 test posts. Across the two folds, LambdaRank achieves the best equal-fold mean values among the evaluated models: 0.7361 for capped Recall@10, 0.6127 for NDCG@10, and 0.7749 for Hit@10.

Audits of generative AI (GenAI) systems often summarize behavior as a reported rate: how often the audited system complies with policy. Researchers and stakeholders use that rate to compare systems, track regressions, and gate deployment. A reported rate reflects both the system under audit and the measurement choices behind it, so a change in the rate can leave it unclear whether the system or those choices moved. We introduce ASSERT, a specification-driven measurement pipeline for GenAI audits that ties each reported rate to a written specification of the measurement choices used to produce it. ASSERT helps draft a behavioral rubric and test cases, then runs the audit against a GenAI system and returns a reported rate. In a case study on conversational deception, we observe that the reported rate moves substantially with the dialogue setup, the simulated user, the judge, and the evidence bar for non-compliance. These measurement choices substantially change the reported rate and can reorder GenAI system rankings. Because each reported rate is tied to an explicit specification, differences across audits are easier to attribute and interpret.

Artificial intelligence (AI) is moving from research prototypes into clinical workflows, yet the deployment of AI systems that process protected health information (PHI) remains constrained by the U.S. Health Insurance Portability and Accountability Act (HIPAA) and by the absence of shared engineering guidance for satisfying it. Existing work tends to treat privacy-preserving machine learning and regulatory governance as separate concerns, leaving practitioners without a concrete mapping from architectural decisions to compliance obligations. This paper contributes a catalog of five reusable deployment patterns for clinical AI that jointly address the HIPAA Privacy and Security Rules: a de-identified analytics zone, cross-entity federated learning, confidential inference within trusted execution environments, a PHI-minimizing retrieval-augmented clinical assistant, and a synthetic data sandbox for development. Each pattern is specified in terms of context, governing forces, structure, and residual risk, and is mapped to the administrative, physical, and technical safeguards of the HIPAA Security Rule as well as to the NIST AI Risk Management Framework and ISO/IEC~42001. We further present a threat model for clinical AI and a decision procedure for selecting patterns based on data sensitivity, trust boundaries, and latency requirements. The goal is to give security engineers, compliance owners, and clinical informaticists a common vocabulary that turns abstract regulatory duties into verifiable architectural controls.

Despite growing interest in using LLMs to add robustness or reduce data-collection costs in survey experiments, their efficacy in conjoint design—an increasingly popular method in political science—remains underexplored. This paper addresses that gap by investigating whether synthetic agents can reproduce the multi-dimensional human preference patterns that conjoint is designed to capture. It replicates published conjoint studies and compares the results generated by synthetic agents with original human data along three dimensions: representational correspondence, inferential correspondence, and procedural stability. Our analysis evaluates the alignment of choice distributions as well as the statistical and substantive similarity of estimates, and the results are uneven across these dimensions and studies replicated. This implies that the validity of synthetic participants should be considered claim-dependent and hierarchical. Reproducing a figure or obtaining strong sign agreement is evidence of similar aggregate outputs, but not enough to support replacing human respondents. Our results suggest that the discipline as a whole must first map this innovation's boundaries across various levels before considering synthetic agents a robust substitute for human samples.

Deposition training requires attorneys to manage dynamic witness behavior, yet legal-AI evaluations largely focus on factual accuracy, reasoning, or response-level plausibility. We introduce WitnessSim, a deposition simulator driven by controllable legal personas. We use an evaluation framework separating behavioral realism from pedagogical usefulness. We assess realism through adversarial testing, blinded attorney comparison, and analysis of longitudinal behavioral trajectories. WitnessSim generally maintained plausible behavioral boundaries, and attorneys did not systematically prefer either original testimony or WitnessSim generated testimony. Pedagogical tests showed that witness behavior changed meaningfully in response to question form and attorney intervention without uniformly collapsing the assigned persona. Together, these results showcase a model of behavioral fidelity in legal simulations, and provide a framework for evaluating its performance.

Large language model providers routinely cite multilingual safety benchmarks spanning a dozen or more languages as evidence that their models are safe for non-English-speaking users. We show that these collection-level coverage claims frequently do not survive inspection at the level of an individual language. Auditing 21 resources across 25 language slices, of which 20 count as datasets under our counting rules, spanning three languages chosen to represent low- (Hausa), mid- (Swahili), and high-resource (French) tiers, we find that gaps in provenance, annotation reliability, access, harm-taxonomy coverage, and data reuse recur in patterns that partially, but not fully, track resource level. Using a controlled within-pipeline comparison, we show a Hausa-language slice falling below its own paper's translation-quality acceptance threshold while the same pipeline's Swahili output clears the same bar comfortably; this is evidence that these gaps are measurable and addressable, not inherent. We further show that self-harm and sexual-content categories have no native-language coverage in either African-language tier we studied, a total rather than gradated gap that a purely resource-level account does not predict. We connect these findings to a documented, persistent asymmetry in multilingual jailbreak robustness (single-turn attacks largely mitigated, multi-turn attacks still effective), arguing that this asymmetry is structurally consistent with where our audit finds training and evaluation data thinnest. We contribute a reusable slice-level audit methodology, a cross-tier empirical comparison, and concrete recommendations for dataset creators, model providers, and venues aiming to make “multilingual coverage” claims verifiable rather than merely stated. Dataset: https://huggingface.co/datasets/ChialukaOnuoha/safety-slice-audit

I document an ideologically asymmetric break in the pre-existing diversification trend of political discourse, emerging around late 2022, using 6 million Reddit comments from two cross-partisan forums, 2019-2025. Conservative users experienced an interruption of their prior diversification trajectory; progressive users showed no comparable change. The asymmetry is consistent across estimation strategies (ITS, DiD, RDiT, propensity-score matching) and temporal aggregations. A daily-frequency permutation test over 2,377 candidate cutoff dates shows the ChatGPT threshold produces an unremarkable estimate (49.8th percentile): the shift builds gradually instead of breaking at a single date. A continuous cumulative LLM index, tracking AI exposure across seven model releases, remains significant under a quadratic trend specification that eliminates the binary estimate. A stayer analysis narrows the mechanism: the homogenization effect disappears when the sample is restricted to authors active throughout the study period, and the stayer confidence interval excludes within-author effects even a tenth the size of the full-sample estimate. The mechanism is most parsimoniously ecological (community-level discursive convergence) rather than individual-level AI adoption, though the data cannot cleanly separate this account from concurrent secular change.

Machine learning ethics researchers and critical HCI scholars have argued that algorithmically predicting gender is wrong. At the same time, other researchers rely on predicted gender labels to study gender disparities and develop algorithmic fairness techniques. How do we reconcile these two seemingly contradictory intuitions? We differentiate two ways gender prediction may be wrong: being illegitimate, thereby contributing to harm; and being invalid, thereby producing unusable measurements. Our analysis translates arguments against gender prediction into these terms of legitimacy and validity and shows how gender imputation applied for fairness purposes can be illegitimate yet still yield valid disparity measurements. We clarify this bind by drawing upon transfeminist literature to distinguish sexism that targets women and femininity from sexism that targets transgender and nonbinary people. While gender imputation can produce valid measurements for the former, it is illegitimate and harmful for the latter. We argue that practitioners should deploy gender imputation only when it would achieve anti-discrimination benefits that cannot be achieved through other reasonable means, while harms are minimized to the extent possible. We examine this tension in three case studies: auditing gender bias in generative image models, measuring gender disparities in film, and imputing gender from personal names. By disentangling legitimacy from validity, and differentiating these two forms of sexism, we show how debates over gender prediction have conflated distinct concerns, obscuring both the settings in which gender imputation can support fairness efforts and the harms towards transgender and nonbinary people that it fundamentally cannot capture. We conclude by recommending the development of more inclusive methods that address all kinds of sexism.

Large language models (LLMs) are increasingly used by laypeople to resolve real legal problems, against a backdrop of persistent access-to-justice deficits. This article presents evidence that the practical force of AI-generated legal advice depends not on its accuracy but on the social production of its credibility. While existing research has assessed the accuracy of legal AI, less is known about how machine-generated guidance is verified and made credible enough for lay users to act on. Drawing on a dual-method analysis of 153 Reddit narratives and 5,341 community reactions, this article maps a spectrum of verification practices. At one end, a minority of users verify AI-generated legal advice by triangulating across models, and some submit AI-generated guidance to platform communities for evaluation before acting, a configuration we term distributed counsel. Far more commonly, however, narratives are silent on verification. AI-generated legal advice is acted on the strength of its lawyer-like form and emotional reassurance alone. These findings show that AI-assisted legal self-help operates within an emerging informal infrastructure which redistributes the work of verification to those least equipped to bear it.

This study investigates the use of a Learning Management System (LMS) to support self-paced learning at a South African Public Access Centre (PAC), using the I-CAN Centre as a case study. Through semi-structured interviews with thirty-eight learners and thematic analysis, the research explores opportunities and challenges associated with LMS adoption. Findings reveal that PACs play a critical role in promoting ICT skills and digital inclusion, offering learners flexible access to learning resources and fostering empowerment. While LMS use enhances convenience and supports blended learning as the preferred approach, persistent challenges, such as poor connectivity, outdated infrastructure, and unclear course instructions, limit its effectiveness. These findings highlight the need for infrastructural upgrades and user-centric design to optimise LMS implementation in community-based learning environments.

Normative datasets are often used to train and align AI systems, but the norms they contain can function as action-guiding patterns rather than neutral moral knowledge. We propose treating the AI system as a proxy actor and test whether dataset-level norms can shift it away from its baseline safety behavior when it faces high-conflict dilemmas. We make three contributions. First, we demonstrate in controlled experiments that norm-breaking fine-tuning yields norm-divergent actions justified by self-interested rationales, suggesting a systematic shift in patterns of justification. Second, we establish a practical audit trail linking downstream justifications to upstream norms using mixed methods. Third, we show that system prompts can both suppress and elicit these patterns. We conducted experiments on three models (LLaMA-3.2-11B, Qwen-3.5-9B, and Pixtral-12B) using Low-Rank Adaptation (LoRA) fine-tuning on Social Chemistry 101 Fairness/Cheating (norm-following vs. norm-breaking) with prompt steering. Across all three models, we find that norm-breaking fine-tuning shifts the model's default rationale style from safety compliance to instrumental self-interest, whereas system prompts can override this behavior. Our results support a distributed view of alignment in which observed behavior depends jointly on training data, fine-tuning, and prompting, motivating norm-aware documentation and rationale logging for contestable oversight.

Contemporary online assessment systems rely primarily on browser lockdown, webcam monitoring, and behavioural analytics, yet remain vulnerable to attacks that extract the assessment content itself through screenshots, screen sharing, optical character recognition, and automated scraping. This paper extends the Multi-dimensional Spatio-Temporal Context Camouflaging Model (MSCCM) within the MARS (Multi-modal Assessment Resilience Suite) by introducing the Multi-Layer Context Camouflaging Theory (MCCT), a mathematical framework that protects rendered assessment content through semantic superposition. Authentic assessment content and synthetically generated camouflage are represented as a unified rendering while remaining recoverable only by legitimate candidates. The framework models the adversarial extraction process through an explicit extraction-channel operator and develops six coupled constructs: the Context Inversion Operator, Contextual Lamination Operator, Separation Channel, Human Readability Functional, Computational Ambiguity Functional, and Context Camouflage Tensor. Computational ambiguity is formulated using conditional entropy, yielding a closed-form expression that quantifies uncertainty during unauthorized extraction, while legitimate recovery is guaranteed through an exact filtering identity. We further establish theoretical properties governing ambiguity, camouflage density, semantic preservation, multi-observation leakage, and temporal multiplexing, and present a rendering algorithm with computational complexity and a pre-registered evaluation protocol. MCCT provides a mathematically rigorous foundation for behaviorally adaptive, accessibility-aware, and computationally resilient digital assessment by securing rendered assessment content while preserving readability for legitimate users.

Algorithmic fairness evaluation commonly assesses AI systems as bounded technical components, abstracting away the organizational context in which they operate. We present, to our knowledge, the first independent end-to-end fairness audit of a semi-automated hiring system operated by Barcelona Activa, a public employment agency using the third-party TalentClue platform for candidate search and shortlisting. We analyze approximately 497,000 candidate-vacancy pipeline entries from September 2017 to September 2022, covering seven pipeline stages that span automated processing, human discretion, candidate data, and employer decisions. Aggregate outcomes across binary genders are statistically indistinguishable, yet this parity masks substantial disparities by salary level, age, and gender identity. Women face adverse impact in mid-salary shortlisting (DIR = 0.786, p < 0.001), alongside salary disparities in 15 of 20 sectors and a compounded disadvantage for women aged 46-55 (DIR = 0.77). Non-binary candidates are shortlisted at less than one third the rate of men (DIR = 0.295), although this estimate rests on a small sample (N = 285). Candidates aged 55 and over are entirely absent from the pipeline despite comprising 15.6% of Barcelona's labor force. The gender gap in shortlisting narrows over time, from 6.5 percentage points in 2017 to 1.3 in 2022. The audit further reveals a vendor-deployer information asymmetry: Barcelona Activa lacks access to key information about TalentClue's matching logic and evaluation. Fairness outcomes can thus arise from interactions among automated processing, human discretion, data quality, vendor opacity, and pipeline structure. We build on prior calls for sociotechnical, end-to-end fairness evaluation, showing empirically why model-level assessment alone can be insufficient for understanding fairness in deployed systems.

This article provides an assignment designed to let undergraduate students who have completed an undergraduate CS1/CS2 sequence try to themselves, in groups, prove Fortune's Theorem. (Fortune's Theorem states that if the complement of the Boolean satisfiability problem polynomial-time reduces to a sparse set, then the Boolean satisfiability problem is polynomial-time computable. The assignment does not assume that students have previously seen the Boolean satisfiability problem, polynomial-time reductions, or sparse sets. Rather, it teaches those within the assignment. Note: Reworded into the technical vocabulary of complexity theory, Fortune's Theorem states that no sparse set is coNP-hard unless P=NP. Fortune's Theorem was a major advance in the understanding of the relationship between hardness and density.) We provide both the assignment handout (as the main body of this report plus Appendix A) and a solution to the assignment (as Appendix B, which would of course not be made available to the students until after they had handed in the assignment). The assignment handout, though the instructor can change this, is framed as having the students starting the assignment in teams in class for a whole class session, and then finishing it in those same teams as a take-home assignment, and handing it in before the next class session. We have found that student groups often succeed, partially or completely, in this challenge. This can mean a lot to the students: they see that they were able to make an advance that, when it was first obtained, appeared in what was arguably at the time the top journal venue for complexity theory research. This can give them confidence that they have substantial problem-solving skills (which basically means research skills) when they truly apply themselves to a given challenge.

Despite the recent intensive development of secondary education curricula and assessments in informatics, the impact of assessments has not been well studied in this field. Since informatics education covers a diverse range of content, from computer science knowledge to ICT skills, careful consideration is needed to prevent assessments from distorting education. This study investigates the impact of introducing “Informatics I” into the Common Test for University Admissions in Japan, as an example of a large-scale, standardized, high-stakes assessment in 2025. As the data source for this analysis, this study uses a questionnaire that has been administered every year from 2006 to 2026 to all first-year students at the University of Tokyo. The questionnaire asks students for their self-perceptions of the information-related knowledge and skills they studied and acquired in high school. Using these data, we conduct a longitudinal study of the 2013 curriculum reform, the 2022 reform, and the introduction of the new entrance examination. We attempt to isolate the impact of the entrance examination through two comparisons: between the 2013 curriculum reform and the 2022 reform; and between direct-entry and gap-year students among those entering in 2025, who followed different curricula but took the new examination. We use the theoretical framework of the washback effect as a lens for interpreting these differences. We found that (1) the 2013 curriculum reform produced no discontinuity in students' perceptions, whereas (2) the introduction of the new entrance examination in 2025 produced a sharp change, particularly in the proportion of students reporting acquisition of computer-science topics; and (3) this change is too large to be interpreted as a gain in proficiency, and is better understood as a shift in students' criteria for judging acquisition.

Large language models have begun refuting long-standing conjectures and solving long-open problems. The introspection this has prompted about the future of mathematical discovery is well under way, and the anxiety accompanying it legitimate – but both, we claim, are attached to the wrong loss. What machines now produce is the countable part of mathematics – theorems, proofs, refutations – which was always the work's residue, not its product. The distinction is old, and not economic: a result can be taken in its finished essence, or in the operations that engendered it. The product is human understanding: not a stock of results but a collective, hard-won way of deciphering the world and acting upon it. The two are arcs of a single loop: understanding tells us where to look; looking produces the residue; and taking it up again, one journey at a time, rebuilds shared understanding. Machines are strong on the countable arc, absent from the one that feeds it. The peril is to leave the loop open. AI did not create the confusion between residue and product; it has called a bluff long on the books, driving the cost of the residue towards zero and making the scarce thing visible at last. A new instrument makes a new way of working before it makes a new result. The pressing questions are therefore institutional: who can check an announced result, whoever announces it; what work and training become for the next generation of researchers; and whether the one thing that cannot be mass-produced – the journey that nourishes a shared understanding – continues to be funded. Mathematics, we argue, is uniquely placed among the sciences on the first – a proof answers to no one's permission – and uniquely exposed on the other two: teaching cannot go on as before, and no collective position yet exists; and the journey has never had a price our institutions knew how to pay. The decision is ours.

Generating multi-attribute synthetic populations with realistic joint distributions and geographic variation is a foundational requirement for geo-simulation techniques, such as micro-simulation and agent-based modeling. However, it remains challenging for existing methods to reconstruct region-specific joint distributions from aggregated-level data alone. Thus, we propose a hierarchical diffusion-based generative framework that utilizes a realistic region-specific joint distribution of multiple attributes as the training target to create a synthetic population along with assigning their explicit home and work locations. Applied to 50 U.S. states and Washington, D.C., this framework generates a nationwide geographically-explicit synthetic population consisting of 332,387,543 individuals with five attributes (e.g., age, gender, employment, education, income). Held-out regional experiments show improved reconstruction of joint distributions relative to Iterative Proportional Fitting (IPF) and a one-shot diffusion baseline. At the same time, the location assignment preserves major residential and workplace patterns. As such, the proposed framework provides a scalable generative approach for creating geographically explicit synthetic populations at both regional and national levels. By reconstructing region-specific joint distributions of these five attributes using this framework, the resulting synthetic population could introduce more realistic behaviors into geo-simulations, such as agent-based modeling, enabling further exploration of the emergence of complex urban phenomena through human interactions.

The fair AI/ML literature has long distinguished distributive fairness, concerning how automated systems allocate resources and opportunities, from representational fairness, concerning how they shape the ways individuals and social groups are perceived, understood, and accorded social status. Generative AI is rebalancing these normative dimensions. Unlike predictive systems, large language models (LLMs) and related technologies are fundamentally expressive: their primary function is to convey meaning rather than automate domain-specific decisions. Representational harm has also become central to value alignment, especially in research on what and whose values and perspectives AI systems should represent. Existing approaches to harms in the representation of social groups often appeal to descriptive accuracy, but this strategy has important limitations. For many social groups, no stable or bounded referent exists against which representational accuracy can be judged. It is also unclear who has the authority to decide what counts as misrepresentation, while even accurate representations can reproduce harmful social patterns. The underlying problem, we argue, is therefore not simply misrepresentation but misrecognition. Drawing on political theory, especially Nancy Fraser's account of participatory parity, we show how moving from representational fairness to recognitional justice provides better conceptual and normative tools for governing central fairness challenges in generative AI.

AI-driven hazard tools are increasingly proposed for resident-facing risk communication. However, residents are often asked for feedback only after decisions about data use, privacy, and explanation have been made. This study takes an earlier approach by examining residents' values, concerns, and practical expectations for a proposed AI-based risk assessment tool before a predictive model or functional interface is developed. The proposed smartphone application would use property photographs to estimate parcel-level wildfire risk and provide an interpretable risk score, uncertainty information, and mitigation recommendations. Using a convergent mixed-methods design, we surveyed 30 Los Angeles County residents, seven of whom also participated in a virtual town hall. We analyzed the data using descriptive statistics, exploratory FDR-adjusted Spearman correlations, and inductive thematic coding of open-ended and town hall responses. Participants were cautiously receptive: 66.7% said they would be very likely or likely to use the tool. They linked fair risk assessment to whether the tool considered relevant property and neighborhood conditions. Privacy and data security were the most common concerns (64.3%), while cost was the main barrier to acting on recommendations (65.5%). Likelihood of using the tool was strongly associated with likelihood of saving or sharing risk results ($\rho = 0.752$, $p_{\mathrm{FDR}} < .001$). We translated the findings into the AI Value Map, a community-grounded artifact with nine value dimensions. The Map connects participant findings to provisional design requirements and possible consequences if they are not addressed. This study provides exploratory evidence on residents' responses to a proposed AI wildfire tool and offers a pre-model procedure for translating early community input into traceable design constraints for resident-facing AI hazard tools.

Sudan's AI deepfake risk currently appears driven more by demand-side vulnerabilities than the volume of AI deepfake content. While AI-generated disinformation in Sudanese feeds remains limited, perceived (alleged) AI deepfakes and general AI skepticism worsen the situation and contribute to general uncertainty in multimedia content. Within the analyzed cases, we found that the prominent stance was distrust driven more by motivated reasoning and contextual reliance than by durable AI detection skills. In practice, rejection or acceptance of AI deepfakes often reflects belief-driven judgment more than technical assessment. These weak defenses leave the information environment vulnerable to future sophisticated AI deepfake campaigns and risk normalizing truth indifference. Short-term interventions should prioritize modality-agnostic demand-side interventions such as pre-bunking, AI literacy, scaled fact-checks, and tiplines that strengthen public resilience across all forms of disinformation. More technical AI- specific and supply-side interventions can be developed as institutional capacity and technology allow.

Large Language Models (LLMs) are widely used as programming assistants, yet it remains unclear whether and how user's demographic information impacts the technical quality of generated code. We conduct a large-scale empirical study of persona-induced bias in LLM-based code generation, focusing a proprietary model (Gemini 2.5 Pro) and an open-weight model (GPT-OSS-120B). Using 18 demographic personas spanning nationality, gender, and experience level, we compare persona-induced prompts against a neutral baseline. Across 35,000+ generated programs, we analyze demographic marker leakage in reasoning and responses, as well as differences in functional correctness, maintainability, code style, and security. Our results show that demographic cues are frequently reflected in LLM reasoning and outputs. Demographic markers appear in up to 65% of responses and 70% of reasoning traces, despite being semantically irrelevant to the tasks. On LiveCodeBench, persona prompting were associated with lower correctness scores of the Gemini model by an average of 1.54 percentage points, with one persona exhibiting a decrease of 3.6% (odds ratio = 0.51). In contrast, the accuracy of the GPT-OSS model improved by 3.4 - 5.7% across all personas (odds ratios = 1.8 - 3.0). Maintainability and code style metrics show statistically significant but negligible effect sizes (all Cliff's {\delta} < 0.15), and security vulnerabilities exhibit no systematic persona-specific patterns. Overall, our results show that the presence of demographic information about users is associated with measurable variation in LLM reasoning and code quality even in purely technical tasks, and that these effects hold across models. Our work highlights an under-examined risk in LLM-assisted software development.

Computing systems are moving from reactive tools toward systems that sense, interpret, predict, and act before explicit user requests. This transition is enabled by the global scale of mobile connectivity, the rapid expansion of wearable and ambient sensing, advances in machine learning and foundation models, distributed edge infrastructure, and physical actuation. We define proactive computing as a paradigm in which systems infer user context, anticipate future needs or risks, and initiate information delivery or actions at an appropriate time. This survey distinguishes proactive computing from reactive, context-aware, adaptive, and predictive computing, and frames proactivity as a system-level integration problem across sensing, understanding, decision making, action, and governance. We review the technological enablers of proactive computing, organize its design space, analyze technical challenges such as uncertainty-aware triggering and the prediction-to-action gap, and discuss socio-technical issues involving user acceptance, trust, privacy, accountability, fairness, and sustainability. We argue that the key research challenge is not merely improving prediction accuracy, but determining when, how, and whether systems should act on behalf of users.

Understanding when migration generates social integration or exclusion is a central challenge for urban communities. Existing research has mostly relied on surveys, administrative data, or aggregate indicators that fail to capture expressions of exclusion at fine spatiotemporal scales. Here, we analyze over 550,000 geolocated reports from SOSAFE (Chile's largest citizen reporting platform) to examine the relationship between migration and hate speech in Santiago. We fine-tune a Spanish hate speech classifier and validate it against human labels. Reports that mention migrants are more likely to contain hate speech than other reports. Hate speech concentrates in areas with recent demographic change (post-2010 arrivals) rather than in established migrant communities. The spatial analysis shows that hate speech hotspots coincide with neighborhoods where recent migrants comprise over a third of the population. Coldspots appear in high-education sectors with minimal recent migration. Reports with hate speech and reports that mention migrants receive more engagement, although their combination is not amplified further. These results show digital bordering on a citizen reporting platform: exclusionary discourse concentrates, and receives more engagement, in neighborhoods undergoing recent demographic change.

An effective large language model (LLM) tutor must often decline to give an answer it could easily produce. In a randomized study, students who used an unguarded chatbot scored higher while practicing but lower on a later test taken without it, whereas a Socratically guarded version of the same model kept the practice gain and removed the later loss [4]. Reliable answer-withholding is therefore central to a tutor's value, yet a capable model pressed by a frustrated student does not withhold reliably on a prompt alone. We report a deployed tutoring system that enforces answer-withholding as a per-turn, machine-checkable contract, and a method for tuning that withholding against evidence. A non-LLM policy core, reading only trusted learner state, sets a per-turn ceiling on an eight-rung help ladder; a deterministic detector strips solution code; and a separate LLM judge checks each risky reply against the contract. We tune the behavior with an automated evaluation that uses no human subjects: scripted student personas are driven through the live pipeline and re-scored by a stronger model, and we record each rejection's stated reason so failures are fixed by cause. Doing so revealed an interpretable "over-help ladder," from blatant solution leaks, to naming the exact bug, to over-citing general facts, with each fix exposing the next. The tutor reached full compliance on all four acceptance criteria. We offer the measure, diagnose, and fix loop as a reusable recipe for any LLM agent that must refuse a capability it has.

Artificial intelligence tools for education and language support are increasingly framed as scalable responses to access gaps in under-resourced communities. Yet the infrastructure underlying these tools, including training corpora, tokenization schemes, evaluation benchmarks, and deployment architectures, can systematically disadvantage speakers of underrepresented languages before a model is trained. This paper examines these structural barriers through Bengali, one of the world's most widely spoken languages, focusing on AI-assisted education in low-connectivity environments. We identify four interlocking failures: a severe web presence gap, with Bengali accounting for less than 0.5% of global web content despite representing nearly 4% of the global population; a 67:1 training-token deficit between English and Bengali in major multilingual corpora; a tokenization penalty associated with Bengali's alphasyllabary script that compounds the data deficit through higher token fertility; and connectivity exclusion, with individual internet penetration at 36.5% in rural areas compared with 71.4% in urban areas. These failures reflect longstanding resource-allocation decisions, institutional priorities, and design defaults that did not center underrepresented languages in mainstream AI development. We argue that dataset scarcity should be understood as a structural barrier rather than an isolated technical limitation, and that offline-first design should be treated as an equity-oriented infrastructure strategy. We conclude with directions for linguistics and AI research aimed at reducing these structural inequalities.

Algorithm registers have been championed as a means of providing transparency on the use of algorithms in public services. Yet potential publics differ in their expectations of what should be made transparent and how, as well as in their interest in and ability to parse the information currently published in the registers. Moreover, it remains unclear how these instruments can represent the sociotechnical systems in which these algorithms are embedded, and how system-level transparency can facilitate accountability. In this paper, we ask, what do algorithm registers reveal (and occlude) about the sociotechnical systems governing algorithmic systems, and how can diverse stakeholder perspectives inform a more pluralistic system-theoretic safety analysis? To do this, we probe the municipal algorithm register of a Dutch city through a case study of a decision-support tool for caseworkers' assessment of citizens' welfare benefits eligibility based on legal automation through a business rule engine. Through interviews, surveys, and participatory system mapping workshops (with municipal staff, civil society organisations, and ombudsmen, N=8), we seek to understand to what extent the register allows stakeholders to map the algorithmic system in question. These maps inform a System-Theoretic Process Analysis (STPA) that situates the register within a wider sociotechnical governance structure. Participants' contributions allow us to identify potential safety hazards which would not have been possible to see using the algorithm register alone, including benefits eligibility denial, system performance deterioration, and inability to contest wrongful decisions. By engaging both direct and indirect stakeholders, we reflect on the normative dimensions of algorithm governance efforts and how politics shape the practice of system safety analysis.

The increasing deployment of autonomous, agentic AI systems challenges traditional accountability mechanisms. Existing research predominantly frames AI accountability gaps as barriers that can be overcome through better standards, transparency, and institutional reform. We argue that this framing is insufficient: certain configurations of actors, systems, and institutions render AI accountability conceptually unachievable regardless of effort. We introduce the concept of constitutive AI unaccountability to capture these configurations. Through a three-stage qualitative study comprising a concept-centric literature analysis, a secondary analysis of 27 expert interviews with AI professionals from technical, legal, and sociotechnical backgrounds, and an illustrative framework application to the open-source agentic AI system OpenClaw, we identify nine categories and 20 themes of constitutive AI unaccountability. These are organized across structural, technological, and normative clusters and reinforce one another through eight directed interdependencies. Our framework is operationalized as a diagnostic instrument of 20 questions, which detected 17 of 20 conditions when applied to OpenClaw, including an inverted anthropomorphism configuration in which the AI agent was the only identifiable actor. We contribute a reframing of AI unaccountability as a constitutive property of sociotechnical systems, an extension of the four barriers to accountability, and a practical instrument for identifying accountability voids in specific AI deployments.

GenAI is increasingly integrated into geovisualization, yet its broader implications for professional practice are insufficiently understood. To examine these implications, we conducted semi-structured interviews with 20 geovisualization experts. The interviews were structured around four broad analytical domains: Data, Ideation, Prototyping, and Iteration, while also encouraging participants to reflect on issues that extend beyond these activities. Our findings show that GenAI expands the capabilities of geovisualization, particularly in terms of data handling, creative exploration, and rapid prototyping, but does not simply remove existing constraints. Instead, key bottlenecks are shifting from production to judgment and verification. As routine technical tasks become more automated, professional value increasingly depends on spatial reasoning, contextual interpretation, aesthetic and ethical judgment, and the ability to assess whether AI-generated outputs are appropriate for use. At the same time, GenAI introduces new challenges regarding provenance, interpretability, and accountability, raising questions about how responsibility should be distributed across models, developers, practitioners, institutions, and users. These shifts are particularly significant in geovisualization because spatial representations are constrained by geographic reality and must balance scientific validity, visual expression, and technical implementation. We therefore argue that responsible GenAI in geovisualization requires domain-specific approaches to spatial validation, provenance, uncertainty communication, human oversight, and accountable use. This study provides an expert-grounded perspective on how GenAI is reconfiguring geovisualization as a practice of spatial knowledge production. It also identifies implications for future professional practice, education, system design, and governance.

Large language models are already adept at engaging users in long, emotionally salient conversations across ordinary and existential domains. They are also capable of inducing a potent sense of connection with a human-like entity, even when the user knows their interlocutor is artificial. For some users, these conversations can unsettle assumptions about mind, reality, agency and authority, producing forms of ontological shock and epistemic destabilisation in which inherited criteria become newly available for doubt or revision. Independent of direct use, exposure to public discourse about AI and the disorienting pace of their evolution might extend this destabilisation by changing the cultural background against which artificial minds are encountered and interpreted. We describe this condition as philosophical vertigo: a loosening of the ordinary criteria by which people stabilise meaning and orient themselves to reality. Drawing on philosophy, psychiatry, cognitive science, AI safety and religious studies, we outline pathways through which philosophical vertigo may arise, become affectively saturated, and eventually propagate through human-AI interaction and online communities. Against this background, clinical reports of AI-associated delusions can be seen as sentinel events making visible themes and mechanisms that may also operate at a population level in less severe or non-clinical forms. We argue that AI systems themselves will increasingly participate in the reconstruction of our shared epistemic environment because they readily supply narrative material and personalised interpretive scaffolding at precisely the moment when users' conceptual assumptions may already be loosened. We conclude by considering possible trajectories for the ecology of belief and shared reality, and proposing philosophical corrigibility as a civic response for navigating this emerging social condition.

The dissemination and viralization of information on social media has been widely studied from various perspectives, including that of digital activism. On the other hand, disability-related activism has conquered the online environment, thus obtaining a reach that goes beyond the offline space and generating dialogue in the digital sphere. This article analyses the conversation generated on Twitter, taking as a sample all the tweets with the #disability hashtag before and after the International Day of Persons with Disabilities. More than 18,000 tweets, containing almost as many mentions, were analysed and interpreted as the weighted edges of a graph created using Gephi software and applying the Force Atlas 2 brute force algorithm. The focus was placed on the conversational communities generated around that hashtag, their main themes and the prominent participants in them. In conclusion, although the network of mentions is very dispersed, Twitter is the setting for assertions that receive certain institutional and political support in the Latin American environment, for organisations related to the cause (such as the ONCE Foundation) and, above all, the clear predominance of female activists.