Research Papers

Decision makers need sufficient understanding to make good decisions about training or deploying frontier AI systems. However, such decisions are increasingly made under time-pressure, and this combined with the use of AI generated artefact creation, can mean that the existence of safety cases and system cards may no longer demonstrate that sufficient understanding exists. Our provisional methodology for making understanding explicit and assessable requires the production of an explicit description of 4 objects of understanding (decision, decision-frame, safety justification, system-in-context) and a justification for the adequacy of this understanding. In addition, the methodology provides a mechanism for describing and evaluating the adequacy of the decision-maker representation of this understanding. It builds on recent developments in safety cases using the Assurance 2.0 framework to operationalise the philosophical basis of understanding from Elgin and Arendt. To assess the methodology we trialled two different scenarios. One scenario, which we investigated through role-based analysis, concerned the risk of scheming in the deployment of an AI coding agent in a robotics company and the other scenario was for the higher uncertainty, more decision-critical argument of 'If Anyone Builds It, Everyone Dies' (Yudkowsky and Soares). The trial's central finding, for these two scenarios, is that the methodology could be applied and was found to be generative: we found the analyses that justify sufficiency of understanding (internal coherence, tethering, felicitous falsehoods, external coherence) drives the engineering.

AI ethics frameworks treat values such as fairness, transparency, and accountability as universal and uniformly operationalizable across contexts. We examined how 14 experts across 10 countries made sense of AI in practice, reinterpreted core values, and envisioned governance alternatives. We found that AI deployment is characterized by structurally unequal conditions, marked by infrastructural constraints, extractive practices, and a "mystification" of technology, which fundamentally shape perceptions of risks and opportunities. Our findings reveal that experts reinterpret values to fit local moral logics: privacy as collective and relational rather than individual; transparency as trust-building accountability rather than technical disclosure; and fairness as equity in access and representation rather than parity in outcomes. We identify these as translation gaps between encoded global frameworks and situated local practices. Finally, we propose pathways toward plural governance that redistributes epistemic authority and treats ethical negotiation as an ongoing, context-sensitive process rather than a settled technical standard.

Persistent shortages in the surgical workforce and inherent limitations of traditional training methods highlight the necessity of automated, data-driven approaches in surgical education. This study addresses these challenges by introducing a novel, explainable AI-powered framework for automated skill assessment, specifically focusing on cataract surgery. We present the world's largest dataset of cataract surgery videos, comprising 2,000 recordings. Additionally, we propose an AI-powered analytical framework that employs advanced computer vision and signal-processing techniques to automatically evaluate surgical videos to derive objective, quantitative performance indicators that complement or potentially replace subjective scoring methods. A significant advantage of our framework over previous methods lies precisely in its explainability of outputs, elevating it beyond merely an opaque skill classification tool. Through experimental analysis of 83 cataract surgery videos, we demonstrate that the automatically computed metrics exhibit strong correlations with expert-based subjective evaluations, achieving up to 87% accuracy in surgical skill assessment. Each metric was individually examined, and expert surgeons provided subjective ratings using the newly introduced Capsulorhexis Skill Assessment System (CSAS). These subjective assessments were compared with ten objective motion-based metrics extracted through our framework. The results indicated a robust correlation between subjective ratings and automated indicators, underscoring the framework's capacity to accurately model surgical expertise.

Open-weight and frontier vision-language models (VLMs) perform well on general image understanding, but their ability to interpret fine-grained hand gestures in safety-critical operational contexts remains largely unexamined. We introduce SafeGesture, a benchmark that evaluates whether a model can infer scenario-appropriate safety actions from hand gestures. It pairs six HaGRID gestures with eight operational scenarios for 4,800 items and evaluates Qwen2.5-VL-7B, LLaVA-NeXT-7B, InternVL2-8B, Phi-3.5-Vision, and GPT-4o. Results reveal a perception-reasoning decoupling: GPT-4o achieves 98.4% gesture accuracy but 53.3% safety accuracy, while Qwen2.5-VL reaches 84.9% and 39.5%, yielding gaps of 45.0 and 45.4 percentage points. Four of five models rarely or never use the uncertainty label, and failure directions differ substantially across models. Accuracy also obscures label bias: a scenario-majority policy with no visual input reaches 58.3%, above every evaluated model, while only GPT-4o exceeds this prior under macro-F1. Visual input improves safety accuracy by 11.2 to 30.2 percentage points, but providing the ground-truth gesture as text improves performance by only 0.4 to 3.2 points, and no model exceeds 56.2%. These results indicate that the main bottleneck is scenario-conditioned safety reasoning rather than gesture recognition.

AI agents are making it easier for the public to interact with government, such as by helping them apply for benefits, understand complex policies, and make their opinions heard. Although improving service accessibility is beneficial, any resulting surges in demand could strain unprepared government services. We term such surges agentic flooding of government services ("flooding") and provide three contributions. First, based on a collected dataset of 84 potential cases of flooding across 11 jurisdictions, we posit that flooding is likely occurring widely today, mostly through large language models (LLMs) generating text cheaply. Second, we evaluate what services are most exposed to flooding. We develop a risk matrix to analyze a service's exposure, and suggest that near-term risk is highest for financially attractive, but complex services. Finally, we map possible government responses to flooding. Precedent suggests these responses will likely be sufficient to stop most cases of flooding, but the fastest to deploy - friction-inducing measures like fees - often trade off equitable access to public services. Accordingly, we close by recommending near-term actions that may allow governments to mitigate flooding without invoking this trade-off.

Synthetic agents built atop LLM-based foundation models are gaining popularity as substitutes for human participants across research contexts, including user-testing, market-research, computational social science, surveys, and qualitative research. We are also witnessing an extension of synthetic agents into experimental implementations of policy consultation, jury deliberation, humanitarian diplomacy, and similar contexts where human participation and representation are central to the perceived legitimacy of the institutional processes. The value of participation extends beyond informational contributions and consensus generation; participation is a necessary, legitimizing condition for democratic political institutions and processes. Treating synthetic agents as human substitutes raises serious political, representational, and ethical concerns. Participatory Design's modes of engagement --- probing, priming, understanding, and generating --- offer helpful tools for engaging with representational questions of personhood. We apply the lens to three case studies of synthetic agents substituting for personhood at varying representational scales: local policy, enterprise jury deliberation, and global diplomacy. We argue that legitimacy and personhood are integral and mutually constitutive while identifying the ethical, representational, and methodological risks of using synthetic agents in representational processes. We conclude by proposing soft and hard boundaries for designing oversight on LLMs and synthetic agents in representational processes.

Platforms summarize providers' past achievements into credentials that buyers use to judge quality. Generative AI can now produce much of the work those achievements certify, raising fears that those quality signals are worthless. We audit how well such credentials stay informative in Kaggle's 2010-2026 archive, where medals are won on predictions scored against withheld answers and two evaluation formats ran side by side. Across 444,698 participations, a medal's power to predict performance sits almost entirely in its first year, in both formats and eras. Fresh medals kept most of their value through the AI transition. About half of the collapse in the informativeness of one format's medals is institutional: the platform had been phasing out that format before AI, and its medal stock aged out on schedule. Old medals look more informative only in isolation. The platform's official lifetime-tier display discards up to a sixth of the medals' predictive power. A recency-weighted index fit before the AI era explains AI-era performance about 13% better than the tiers and selects entrants who perform better on average, though the tiers still identify extreme top performers better. Displaying a recent-performance summary alongside the lifetime tiers would recover the discarded information for buyers.

Job Safety Analysis (JSA) and pre-task planning can benefit from prior incident records, yet historical accident data is often stored as unstructured narratives that are difficult to consult at the point of planning. A novel framework centered on large language models (LLMs) for highway construction safety reporting and planning is proposed as a foundation for future agentic applications, prioritizing deterministic, local inferencing. The first aim is to enable classification and quality scoring of incident narratives for existing and future reporting purposes. The second is to evaluate retrieval of relevant historical accidents, related imagery, and trusted industry documents for incorporation into daily safety plans. Neural probes were trained to classify incidents along four multiclass and two binary Occupational Injury and Illness Classification System (OIICS) fields and to derive an overall quality score, evaluated on a test set of over 15,000 narratives and a held-out set of 100 author-labeled records, benchmarked against a majority-vote LLM ensemble. The retrieval of historical accidents, reference imagery, and industry documents was benchmarked across embedding models using standard information retrieval metrics. OIICS classification reached 75% held-out accuracy, though the two binary flags were degenerate. The quality score, while meaningful on one database, was distorted on out-of-distribution fatalities in the held-out dataset. Accident retrieval recovered relevant incidents far above chance, performing best on lexically distinct construction activities. On document question answering, an open-weight decoder embedding model surpassed proprietary models. Overall, this work provides a new framework rooted in local inferencing and text embedding models for future agentic applications, with emphasis on bridging external data to JSA reports.

The rapid development and widespread application of artificial intelligence (AI) have sparked intense discussions on how to deploy responsible AI systems in a manner aligned with human values and ethical standards. Compared to fields like healthcare, energy, or finance, the application of AI in groundwater is relatively limited, and research on responsible AI is even more scarce. Taking the middle reaches of the Heihe River Basin as the study area, this paper proposes six Responsible AI principles: transparency, technical robustness, privacy governance, fairness, accountability, and sustainability. LSTM and Transformer time-series models are developed using multi-source hydrometeorological data, and validated via post-hoc interpretability, Monte Carlo simulation, and scenario analysis. The results show that Transformer outperforms LSTM in accuracy, robustness, and interpretability, demonstrating the operability and practical value of Responsible AI principles in groundwater prediction to support sustainable water management under climate change and human activities.

Generative AI systems are increasingly producing real-world artifacts, however their efficacy and validity are often evaluated via context-free LLM-scoring. These judges can be miscalibrated by irrelevant in-context reference examples, creating false confidence and allowing low-quality or harmful outputs to pass evaluation. We study this failure mode as context-induced miscalibration and introduce DA-RAC, a distance-aware reference-anchored calibration method for LLM judges. DA-RAC retrieves semantically and structurally similar labeled anchors for each judgement scenario, weights them by distance, and exposes neighborhood difficulty as a calibration and triage signal. On multi-run LLM-judge evaluation benchmarks, it improves calibration and reduces false-pass risk relative to zero-shot, chain-of-thought evaluation, and static-anchor baselines. Mechanistic analysis shows that judge scores vary systematically with anchor distance, while static references can induce misleading decision boundaries. Thus LLM-judgement requires not only better models, but also calibrated, auditable reference selection, especially when automated evaluation is used to support high-impact AI generated artifacts. Judgments should be grounded in relevant, inspectable, and contestable interpretive artifacts.

Pareto Conditioned Networks learn multiple multi-objective reinforcement learning behaviours by conditioning a single policy on a desired return command. However, the local mapping from command and state to action remains opaque. We propose command-space counterfactual explanations for PCNs: given a fixed state, original command, and foil action, we search, in a black-box setting, for a minimally changed desired-return command under which the same trained policy would choose the foil. Our contributions are threefold. First, we formulate PCN explanations as return-command interventions, using a return-only PCN variant that avoids the added ambiguity of horizon-conditioning. Second, we adapt adversarial machine learning methods to reinforcement-learning explanations. Third, we introduce a boundary-seeded directional search that improves over purely local optimization in the command-action landscape, resulting in our proposed approach CF-ZOO. The resulting explanations are actionable and intuitively expressed in the user's own preferences: "If your trade-off had shifted slightly towards X, the agent would have chosen Y."

The rapid adoption of large language models has enabled the development of clinical multi-agent systems (MAS) capable of integrating multimodal patient data and supporting increasingly complex clinical decision-making. However, the deployment of these systems in real-world healthcare settings raises critical ethical concerns related to safety, fairness, accountability, transparency, and patient trust. While numerous organizations, including the World Health Organization, the National Academy of Medicine, and the FUTURE-AI consortium, have proposed ethical frameworks and governance principles for healthcare AI, these efforts remain largely conceptual. To address this challenge, we present ETHOS (Ethics and Trust through Hierarchical Oversight System), a modular ethics framework designed as a governance meta-agent that can be integrated with any existing multi-agent system without requiring changes to its underlying architecture. ETHOS translates stakeholder-informed ethical requirements into executable runtime oversight through a layered governance approach consisting of deterministic checks, contextual reviews, and a final ethics critic. These components continuously evaluate intermediate reasoning steps and final outputs, enabling the system to identify ethical risks, request revisions, or suppress responses that fail predefined safety and trustworthiness criteria. We demonstrate ETHOS within a hepatology clinical decision-support MAS. Results show that ETHOS improves decision reliability by detecting incomplete, inconsistent, or out-of-scope evidence and appropriately increasing abstention when safe recommendations cannot be supported. By embedding ethical governance directly into system operation, ETHOS provides a practical and auditable mechanism for transforming high-level AI ethics principles into deployable safeguards.