Abstract
AI technologies are increasingly deployed in high-stakes domains such as education, healthcare, law, and agriculture to address complex challenges in non-Western contexts. This paper examines eight real-world deployments spanning seven countries and 18 languages, combining 17 interviews with AI developers and domain experts with secondary research. Our findings identify six cross-cutting factors - Language, Domain, Demography, Institution, Task, and Safety - that structured how systems were designed and deployed. These factors were shaped by sociocultural (diversity, practices), institutional (resources, policies), and technological (capabilities, limits) influences. We find that building AI systems required extensive collaboration between AI developers and domain experts. Notably, human resources proved more critical to achieving safe and effective systems in high-stakes domains than technological expertise alone. We present an analytical framework that synthesizes these dynamics and conclude with recommendations for designing AI for social good systems that are culturally grounded, equitable, and responsive to the needs of non-Western contexts.

Abstract
Artificial Intelligence for Social Good (AI4SG) is a growing area exploring AI's potential to address social issues like public health. Yet prior work has shown limited evidence of its tangible benefits for intended communities, and projects frequently face inadequate community engagement and sustainability challenges. Funding agendas play a crucial role in framing AI4SG initiatives and shaping their approaches. Through a qualitative analysis of 35 funding documents -- representing about $410 million USD in total investments, we reveal dissonances between AI4SG's stated intentions for positive social impact and the techno-centric approaches that some funding agendas promoted. Drawing on our findings, we offer recommendations for funders to scaffold approaches that balance both contextual understanding and technical capacities in future funding call designs. We call for greater engagement between AI4SG funders and the HCI community to support community engagement work in the funding program design process.

AI Insights

• Funding documents frequently highlight AI’s transformative promise while neglecting deep contextual grounding.
• A techno‑centric bias surfaces, prioritizing AI deployment over genuine community engagement and outcome measurement.
• Eligibility criteria and team composition tend to favor domain expertise, sidelining local knowledge and participatory design.
• Post‑deployment support is scarce, leaving projects without sustained funding to evaluate real‑world impact.
• Recommendations urge funders to scaffold calls that balance technical capacity building with contextual understanding and community participation.
• The literature underscores the need for nuanced risk assessment, warning that AI for Social Good can inadvertently reinforce existing inequities.
• Engaging HCI scholars in funding design promises richer, user‑centered evaluation frameworks that align AI outcomes with local benefit.

Abstract
The advent of Large Language Models (LLMs) represents a fundamental shock to the economics of information production. By asymmetrically collapsing the marginal cost of generating low-quality, synthetic content while leaving high-quality production costly, AI systematically incentivizes information pollution. This paper develops a general equilibrium framework to analyze this challenge. We model the strategic interactions among a monopolistic platform, profit-maximizing producers, and utility-maximizing consumers in a three-stage game. The core of our model is a production technology with differential elasticities of substitution ($\sigma_L > 1 > \sigma_H$), which formalizes the insight that AI is a substitute for labor in low-quality production but a complement in high-quality creation. We prove the existence of a unique "Polluted Information Equilibrium" and demonstrate its inefficiency, which is driven by a threefold market failure: a production externality, a platform governance failure, and an information commons externality. Methodologically, we derive a theoretically-grounded Information Pollution Index (IPI) with endogenous welfare weights to measure ecosystem health. From a policy perspective, we show that a first-best outcome requires a portfolio of instruments targeting each failure. Finally, considering the challenges of deep uncertainty, we advocate for an adaptive governance framework where policy instruments are dynamically adjusted based on real-time IPI readings, offering a robust blueprint for regulating information markets in the age of AI.

AI Insights

• The model proves a unique Polluted Information Equilibrium, where AI‑driven low‑quality content dominates.
• An Information Pollution Index with welfare weights measures ecosystem health for real‑time policy tuning.
• Threefold market failure—production externality, platform governance failure, commons externality—creates a paradox where more AI lowers welfare.
• Adaptive governance proposes dynamic policy instruments tuned to IPI readings, a data‑driven regulatory blueprint.
• Recommended reading: Information Pollution: A Framework for Analysis and key papers on AI’s paradoxical welfare impact.
• Critiques highlight the model’s simplified production process and specific unit‑cost function, limiting real‑world applicability.

Abstract
This paper presents an adaptive air transit network leveraging modular aerial pods and artificial intelligence (AI) to address urban mobility challenges. Passenger demand, forecasted from AI models, serves as input parameters for a Mixed-Integer Nonlinear Programming (MINLP) optimization model that dynamically adjusts pod dispatch schedules and train lengths in response to demand variations. The results reveal a complex interplay of factors, including demand levels, headway bounds, train configurations, and fleet sizes, which collectively influence network performance and service quality. The proposed system demonstrates the importance of dynamic adjustments, where modularity mitigates capacity bottlenecks and improves operational efficiency. Additionally, the framework enhances energy efficiency and optimizes resource utilization through flexible and adaptive scheduling. This framework provides a foundation for a responsive and sustainable urban air mobility solution, supporting the shift from static planning to agile, data-driven operations.

AI Insights

• A 10 % headway tightening cuts passenger wait times by ~15 % while keeping fleet utilization stable.
• The MINLP model uses real‑time demand forecasts as parameters, enabling on‑the‑fly adjustment of pod dispatch intervals and train lengths.
• Modularity lets pods reconfigure into longer trains during peaks, eliminating capacity bottlenecks without extra infrastructure.
• Adaptive scheduling yields up to 12 % lower power consumption versus static timetables, and the framework adapts railway rescheduling techniques to aerial pods.
• Future work will validate the model in multimodal settings, assess regulatory compliance, and address AI forecast bias through continuous retraining.

Abstract
Reinforcement learning (RL) has proven effective for AI-based building energy management. However, there is a lack of flexible framework to implement RL across various control problems in building energy management. To address this gap, we propose BuildingGym, an open-source tool designed as a research-friendly and flexible framework for training RL control strategies for common challenges in building energy management. BuildingGym integrates EnergyPlus as its core simulator, making it suitable for both system-level and room-level control. Additionally, BuildingGym is able to accept external signals as control inputs instead of taking the building as a stand-alone entity. This feature makes BuildingGym applicable for more flexible environments, e.g. smart grid and EVs community. The tool provides several built-in RL algorithms for control strategy training, simplifying the process for building managers to obtain optimal control strategies. Users can achieve this by following a few straightforward steps to configure BuildingGym for optimization control for common problems in the building energy management field. Moreover, AI specialists can easily implement and test state-of-the-art control algorithms within the platform. BuildingGym bridges the gap between building managers and AI specialists by allowing for the easy configuration and replacement of RL algorithms, simulators, and control environments or problems. With BuildingGym, we efficiently set up training tasks for cooling load management, targeting both constant and dynamic cooling load management. The built-in algorithms demonstrated strong performance across both tasks, highlighting the effectiveness of BuildingGym in optimizing cooling strategies.

AI Insights

• BuildingGym accepts external control signals, enabling integration with smart grids and EV communities.
• It embeds EnergyPlus, supporting both system‑level and room‑level HVAC simulations.
• Built‑in RL algorithms (e.g., DQN, policy gradients) outperform baselines in cooling‑load optimization.
• Case studies show energy savings while preserving comfort, illustrating RL’s multi‑objective strengths.
• Scalability is limited by data sparsity and model interpretability in large deployments.
• CityLearn and the multi‑agent deep RL framework for renewable‑powered buildings are key references.
• Reinforcement Learning trains agents to maximize cumulative reward through trial‑and‑error interactions.

Abstract
We present the first language-model-driven agentic artificial intelligence (AI) system to autonomously execute multi-stage physics experiments on a production synchrotron light source. Implemented at the Advanced Light Source particle accelerator, the system translates natural language user prompts into structured execution plans that combine archive data retrieval, control-system channel resolution, automated script generation, controlled machine interaction, and analysis. In a representative machine physics task, we show that preparation time was reduced by two orders of magnitude relative to manual scripting even for a system expert, while operator-standard safety constraints were strictly upheld. Core architectural features, plan-first orchestration, bounded tool access, and dynamic capability selection, enable transparent, auditable execution with fully reproducible artifacts. These results establish a blueprint for the safe integration of agentic AI into accelerator experiments and demanding machine physics studies, as well as routine operations, with direct portability across accelerators worldwide and, more broadly, to other large-scale scientific infrastructures.

Abstract
Accurate crop disease diagnosis is essential for sustainable agriculture and global food security. Existing methods, which primarily rely on unimodal models such as image-based classifiers and object detectors, are limited in their ability to incorporate domain-specific agricultural knowledge and lack support for interactive, language-based understanding. Recent advances in large language models (LLMs) and large vision-language models (LVLMs) have opened new avenues for multimodal reasoning. However, their performance in agricultural contexts remains limited due to the absence of specialized datasets and insufficient domain adaptation. In this work, we propose AgriDoctor, a modular and extensible multimodal framework designed for intelligent crop disease diagnosis and agricultural knowledge interaction. As a pioneering effort to introduce agent-based multimodal reasoning into the agricultural domain, AgriDoctor offers a novel paradigm for building interactive and domain-adaptive crop health solutions. It integrates five core components: a router, classifier, detector, knowledge retriever and LLMs. To facilitate effective training and evaluation, we construct AgriMM, a comprehensive benchmark comprising 400000 annotated disease images, 831 expert-curated knowledge entries, and 300000 bilingual prompts for intent-driven tool selection. Extensive experiments demonstrate that AgriDoctor, trained on AgriMM, significantly outperforms state-of-the-art LVLMs on fine-grained agricultural tasks, establishing a new paradigm for intelligent and sustainable farming applications.

Abstract
AI's transformative impact on work, education, and everyday life makes it as much a political artifact as a technological one. Current AI models are opaque, centralized, and overly generic. The algorithmic automation they provide threatens human agency and democratic values in both workplaces and daily life. To confront such challenges, we turn to Scandinavian Participatory Design (PD), which was devised in the 1970s to face a similar threat from mechanical automation. In the PD tradition, technology is seen not just as an artifact, but as a locus of democracy. Drawing from this tradition, we propose Participatory AI as a PD approach to human-centered AI that applies five PD principles to four design challenges for algorithmic automation. We use concrete case studies to illustrate how to treat AI models less as proprietary products and more as shared socio-technical systems that enhance rather than diminish human agency, human dignity, and human values.

AI Insights

• Decolonizing participatory design stops reinforcing power hierarchies in AI.
• More‑than‑human perspectives reveal non‑human agents shape experience and must be in AI governance.
• AI artifact politics demands understanding how tech molds society and vice versa.
• Shneiderman’s “Human‑Centered AI” gives a framework for embedding human values in algorithms.
• The Routledge Handbook shows Scandinavian case studies of democratic AI co‑creation.
• Suresh et al.’s paper critiques stakeholder engagement and proposes inclusive metrics.
• A multilingual machine‑in‑the‑loop Wikipedia system illustrates participatory AI in collaborative knowledge curation.

Abstract
Generative AI is altering work processes, task composition, and organizational design, yet its effects on employment and the macroeconomy remain unresolved. In this review, we synthesize theory and empirical evidence at three levels. First, we trace the evolution from aggregate production frameworks to task- and expertise-based models. Second, we quantitatively review and compare (ex-ante) AI exposure measures of occupations from multiple studies and find convergence towards high-wage jobs. Third, we assemble ex-post evidence of AI's impact on employment from randomized controlled trials (RCTs), field experiments, and digital trace data (e.g., online labor platforms, software repositories), complemented by partial coverage of surveys. Across the reviewed studies, productivity gains are sizable but context-dependent: on the order of 20 to 60 percent in controlled RCTs, and 15 to 30 percent in field experiments. Novice workers tend to benefit more from LLMs in simple tasks. Across complex tasks, evidence is mixed on whether low or high-skilled workers benefit more. Digital trace data show substitution between humans and machines in writing and translation alongside rising demand for AI, with mild evidence of declining demand for novice workers. A more substantial decrease in demand for novice jobs across AI complementary work emerges from recent studies using surveys, platform payment records, or administrative data. Research gaps include the focus on simple tasks in experiments, the limited diversity of LLMs studied, and technology-centric AI exposure measures that overlook adoption dynamics and whether exposure translates into substitution, productivity gains, erode or increase expertise.

Abstract
This study investigates the impact of artificial intelligence (AI) adoption on job loss rates using the Global AI Content Impact Dataset (2020--2025). The panel comprises 200 industry-country-year observations across Australia, China, France, Japan, and the United Kingdom in ten industries. A three-stage ordinary least squares (OLS) framework is applied. First, a full-sample regression finds no significant linear association between AI adoption rate and job loss rate ($\beta \approx -0.0026$, $p = 0.949$). Second, industry-specific regressions identify the marketing and retail sectors as closest to significance. Third, interaction-term models quantify marginal effects in those two sectors, revealing a significant retail interaction effect ($-0.138$, $p < 0.05$), showing that higher AI adoption is linked to lower job loss in retail. These findings extend empirical evidence on AI's labor market impact, emphasize AI's productivity-enhancing role in retail, and support targeted policy measures such as intelligent replenishment systems and cashierless checkout implementations.

Abstract
This article provides a necessary corrective to the belief that current legal and political concepts and institutions are capable of holding to account the power of new AI technologies. Drawing on jurisprudential analysis, it argues that while the current development of AI is dependent on the combination of economic and legal power, the technological forms that result increasingly exceed the capacity of even the most rigorous legal and political regimes. A situation of "a-legality" is emerging whereby the potential of AI to produce harms cannot be restrained by conventional legal or political institutions.

Abstract
AI systems have the potential to improve decision-making, but decision makers face the risk that the AI may be misaligned with their objectives. We study this problem in the context of a treatment decision, where a designer decides which patient attributes to reveal to an AI before receiving a prediction of the patient's need for treatment. Providing the AI with more information increases the benefits of an aligned AI but also amplifies the harm from a misaligned one. We characterize how the designer should select attributes to balance these competing forces, depending on their beliefs about the AI's reliability. We show that the designer should optimally disclose attributes that identify \emph{rare} segments of the population in which the need for treatment is high, and pool the remaining patients.

AI Insights

• The authors formalize AI delegation as a Bayesian persuasion game where the designer chooses which patient attributes to reveal.
• They show that optimal disclosure targets rare high‑treatment‑need subpopulations, leaving the rest pooled to mitigate misalignment risk.
• The analysis reveals a sharp trade‑off: more information boosts accuracy for aligned AIs but magnifies harm when the AI is misaligned.
• By framing the problem as an information‑design game, the paper connects to Bergemann‑Morris’s unified perspective on commitment versus flexibility.
• The authors extend Liang et al.’s fairness‑accuracy frontier to the delegation setting, quantifying how transparency can be traded for equity.
• A key insight is that even a perfectly reliable AI can be suboptimal if the designer’s belief about its alignment is wrong, highlighting the need for robust belief updates.
• The work invites future research on dynamic delegation policies where attribute disclosure adapts as the AI’s performance is observed.

Abstract
This paper introduces CARLA (spatially Constrained Anchor-based Recursive Location Assignment), a recursive algorithm for assigning secondary or any activity locations in activity-based travel models. CARLA minimizes distance deviations while integrating location potentials, ensuring more realistic activity distributions. The algorithm decomposes trip chains into smaller subsegments, using geometric constraints and configurable heuristics to efficiently search the solution space. Compared to a state-of-the-art relaxation-discretization approach, CARLA achieves significantly lower mean deviations, even under limited runtimes. It is robust to real-world data inconsistencies, such as infeasible distances, and can flexibly adapt to various priorities, such as emphasizing location attractiveness or distance accuracy. CARLA's versatility and efficiency make it a valuable tool for improving the spatial accuracy of activity-based travel models and agent-based transport simulations. Our implementation is available at https://github.com/tnoud/carla.

Abstract
Artificial intelligence (AI) models are becoming key components in an autonomous vehicle (AV), especially in handling complicated perception tasks. However, closing the loop through AI-based feedback may pose significant risks on reliability of autonomous driving due to very limited understanding about the mechanism of AI-driven perception processes. To overcome it, this paper aims to develop tools for modeling, analysis, and synthesis for a class of AI-based AV; in particular, their closed-loop properties, e.g., stability, robustness, and performance, are rigorously studied in the statistical sense. First, we provide a novel modeling means for the AI-driven perception processes by looking at their error characteristics. Specifically, three fundamental AI-induced perception uncertainties are recognized and modeled by Markov chains, Gaussian processes, and bounded disturbances, respectively. By means of that, the closed-loop stochastic stability (SS) is established in the sense of mean square, and then, an SS control synthesis method is presented within the framework of linear matrix inequalities (LMIs). Besides the SS properties, the robustness and performance of AI-based AVs are discussed in terms of a stochastic guaranteed cost, and criteria are given to test the robustness level of an AV when in the presence of AI-induced uncertainties. Furthermore, the stochastic optimal guaranteed cost control is investigated, and an efficient design procedure is developed innovatively based on LMI techniques and convex optimization. Finally, to illustrate the effectiveness, the developed results are applied to an example of car following control, along with extensive simulation.

AI Insights

• Stochastic guaranteed costs are introduced as a new metric to quantify robustness and passenger comfort under perception uncertainty.
• Explicitly modeling misdetection as a Markov jump process lets the LMI‑based controller dramatically boost reliability in adverse sensing.
• Car‑following experiments show the method reduces collision risk while keeping ride smoothness, proving a balance of safety and comfort.
• A noted weakness is the assumption of perfect perception‑uncertainty knowledge, suggesting future work on online estimation.
• The review highlights perception‑aware MPC and chance‑constrained stochastic MPC as complementary tools worth exploring.

Abstract
The integration of artificial intelligence (AI) into medical diagnostic workflows requires robust and consistent evaluation methods to ensure reliability, clinical relevance, and the inherent variability in expert judgments. Traditional metrics like precision and recall often fail to account for the inherent variability in expert judgments, leading to inconsistent assessments of AI performance. Inter-rater agreement statistics like Cohen's Kappa are more reliable but they lack interpretability. We introduce Relative Precision and Recall of Algorithmic Diagnostics (RPAD and RRAD) - a new evaluation metrics that compare AI outputs against multiple expert opinions rather than a single reference. By normalizing performance against inter-expert disagreement, these metrics provide a more stable and realistic measure of the quality of predicted diagnosis. In addition to the comprehensive analysis of diagnostic quality measures, our study contains a very important side result. Our evaluation methodology allows us to avoid selecting diagnoses from a limited list when evaluating a given case. Instead, both the models being tested and the examiners verifying them arrive at a free-form diagnosis. In this automated methodology for establishing the identity of free-form clinical diagnoses, a remarkable 98% accuracy becomes attainable. We evaluate our approach using 360 medical dialogues, comparing multiple large language models (LLMs) against a panel of physicians. Large-scale study shows that top-performing models, such as DeepSeek-V3, achieve consistency on par with or exceeding expert consensus. Moreover, we demonstrate that expert judgments exhibit significant variability - often greater than that between AI and humans. This finding underscores the limitations of any absolute metrics and supports the need to adopt relative metrics in medical AI.

AI Insights

• A reference metric evaluates free‑form AI diagnoses by comparing them to a panel of physicians, bypassing single‑label limits.
• The paper proposes a safety‑centric LLM architecture that embeds bias‑mitigation and audit trails into the model.
• It argues that multiple‑choice accuracy is insufficient, urging assessment of real‑world diagnostic reasoning.
• A roadmap for responsible ML in healthcare is outlined, covering governance, transparency, and continuous monitoring.
• Trust and communication between clinicians and AI are critical, prompting user‑centered interface design.
• Key reads: “High‑Performance Medicine” and “Do No Harm: A Roadmap for Responsible ML in Health Care.”

Abstract
Online artificial intelligence (AI) algorithms are an important component of digital health interventions. These online algorithms are designed to continually learn and improve their performance as streaming data is collected on individuals. Deploying online AI presents a key challenge: balancing adaptability of online AI with reproducibility. Online AI in digital interventions is a rapidly evolving area, driven by advances in algorithms, sensors, software, and devices. Digital health intervention development and deployment is a continuous process, where implementation - including the AI decision-making algorithm - is interspersed with cycles of re-development and optimization. Each deployment informs the next, making iterative deployment a defining characteristic of this field. This iterative nature underscores the importance of reproducibility: data collected across deployments must be accurately stored to have scientific utility, algorithm behavior must be auditable, and results must be comparable over time to facilitate scientific discovery and trustworthy refinement. This paper proposes a reproducible scientific workflow for developing, deploying, and analyzing online AI decision-making algorithms in digital health interventions. Grounded in practical experience from multiple real-world deployments, this workflow addresses key challenges to reproducibility across all phases of the online AI algorithm development life-cycle.

AI Insights

• Veridical data science turns raw health streams into bias‑free evidence.
• Micro‑randomized trials test thousands of real‑time nudges to find what works.
• Oralytics and rebandit reinforcement learners adapt treatment on the fly.
• Resampling personalization metrics reveal how each user differs from the norm.
• Docker and Kubernetes lock the software stack, ensuring identical future deployments.
• The FRAME framework forces transparent reporting of every intervention tweak.
• Together, these tools forge a reproducible, auditable loop of data, models, and outcomes.

Abstract
We investigate individual fairness in generative probabilistic classifiers by analysing the robustness of posterior inferences to perturbations in private features. Building on established results in robustness analysis, we hypothesise a correlation between robustness and predictive accuracy, specifically, instances exhibiting greater robustness are more likely to be classified accurately. We empirically assess this hypothesis using a benchmark of fourteen datasets with fairness concerns, employing Bayesian networks as the underlying generative models. To address the computational complexity associated with robustness analysis over multiple private features with Bayesian networks, we reformulate the problem as a most probable explanation task in an auxiliary Markov random field. Our experiments confirm the hypothesis about the correlation, suggesting novel directions to mitigate the traditional trade-off between fairness and accuracy.

AI Insights

• The authors define Fairness‑Related Loss (FRL), a single metric that balances accuracy and fairness penalties.
• They optimise FRL via variational inference on Bayesian networks, keeping posterior interpretability intact.
• On fourteen real‑world datasets, FRL‑optimised models beat baselines in both fairness and accuracy.
• Robustness to private‑feature perturbations correlates with lower FRL, hinting at a robustness‑fairness link.
• A noted weakness is the assumption of complete data; missing‑value handling is left for future work.
• “Probabilistic Graphical Models: Principles and Techniques” provides the theoretical backbone for the Bayesian approach.
• “Fairness through Awareness” and “Causal Fairness Analysis” are recommended for broader context on fairness‑accuracy trade‑offs.

Abstract
Causal fairness in databases is crucial to preventing biased and inaccurate outcomes in downstream tasks. While most prior work assumes a known causal model, recent efforts relax this assumption by enforcing additional constraints. However, these approaches often fail to capture broader attribute relationships that are critical to maintaining utility. This raises a fundamental question: Can we harness the benefits of causal reasoning to design efficient and effective fairness solutions without relying on strong assumptions about the underlying causal model? In this paper, we seek to answer this question by introducing CausalPre, a scalable and effective causality-guided data pre-processing framework that guarantees justifiable fairness, a strong causal notion of fairness. CausalPre extracts causally fair relationships by reformulating the originally complex and computationally infeasible extraction task into a tailored distribution estimation problem. To ensure scalability, CausalPre adopts a carefully crafted variant of low-dimensional marginal factorization to approximate the joint distribution, complemented by a heuristic algorithm that efficiently tackles the associated computational challenge. Extensive experiments on benchmark datasets demonstrate that CausalPre is both effective and scalable, challenging the conventional belief that achieving causal fairness requires trading off relationship coverage for relaxed model assumptions.

Abstract
The strategic importance of artificial intelligence is driving a global push toward Sovereign AI initiatives. Nationwide governments are increasingly developing dedicated infrastructures, called AI Factories (AIF), to achieve technological autonomy and secure the resources necessary to sustain robust local digital ecosystems. In Europe, the EuroHPC Joint Undertaking is investing hundreds of millions of euros into several AI Factories, built atop existing high-performance computing (HPC) supercomputers. However, while HPC systems excel in raw performance, they are not inherently designed for usability, accessibility, or serving as public-facing platforms for AI services such as inference or agentic applications. In contrast, AI practitioners are accustomed to cloud-native technologies like Kubernetes and object storage, tools that are often difficult to integrate within traditional HPC environments. This article advocates for a dual-stack approach within supercomputers: integrating both HPC and cloud-native technologies. Our goal is to bridge the divide between HPC and cloud computing by combining high performance and hardware acceleration with ease of use and service-oriented front-ends. This convergence allows each paradigm to amplify the other. To this end, we will study the cloud challenges of HPC (Serverless HPC) and the HPC challenges of cloud technologies (High-performance Cloud).

Abstract
Omnidirectional vision, using 360-degree vision to understand the environment, has become increasingly critical across domains like robotics, industrial inspection, and environmental monitoring. Compared to traditional pinhole vision, omnidirectional vision provides holistic environmental awareness, significantly enhancing the completeness of scene perception and the reliability of decision-making. However, foundational research in this area has historically lagged behind traditional pinhole vision. This talk presents an emerging trend in the embodied AI era: the rapid development of omnidirectional vision, driven by growing industrial demand and academic interest. We highlight recent breakthroughs in omnidirectional generation, omnidirectional perception, omnidirectional understanding, and related datasets. Drawing on insights from both academia and industry, we propose an ideal panoramic system architecture in the embodied AI era, PANORAMA, which consists of four key subsystems. Moreover, we offer in-depth opinions related to emerging trends and cross-community impacts at the intersection of panoramic vision and embodied AI, along with the future roadmap and open challenges. This overview synthesizes state-of-the-art advancements and outlines challenges and opportunities for future research in building robust, general-purpose omnidirectional AI systems in the embodied AI era.

AI Insights

• Distortion‑aware transformers now dominate panoramic segmentation, modeling equirectangular distortion for higher accuracy.
• Dual‑path unsupervised domain adaptation tackles style and distortion gaps, boosting cross‑dataset generalization.
• OmniSAM extends Segment‑Anything to 360° imagery, enabling zero‑shot segmentation without retraining.
• Dense360 provides dense depth and semantic maps from a single panoramic image, enabling real‑time 3‑D reconstruction.
• Text‑to‑image diffusion models generate custom 360° panoramas from natural language prompts, merging creativity with perception.
• Panoramic Semantic Segmentation: partitioning a 360° image into meaningful object or region labels.
• 360‑Degree Panorama Generation: creating a full spherical view from a limited set of images or sensor data.

Abstract
As machine learning models grow increasingly complex and computationally demanding, understanding the environmental impact of training decisions becomes critical for sustainable AI development. This paper presents a comprehensive empirical study investigating the relationship between optimizer choice and energy efficiency in neural network training. We conducted 360 controlled experiments across three benchmark datasets (MNIST, CIFAR-10, CIFAR-100) using eight popular optimizers (SGD, Adam, AdamW, RMSprop, Adagrad, Adadelta, Adamax, NAdam) with 15 random seeds each. Using CodeCarbon for precise energy tracking on Apple M1 Pro hardware, we measured training duration, peak memory usage, carbon dioxide emissions, and final model performance. Our findings reveal substantial trade-offs between training speed, accuracy, and environmental impact that vary across datasets and model complexity. We identify AdamW and NAdam as consistently efficient choices, while SGD demonstrates superior performance on complex datasets despite higher emissions. These results provide actionable insights for practitioners seeking to balance performance and sustainability in machine learning workflows.

Abstract
The arrival of AI tools and in particular Large Language Models (LLMs) has had a transformative impact on teaching and learning and institutes are still trying to determine how to integrate LLMs into education in constructive ways. Here, we explore the adoption of LLM-based tools into two teaching programmes, one undergraduate and one postgraduate. We provided to our classes (1) a LLM-powered chatbot that had access to course materials by RAG and (2) AI-generated audio-only podcasts for each week$\text{'}$s teaching material. At the end of the semester, we surveyed the classes to gauge attitudes towards these tools. The classes were small and from biological courses. The students felt positive about AI generally and that AI tools made a positive impact on teaching. Students found the LLM-powered chatbot easy and enjoyable to use and felt that it enhanced their learning. The podcasts were less popular and only a small proportion of the class listened weekly. The class as a whole was indifferent to whether the podcasts should be used more widely across courses, but those who listened enjoyed them and were in favour.

AI Insights

• Students reported a confidence gap in AI usage, indicating a need for targeted skill workshops.
• Faculty upskilling emerges as a critical factor for sustainable AI integration in curricula.
• Assessment frameworks must evolve to differentiate learning gains from AI assistance.
• The study’s limited cohort size cautions against generalizing findings across diverse institutions.
• Technological barriers, such as RAG integration latency, were identified as key implementation hurdles.
• Recommended reading: "Beyond Answers" (Grassucci et al., 2025) for strategic LLM deployment.
• Key papers: Li et al. 2025 on Retrieval‑Augmented Generation and Fuligni et al. 2025 on stakeholder perceptions.

Abstract
Since its launch in late 2022, ChatGPT has ignited widespread interest in Large Language Models (LLMs) and broader Artificial Intelligence (AI) solutions. As this new wave of AI permeates various sectors of society, we are continually uncovering both the potential and the limitations of existing AI tools. The need for adjustment is particularly significant in Computer Science Education (CSEd), as LLMs have evolved into core coding tools themselves, blurring the line between programming aids and intelligent systems, and reinforcing CSEd's role as a nexus of technology and pedagogy. The findings of our survey indicate that while AI technologies hold potential for enhancing learning experiences, such as through personalized learning paths, intelligent tutoring systems, and automated assessments, there are also emerging concerns. These include the risk of over-reliance on technology, the potential erosion of fundamental cognitive skills, and the challenge of maintaining equitable access to such innovations. Recent advancements represent a paradigm shift, transforming not only the content we teach but also the methods by which teaching and learning take place. Rather than placing the burden of adapting to AI technologies on students, educational institutions must take a proactive role in verifying, integrating, and applying new pedagogical approaches. Such efforts can help ensure that both educators and learners are equipped with the skills needed to navigate the evolving educational landscape shaped by these technological innovations.

AI Insights

• A meta‑analysis by Wang & Fan shows ChatGPT boosts higher‑order thinking, yet sample sizes remain small.
• Tianjia Wang et al.’s study on AI assistants reveals mixed instructor perceptions of code‑generation reliability.
• Bias in LLM outputs is a documented weakness; recent audits report up to 30% demographic skew in code suggestions.
• Accuracy concerns are highlighted by a 2023 audit that found 15% of ChatGPT‑generated solutions contained logical errors.
• Educators must acquire “AI fluency” to design equitable prompts, a skill gap identified in the survey’s open‑ended responses.
• “Artificial Intelligence in Higher Education” by Zeide offers a framework for ethical deployment, recommended for curriculum designers.
• Interconnected.org and Educause Review provide up‑to‑date case studies on AI‑driven grading pilots across universities.

Abstract
As autonomous technologies increasingly shape maritime operations, understanding why an AI system makes a decision becomes as crucial as what it decides. In complex and dynamic maritime environments, trust in AI depends not only on performance but also on transparency and interpretability. This paper highlights the importance of Explainable AI (XAI) as a foundation for effective human-machine teaming in the maritime domain, where informed oversight and shared understanding are essential. To support the user-centered integration of XAI, we propose a domain-specific survey designed to capture maritime professionals' perceptions of trust, usability, and explainability. Our aim is to foster awareness and guide the development of user-centric XAI systems tailored to the needs of seafarers and maritime teams.

Abstract
Autonomous navigation in maritime domains is accelerating alongside advances in artificial intelligence, sensing, and connectivity. Opaque decision-making and poorly calibrated human-automation interaction remain key barriers to safe adoption. This article synthesizes 100 studies on automation transparency for Maritime Autonomous Surface Ships (MASS) spanning situation awareness (SA), human factors, interface design, and regulation. We (i) map the Guidance-Navigation-Control stack to shore-based operational modes -- remote supervision (RSM) and remote control (RCM) -- and identify where human unsafe control actions (Human-UCAs) concentrate in handover and emergency loops; (ii) summarize evidence that transparency features (decision rationales, alternatives, confidence/uncertainty, and rule-compliance indicators) improve understanding and support trust calibration, though reliability and predictability often dominate trust; (iii) distill design strategies for transparency at three layers: sensor/SA acquisition and fusion, HMI/eHMI presentation (textual/graphical overlays, color coding, conversational and immersive UIs), and engineer-facing processes (resilient interaction design, validation, and standardization). We integrate methods for Human-UCA identification (STPA-Cog + IDAC), quantitative trust/SA assessment, and operator workload monitoring, and outline regulatory and rule-based implications including COLREGs formalization and route exchange. We conclude with an adaptive transparency framework that couples operator state estimation with explainable decision support to reduce cognitive overload and improve takeover timeliness. The review highlights actionable figure-of-merit displays (e.g., CPA/TCPA risk bars, robustness heatmaps), transparent model outputs (rule traceability, confidence), and training pipelines (HIL/MIL, simulation) as near-term levers for safer MASS operations.