Search Results (299)

We present a hybrid system that combines large language models (LLMs) with formal graph-analytic methods to generate and automatically repair interactive fiction (IF) stories written in the Twine/Twee format. We chronologically describe the practical challenges encountered when attempting to produce fully playable branching narratives using contemporary state-of-the-art LLMs, including missing passages, trap-like cycles without exits, dead-end passages, narrative discontinuities, incorrect use of Twine macro commands, and inconsistent handling of story variables. To address these limitations, we deliberately abandon all macro- and variable-based logic and instead encode story state directly within passage names through structured, token-based naming. We formalize symmetry and asymmetry in the resulting narrative graphs: symmetrical convergence occurs when multiple branches with compatible states merge into a common passage, whereas asymmetry reveals incorrect or logically inconsistent merging of divergent states (for example, entering a scene in which an item or companion is present via paths where they were never acquired or met). We propose algorithms to detect naming-based asymmetries, cycles, unreachable endings, and structurally defective branches, and we integrate these diagnostics into a repair loop that prompts the LLM to rewrite missing or inconsistent parts of the story. Experiments with several LLM backends indicate that this approach can yield structurally robust and locally coherent interactive stories by reducing state inconsistencies and structural defects. Beyond the specific case of Twine, we argue that symmetry/asymmetry analysis offers a powerful lens for evaluating and correcting AI-generated narrative graphs in general. Full article

Seconds count differently for people in danger. We present a real-time streaming pipeline for audio-based detection of hazardous life events affecting life and property. The system operates online rather than as a retrospective analysis tool. Its objective is to reduce the latency between the occurrence of a crime, conflict, or accident and the corresponding response by authorities. The key idea is to map reality as perceived by audio into a written story and question the text via a large language model. The method integrates streaming, zero-shot algorithms in an online decoding mode that convert sound into short, interpretable tokens, which are processed by a lightweight language model. CLAP text–audio prompting identifies agitation, panic, and distress cues, combined with conversational dynamics derived from speaker diarization. Lexical information is obtained through streaming automatic speech recognition, while general audio events are detected by a streaming version of Audio Spectrogram Transformer tagger. Prosodic features are incorporated using pitch- and energy-based rules derived from robust F0 tracking and periodicity measures. The system uses a large language model configured for online decoding and outputs binary (YES/NO) life-threatening risk decisions every two seconds, along with a brief justification and a final session-level verdict. The system emphasizes interpretability and accountability. We evaluate it on a subset of the X-Violence dataset, comprising only real-world videos. We release code, prompts, decision policies, evaluation splits, and example logs to enable the community to replicate, critique, and extend our blueprint. Full article

The rapid integration of large language models into higher education creates opportunities for mathematics instruction, but also raises the need for structured interaction strategies that support reflective learning rather than passive answer consumption. This study, conducted within the Erasmus+ MAESTRO-AI project, examines how selected AI prompt patterns can be implemented in concrete university mathematics activities and how students evaluate these AI-supported experiences. Two experimental modules were compared: complex numbers for first-semester Applied Mathematics students in Poland ($n=100$) and conditional probability for second-year Computer Science students in Romania ($n=213$). After completing AI-assisted learning activities with ChatGPT and/or Gemini, students completed a common evaluation questionnaire assessing engagement, perceived usefulness, and reflections on AI as a tutor. Group comparisons and experience-based analyses were performed using the Mann–Whitney test. Results indicate that students who reported regular prior use of AI tools evaluated AI-supported learning significantly more positively than those with occasional or no prior experience. They gave higher ratings across most questionnaire items as well as for the overall score. The findings suggest that prompt-pattern-based designs can support engaging AI-assisted mathematics activities. They also indicate that such designs can provide a structured learning experience, while introductory guidance may be important to ensure comparable benefits for less experienced students. Full article

With the deep integration of digital technology and the tourism industry, the transformation of the spatial form of smart tourism destinations and the research on their system structure have become the focus. This study adopts a mixed research approach, taking villages in the mountainous areas of southeastern China as examples, and collects empirical data through semi-structured interviews, participant observation and literature collection. This study draws on structuralist location theory to construct a four-dimensional spatial analysis model of natural environment, production economy, social norms and cultural values and incorporates a historical perspective to make up for the limitations of this theory in explaining regional dynamic changes caused by the lack of a time dimension. This study finds that digital tourism provides external resources such as the consumer market, tourism capital and information technology prompting the reconfiguration of the rural internal system. By absorbing external resources and upgrading traditional industries, rural areas have formed a more diversified, inclusive, and dynamically balanced spatial form. Furthermore, phenomena such as villagers’ relocation, e-commerce employment and local tea-growing knowledge indicate that certain predicaments still exist in the construction of digital tourism. This research can provide practical references for the development and spatial optimization of rural digital tourism. Full article

This exploratory study examines how discursive design—using provocative, speculative artifacts to spark reflection and discussion—might expand public health experts’ problematization of approaches to tailoring and targeting interventions. Cultural tailoring and targeting (CTT) refers to adapting interventions for specific sociocultural populations. Because LGBTQ+ communities experience disproportionately high rates of tobacco use, this study applies discursive intervention concepts within this context to explore how they might help experts critically engage with CTT strategies for reaching LGBTQ+ populations more effectively. To investigate this, two pairs of discursive intervention concepts were designed and presented to three focus groups of public health experts. Each pair juxtaposed a conventional intervention approach with a more provocative, unfamiliar one—for example, deepfake-driven behavior disruption. The goal was to document the type of conversation discursive design could stimulate around CTT considerations and generate insights relevant to the value of design methodologies to foster new ways to problematize public health matters. Findings indicate that the concepts prompted critical conversations about CTT, although the depth and focus of engagement varied. Those with greater expertise in LGBTQ+ issues engaged more with CTT mechanisms and implications, while others focused on implementation and feasibility concerns—essential to intervention development but outside the study’s focus. These patterns highlight who should be included in such efforts and how they should be engaged from a facilitation perspective, raising important considerations for methodological refinements and future research. Overall, this initial exploration aims to uncover the potential of discursive design to deepen understanding of CTT interventions and inform more responsive, innovative approaches to addressing tobacco use among priority populations. Full article

Background/Objectives: The treatment of ovarian cancer (OC), which is predominantly diagnosed in advanced stages, poses a significant challenge to modern gynecologic oncology practice. A significant proportion of patients exhibit chemoresistance, underscoring the need for novel therapeutic interventions. This challenge is further compounded by the immunogenic nature of this neoplasm, prompting the exploration of alternative therapies. A notable example is the use of trastuzumab-deruxtecan (T-DXd), an antibody-drug conjugate (ADC), that has demonstrated encouraging outcomes in preliminary studies and has the potential to become a new treatment option. This systematic review aims to prove that. Methods: The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) structure was employed to systematically search the PubMed and Scopus databases from December 2024. Furthermore, authors employed materials from the FDA’s official website and registry of clinical trials that are currently recruiting participants for T-DXd’s studies. Eligible studies included randomized controlled trials and observational studies assessing T-DXd in patients with OC. Outcomes of interest were objective response rate (ORR), median overall survival, adverse effects, and progression-free survival. Data was synthesized narratively. Results: Following a thorough review of available literature, 30 scientific papers were selected for inclusion. A total of 598 patients participated in clinical trials. The most common adverse effects were blurred vision and nausea, generally manageable. The risk of bias was low in most studies. Conclusions: T-DXd shows promising efficacy. A comparison of T-DXd with the ADC currently approved for OC therapy reveals that both demonstrate similar median overall survival and ORRs. However, the drug has exhibited significant adverse effects in breast cancer trials and has been studied on a relatively small number of patients. Therefore, further clinical trials focusing on OC patients are necessary to better assess the safety and efficacy of T-DXd in this population. Full article

While research has established the importance of Self-Regulated Learning (SRL) strategies for student achievement, their effective instruction in classrooms is often lacking. This study adopted a novel methodology that focused on Australian students’ perspectives of their teachers’ promotion of SRL strategies. Eight secondary school teachers completed a professional learning programme aimed at promoting SRL during regular classroom instruction and submitted a video excerpt of their instruction. These videos were used as stimuli for semi-structured stimulated recall interviews conducted with 25 students. Students were asked to describe their teachers’ SRL strategy instruction in terms of ‘What, When, Why and How?’. Associations between instances where students provided a clear description of the purpose and possibilities for transfer of SRL strategies and their teachers’ actions, manner of promotion and choice of strategy type were explored. Results indicate that SRL instruction was most noticed by students when it consisted of naming the strategy, providing a clear process to be followed to apply the strategy, and was accompanied by teachers’ explanations about how and why the strategy improves learning, combined with prompts to encourage students to provide examples of transfer. The implications of these results for how teachers can best promote SRL in the classroom are discussed. Full article

Pandemics emphasize the importance of real-time, interpretable clinical decision-support systems for identifying high-risk patients and assisting with prompt triage, particularly in data-intensive healthcare systems. This paper describes a novel dual big-data pipeline that includes (i) a streaming module for real-time epidemiological hospitalization risk prediction and (ii) a supplementary imaging-based detection and reasoning module for chest X-rays, with COVID-19 as an example. The first pipeline uses state-of-the-art machine learning algorithms to estimate patient-level hospitalization risk based on data from the Centers for Disease Control and Prevention’s (CDC) COVID-19 Case Surveillance dataset. A Bloom filter accelerated triage by constant-time pre-screening of high-risk profiles. Specifically, after significant experimentation and optimization, one of the models, XGBoost, was selected because it achieved the best minority-class F1-score (0.76) and recall (0.80), outperforming baseline models. Synthetic data generation was employed to mimic streaming workloads, including a strategy that used the Conditional Tabular Generative Adversarial Network (CTGAN) to produce the best balanced and realistic distributions. The second pipeline focuses on diagnostic imaging and combines an advanced convolutional neural network, EfficientNet-B0, with Grad-CAM visual explanations, achieving 99.5% internal and 99.3% external accuracy. A lightweight Generative Pre-trained Transformer (GPT)-based reasoning layer converts model predictions into auditable triage comments (ALERT/FLAG/LOG), yielding traceable and interpretable decision logs. This scalable, explainable, and near-real-time framework provides a foundation for future multimodal and genomic advancements in public health readiness. Full article

Personalized medicine regarding the biopsychosocial model can extend to females considering fertility choices through online one-on-one interactions. This finding is relevant, as recent publications suggest that online one-on-one interventions might help them in this regard. An examination of one online one-on-one intervention considers its conceptual appropriateness. The investigation is through a narrative historical analysis of a previous online group meeting, personalized to help researchers reduce their burnout. The finding is that, with an adaptation of the group process to the individual’s schedule, some participants became overwhelmed by being responsible for their schedule. By using a modification of the same process—one that does not depend on them determining their participation schedule—females can respond to writing prompts that reveal their values, from the most objective to those that are increasingly subjective. However, notably, those who are clear about their values would likely experience the least difficulty in assuming responsibility for their participation. In this regard, methodological examples of possible prompts for the modified process are offered. Through the appropriate personalization of an online, one-on-one process, the future aim in testing this process is to improve the likelihood of success in helping females clarify their values for making fertility-related decisions. Full article

The development of systems for the identification of elasmobranchs, including sharks and rays, is crucial for biodiversity conservation and fisheries management, as they represent one of the most threatened marine taxa. This challenge is constrained by data scarcity and the high morphological similarity among species, which limits the applicability of traditional supervised models trained on specific datasets. In this work, we propose an informed zero-shot learning approach that integrates external expert knowledge into the inference process, leveraging the multimodal CLIP framework. The methodology incorporates three main sources of knowledge: detailed text descriptions provided by specialists, schematic illustrations highlighting distinctive morphological traits, and the taxonomic hierarchy that organizes species at different levels. Based on these resources, we design a pipeline for prompt extraction and validation, taxonomy-aware classification strategies, and enriched embeddings through a prototype-guided attention mechanism. The results show significant improvements in CLIP’s discriminative capacity in a complex problem characterized by high inter-class similarity and the absence of annotated examples, demonstrating the value of integrating domain knowledge into methodology development and providing a framework adaptable to other problems with similar constraints. Full article

The rapid uptake of generative AI (e.g., ChatGPT, DALL·E and MS Copilot) is disrupting conventional notions of authenticity in assessment across higher education. The dominant response, surveillance and AI detection, misdiagnoses the problem. In an AI-mediated world, authenticity cannot be policed into existence; it must be redesigned. Situating AI within contemporary knowledge work shaped by digitisation, collaboration and evolving ethical expectations, we reconceptualise authenticity as something constructed in contexts where AI is expected, declared and scrutinised. The emphasis shifts from what students know to how they apply knowledge, make judgement, and justify choices with AI in the loop. We offer practical design for learning moves, i.e., discipline-agnostic learning design patterns that position AI as a collaborator rather than a cheating application: tasks that require students to critique, adapt and verify AI outputs, provide explicit process transparency (prompts, iterations, rationale) and exercise assessable demonstrations of digital discernment and ethical judgement. Examples include asking business students to interrogate a chatbot-generated market analysis and inviting pre-service teachers to evaluate AI-produced lesson plans for inclusivity and pedagogical soundness. Reflective artefacts such as metacognitive commentary, process logs, and oral defences make students’ thinking visible, substantiate attribute, and reduce reliance on punitive “gotcha” approaches. Our contribution is twofold: i. a conceptual account of authenticity fit for an AI-mediated world, and ii. a set of actionable, discipline-agnostic patterns that can be tailored to local contexts. The result is an integrity stance anchored in design rather than detection, enabling assessment that remains meaningful, ethical and intellectually demanding in the presence of AI, while advancing a broader shift toward assessment paradigms that reflect real-world professionalism. Full article

Age-related macular degeneration (AMD) is a leading cause of irreversible blindness, yet current tools for forecasting treatment outcomes remain limited by either the opacity of deep learning or the rigidity of rule-based systems. To address this gap, we propose a hybrid neuro-symbolic and large language model (LLM) framework that combines mechanistic disease knowledge with multimodal ophthalmic data for explainable AMD treatment prognosis. In a pilot cohort of ten surgically managed AMD patients (six men, four women; mean age 67.8 ± 6.3 years), we collected 30 structured clinical documents and 100 paired imaging series (optical coherence tomography, fundus fluorescein angiography, scanning laser ophthalmoscopy, and ocular/superficial B-scan ultrasonography). Texts were semantically annotated and mapped to standardized ontologies, while images underwent rigorous DICOM-based quality control, lesion segmentation, and quantitative biomarker extraction. A domain-specific ophthalmic knowledge graph encoded causal disease and treatment relationships, enabling neuro-symbolic reasoning to constrain and guide neural feature learning. An LLM fine-tuned on ophthalmology literature and electronic health records ingested structured biomarkers and longitudinal clinical narratives through multimodal clinical-profile prompts, producing natural-language risk explanations with explicit evidence citations. On an independent test set, the hybrid model achieved AUROC 0.94 ± 0.03, AUPRC 0.92 ± 0.04, and a Brier score of 0.07, significantly outperforming purely neural and classical Cox regression baselines (p ≤ 0.01). Explainability metrics showed that >85% of predictions were supported by high-confidence knowledge-graph rules, and >90% of generated narratives accurately cited key biomarkers. A detailed case study demonstrated real-time, individualized risk stratification—for example, predicting an >70% probability of requiring three or more anti-VEGF injections within 12 months and a ~45% risk of chronic macular edema if therapy lapsed—with predictions matching the observed clinical course. These results highlight the framework’s ability to integrate multimodal evidence, provide transparent causal reasoning, and support personalized treatment planning. While limited by single-center scope and short-term follow-up, this work establishes a scalable, privacy-aware, and regulator-ready template for explainable, next-generation decision support in AMD management, with potential for expansion to larger, device-diverse cohorts and other complex retinal diseases. Full article

Background: Randomized controlled trials (RCTs) are the foundation of evidence-based medicine. However, the rapid pace of technological innovation in cardiovascular surgery and interventional cardiology challenges the traditional RCT framework. Observational studies may hold renewed value in fields where device evolution outpaces the time required to validate clinical outcomes. Methods: This analysis evaluates 270 randomized and non-randomized studies in transcatheter aortic valve implantation (TAVI), one of the most rapidly evolving areas in cardiovascular medicine. The investigation follows two lines: first, mapping the timeline of major RCTs against the introduction of new prosthetic models; second, comparing the prevalence, duration, and role of randomized (R) versus non-randomized (NR) studies. Results: The timeline reveals a persistent misalignment between innovation and validation. New prosthetic models frequently enter the market while RCTs for prior generations are still ongoing. For example, the Sapien 3 valve was approved, while trials on Sapien XT were still enrolling. Similarly, newer Evolut and Acurate models were introduced during ongoing studies of earlier versions, often prompting new studies before existing ones concluded. This leapfrogging effect fragments the evidence base and delays definitive comparisons. In parallel, randomized trials have increased in number and tend to be shorter in duration, reflecting a maturing field. However, non-randomized studies remain crucial for early testing and post-market surveillance. Conclusions: In a field with rapid technological evolution a sort of Zeno’s paradox occurs: long-term validation cannot keep pace with fast innovation, resetting the evidence base with each new model. To overcome this paradox, a paradigm shift in evidence generation is desirable. Future strategies must augment adaptive trial designs, leverage real-world data and use higher-level, advanced analyses to incorporate subjective variables and phenotypic diversity, to reduce confounding factors and speed up data access. Higher-level, integrative evidence analytics could help Achilles walk alongside the tortoise. Full article

Background: Over the past several decades, the prevalence of gestational diabetes mellitus (GDM) has risen markedly worldwide, posing serious threats to both maternal and child health by increasing adverse pregnancy outcomes and long-term metabolic risks. Developing effective risk prediction tools for early detection and intervention has become the most important clinical priority in this field. The current GDM prediction models primarily rely on non-modifiable factors, for example age and body mass index, while modifiable factors such as lifestyle and health literacy, although strongly associated with GDM, have not been fully utilized in risk assessment. This study sought to establish and validate a nomogram prediction model combining modifiable and non-modifiable risk factors, with the goal of identifying high-risk Chinese pregnant women with GDM at an early stage and promoting targeted prevention and personalized prenatal management. Methods: A multicenter study was conducted across 7 maternal health institutions in Southern China (2021–2023), enrolling 806 singleton pregnant women (14–23⁺⁶ weeks). The collected data included sociodemographic, clinical history, and modifiable factors collected through validated questionnaires: dietary quality, physical activity level, sleep quality, and nutrition and health literacy. GDM was diagnosed via 75 g oral glucose tolerance test at 24–28 weeks. Predictive factors were identified through multi-variable logistic regression. A nomogram model was developed (70% modeling group) and validated (30% validation group). Receiver operator characteristic curves, calibration curves, and decision curve analysis were used to evaluate the prediction ability, the degree of calibration, and the clinical benefit of the model, respectively. Results: The finalized risk prediction model included non-modifiable factors such as maternal age, pre-pregnancy weight, and maternal polycystic ovary syndrome, as well as modifiable factors including dietary quality, physical activity level, sleep quality, nutrition and health literacy. The application of the nomogram in the modeling group and the validation groups showed that the model had high stability, favorable predictive ability, good calibration effect and clinical practicality. Conclusions: Overall, the integrated model demonstrates significant clinical utility as it facilitates the prompt identification of individuals at heightened risk and offers actionable targets for personalized interventions. In terms of future implementation, this model can be integrated into prenatal care as a rapid scoring table during early pregnancy consultations or incorporated into mobile health applications. This approach fosters precise prevention strategies for GDM in maternal health by emphasizing nutrition and health literacy, supplemented by coordinated adjustments in diet, physical activity, and sleep. Full article

Solar photovoltaic systems now produce the lowest-cost electricity in history and coupling with agriculture in agrivoltaics increases crop yields. This indicates solar will continue to experience explosive growth. Concerns exist, however, about the long-term end-of-life decommissioning of solar farms. For example, due to fossil fuel decommissioning mismanagement, Alberta is inundated with orphaned oil and gas wells that have remediation cost estimates of CAD$100 billion. Such comparisons have prompted preemptive legislation targeting solar farms, but is the fear justified? This study addresses this question by (1) analyzing warranted and actual lifespans of key agrivoltaic system components, (2) experimentally measuring microclimate impacts of two agrivoltaic arrays (fully powered with electricity extraction and unpowered to simulate post-inverter-failure conditions) and (3) quantifying agrivoltaic yield gains based on crops previously shown to respond positively to such conditions. Experimental results indicate that unpowered photovoltaic shading not only moderates soil temperatures but also enhances soil moisture conservation relative to unshaded conditions. This study demonstrates that agrivoltaic systems, even after the cessation of power generation, can continue to deliver meaningful agronomic and economic value through passive shading and policy frameworks should adapt to this dual-use reality. Integrating agronomic co-benefits into decommissioning policy supports long-term farm productivity and climate resilience. Full article