Search Results (323)

Background/Objective: Dry eye syndrome affects 16 million Americans with USD 52 billion in annual healthcare costs. With large language models (LLMs) increasingly used for healthcare information, understanding their performance in delivering equitable dry eye guidance across diverse populations is critical. This study aims to evaluate and compare five major LLMs (Grok, ChatGPT, Gemini, Claude.ai, and Meta AI) regarding dry eye syndrome information delivery across different demographic groups. Methods: LLMs were queried using standardized prompts simulating a 62-year-old patient with dry eye symptoms across four demographic categories (White, Black, East Asian, and Hispanic males and females). Responses were analyzed for word count, readability, cultural sensitivity scores (0–3 scale), keyword coverage, and response times. Results: Significant variations existed across LLMs. Word counts ranged from 32 to 346 words, with Gemini being the most comprehensive (653.8 ± 96.2 words) and Claude.ai being the most concise (207.6 ± 10.8 words). Cultural sensitivity scores revealed Grok demonstrated highest awareness for minority populations (scoring 3 for Black and Hispanic demographics), while Meta AI showed minimal cultural tailoring (0.5 ± 0.5). All models recommended specialist consultation, but medical term coverage varied significantly. Response times ranged from 7.41 s (Meta AI) to 25.32 s (Gemini). Conclusions: While all LLMs provided appropriate referral recommendations, substantial disparities exist in cultural sensitivity, content depth, and information delivery across demographic groups. No LLM consistently addressed the full spectrum of dry eye causes across all demographics. These findings underscore the importance for physician oversight and standardization in AI-generated healthcare information to ensure equitable access and prevent care delays. Full article

Modern vehicles are rapidly transforming into interconnected cyber–physical systems that rely on advanced sensor technologies and pervasive connectivity to support autonomous functionality. Yet, despite this evolution, standardized methods for quantifying cybersecurity vulnerabilities across critical automotive components remain scarce. This paper introduces a novel hybrid model that integrates expert-driven insights with generative AI tools to adapt and extend the Common Vulnerability Scoring System (CVSS) specifically for autonomous vehicle sensor systems. Following a three-phase methodology, the study conducted a systematic review of 16 peer-reviewed sources (2018–2024), applied CVSS version 4.0 scoring to 15 representative attack types, and evaluated four free source generative AI models—ChatGPT, DeepSeek, Gemini, and Copilot—on a dataset of 117 annotated automotive-related vulnerabilities. Expert validation from 10 domain professionals reveals that Light Detection and Ranging (LiDAR) sensors are the most vulnerable (9 distinct attack types), followed by Radio Detection And Ranging (radar) (8) and ultrasonic (6). Network-based attacks dominate (104 of 117 cases), with 92.3% of the dataset exhibiting low attack complexity and 82.9% requiring no user interaction. The most severe attack vectors, as scored by experts using CVSS, include eavesdropping (7.19), Sybil attacks (6.76), and replay attacks (6.35). Evaluation of large language models (LLMs) showed that DeepSeek achieved an F1 score of 99.07% on network-based attacks, while all models struggled with minority classes such as high complexity (e.g., ChatGPT F1 = 0%, Gemini F1 = 15.38%). The findings highlight the potential of integrating expert insight with AI efficiency to deliver more scalable and accurate vulnerability assessments for modern vehicular systems.This study offers actionable insights for vehicle manufacturers and cybersecurity practitioners, aiming to inform strategic efforts to fortify sensor integrity, optimize network resilience, and ultimately enhance the cybersecurity posture of next-generation autonomous vehicles. Full article

The adolescent years are pivotal in reproductive and sexual development and maturation, yet the experience of migration can severely disrupt this period, inhibiting young immigrants’ knowledge, access, and engagement with sexual and reproductive health (SRH) services. Further, young immigrants and minority populations often face persistent intersectional barriers, including language difficulties, cultural stigma, and systemic exclusion, that result in adverse SRH outcomes. Recent advances in SRH care, particularly in digital health and community-based interventions, show promise in improving access to culturally appropriate SRH services and information. Co-designing SRH programs with families and young immigrants to adequately acknowledge the unique cultural norms and barriers in SRH is essential in ensuring a high outreach of interventions. Shifts in traditional health policies are needed to ensure that immigrant and minority adolescents are not overlooked and that SRH programs incorporate culturally relevant content that is easily and widely accessible. Despite positive shifts, several barriers remain: limited disaggregated data on diverse populations, inadequate policy attention, and the insufficient scalability and funding of promising interventions. Future research and promotional efforts must prioritise the co-creation of SRH interventions with stakeholders and affected communities, ensuring that services are sustainable, culturally appropriate, and accessible to all adolescents. Full article

In the era of digital healthcare, electronic health records generate vast amounts of data, much of which is unstructured, and therefore, not in a usable format for conventional machine learning and artificial intelligence applications. This study investigates how to extract meaningful insights from unstructured radiology reports written in Finnish, a minority language, using machine learning techniques for text analysis. With this approach, unstructured information could be transformed into a structured format. The results of this research show that relevant information can be effectively extracted from Finnish medical reports using classification algorithms with default parameter values. For the detection of breast tumour mentions from medical texts, classifiers achieved high accuracy, almost 90%. Detection of metastasis mentions, however, proved more challenging, with the best-performing models Support Vector Machine (SVM) and logistic regression achieving an F1-score of 81%. The lower performance in metastasis detection is likely due to the more complex problem, ambiguous labeling, and the smaller dataset size. The results of classical classifiers were also compared with FinBERT, a domain-adapted Finnish BERT model. However, classical classifiers outperformed FinBERT. This highlights the challenge of medical language processing when working with minority languages. Moreover, it was noted that parameter tuning based on translated English reports did not significantly improve the detection rates, likely due to linguistic differences between the datasets. This larger translated dataset used for tuning comes from a different clinical domain and employs noticeably simpler, less nuanced language than the Finnish breast cancer reports, which are written by native Finnish-speaking medical experts. This underscores the need for localised datasets and models, particularly for minority languages with unique grammatical structures. Full article

Background: Human papillomavirus (HPV) is the most common sexually transmitted infection in the United States (U.S.) and a major contributor to several cancers, including cervical, anal, penile, and oropharyngeal cancers. Although a safe and effective vaccine is available, HPV vaccination rates remain suboptimal, particularly among racial, ethnic, and immigrant minority groups. This study explored multiple factors, such as cultural, social, and structural influences, influencing HPV vaccine decision-making among Cape Verdean immigrant parents in the U.S., a population currently underrepresented in HPV research. Methods: Qualitative study using individual, in-depth interviews with Cape Verdean immigrant parents of children aged 11 to 17 years living in the U.S. Interviews were transcribed verbatim and analyzed thematically using the social ecological model (SEM) to identify barriers and facilitators at the intrapersonal, interpersonal, organizational, community, and policy levels. Results: Forty-five Cape Verdean parents (27 mothers, 18 fathers) participated. Fathers were significantly older than mothers (50.0 vs. 41.1 years, p = 0.05). Most were married or partnered (60%), had at least a high school education (84.4%), and reported annual household incomes of US$50,000 or more (66.7%), with no significant gender differences. Nearly all spoke Creole at home (95.6%). Fathers had lower acculturation than mothers (p = 0.05), reflecting less adaptation to U.S. norms and language use. Most parents had limited knowledge of HPV and the vaccine, with gendered beliefs and misconceptions about risk. Only seven mothers (25.9%) reported receiving a provider recommendation; all indicated that their children had initiated vaccination (1 dose or more). Mothers were the primary decision-makers, though joint decision-making was common. Trust in providers was high, but poor communication and the lack of culturally and linguistically appropriate materials limited informed decision-making. Stigma, misinformation, and cultural taboos restricted open dialogue. Trusted sources of information included schools, churches, and Cape Verdean organizations. While parents valued the U.S. healthcare system, they noted gaps in public health messaging and provider engagement. Conclusions: Findings revealed that HPV vaccine uptake and hesitancy among Cape Verdean immigrant parents in the U.S. were influenced by individual beliefs, family dynamics, healthcare provider interactions, cultural norms, and structural barriers. These findings highlight the need for multilevel strategies such as culturally tailored education, community engagement, and improved provider communication to support informed vaccination decisions in this population. Full article

Immersive learning has been recognized as a promising paradigm for enhancing educational experiences through the integration of VR. We propose an architecture for intelligent tutoring in immersive VR environments that employs LLM-based non-playable characters. Key system capabilities are identified, including natural language understanding, real-time adaptive dialogue, and multimodal interaction through hand tracking, gaze detection, and haptic feedback. The system synchronizes speech output with NPC animations, enhancing both interactional realism and cognitive immersion. This design demonstrates that AI-driven VR interactions can significantly improve learner engagement. System performance was generally stable; however, minor latency was observed during speech processing, indicating areas for technical refinement. Overall, this research highlights the transformative potential of VR in education and emphasizes the importance of ongoing optimization to maximize its effectiveness in immersive learning contexts. Full article

Acquiring prosodic focus marking in a second language (L2) is difficult for learners whose native language utilizes strategies that differ from those of the target language. German typically uses pitch accents (L+H*/H*) to mark focus, while (Modern Standard) Arabic preferably employs a syntactic strategy (word order) or lexical means. In Syrian Arabic, a variety which is predominantly oral, pitch accents are used to mark focus, but the distribution and types are different from German. The present study investigates how Syrian Arabic learners of German prosodically mark focus in L2 German. A question–answer paradigm was used to elicit German subject-verb-object (SVO)-sentences with broad, narrow, or contrastive focus. Productions of advanced (C1, N = 17) and intermediate (B1/B2, N = 8) Syrian Arabic learners were compared to those of German controls (N = 12). Like the controls, both learner groups successfully placed pitch accents on focused constituents. However, learners, especially those with lower proficiency, used more pitch accents in non-focal regions than the controls, revealing challenges in de-accentuation. These may result from the larger number of phrase boundaries in learners’ productions, which in turn might be explained by transfer from the L1 or aspects of general fluency. Learners also differed from the controls with respect to accent type. They predominantly used H* for narrow or contrastive focus (instead of L+H*); proficiency effects played only a minor role here. Our study hence reveals an intricate interplay between cross-linguistic influence and proficiency in the L2 acquisition of prosodic focus marking, targeting a language pair so far underrepresented in the literature (German vs. Syrian Arabic). Full article

This paper examines morphosyntactic agreement in gender and number within the spontaneous spoken discourse of Spanish-speaking adults with Williams syndrome (WS), compared to that of typically developing (TD) speakers. Data were collected through natural speech transcriptions from both WS and TD groups. The analysis was conducted using Netlang 1.0.0—a piece of corpus annotation software—based on Dependency Grammar, to capture agreement patterns among determiners, nouns, and adjectives. The findings reveal that WS speakers’ gender and number agreement patterns are closely aligned with those observed in TD speakers, with only minor variations, such as a slight tendency toward unmarked gender forms among TD participants. Additionally, error rates are low in both groups, suggesting that observed discrepancies might be due to individual variation rather than condition-specific deficits, even though the statistical power of the study is limited. These results contribute to the ongoing debate on language abilities in WS, indicating that individuals with WS produce morphosyntactic agreement similarly to individuals with TD. Further research with larger datasets is recommended to validate these results, as individual variability within the WS group underscores the need for a more nuanced approach to analysis. Full article

Background: The aim of this systematic review is to investigate effectiveness and safety of sedation–analgesia techniques in controlling pain during botulinum injections in patients with cerebral palsy (CP). Methods: The Pubmed, Cinahl, and Scopus databases were searched. Inclusion criteria were as follows: cerebral palsy; any type of outcome measure regarding pain and side effects assessment; any type of studies; and English language. RoB2 and Robins-I were applied to assess the risk of bias. Tables and forest plots synthetized the findings. Results: Seventeen reports were included; most regarded pain control, and ten investigated side effects. Three were RCTs, three were controlled, and twelve were observational studies. Several techniques were used, often in combination, such as non-pharmacological approaches (clown care or virtual reality); topical anesthesia with Emla^®®, vapocoolant spray, or ice; and light-to-deep sedation with inhaled nitrous oxide, intranasal fentanyl, rectal, enteral, or intravenous midazolam, or intravenous ketamine or propofol. Vomiting and oxygen desaturation were uncommon complications. Conversely, the pooled incidence of other minor side effects was 6.39% (95% CI: 1.47–14.42%) under the random-effects model, with considerable heterogeneity. Conclusions: All the techniques are safe, if administered in an appropriate setting. Deep sedation is more effective in pain control but requires an anesthetist. A combined individualized approach is preferrable. PROSPERO CRD42025639999. Full article

This study investigates Spanish-language public discourse on YouTube following the unprecedented Iberian Peninsula blackout of 28 April 2025. Leveraging comments extracted via the YouTube Data API and analyzed with the OpenAI GPT-4o-mini model, it systematically examined 76,398 comments from 360 of the most relevant videos posted on the day of the event. The analysis explored emotional responses, sentiment trends, misinformation prevalence, civic engagement, and attributions of blame within the immediate aftermath of the blackout. The results reveal a discourse dominated by negativity and anger, with 43% of comments classified as angry and an overall negative sentiment trend. Misinformation was pervasive, present in 46% of comments, with most falsehoods going unchallenged. The majority of users attributed the blackout to government or political failures rather than technical causes, reflecting a profound distrust in institutions. Notably, while one in five comments included a call to action, only a minority offered constructive solutions, focusing mainly on infrastructure and energy reform. These findings highlight the crucial role of multilingual, real-time crisis communication and the unique information needs of Spanish-speaking populations during emergencies. By illuminating how rumors, emotions, and calls for accountability manifest in digital spaces, this study contributes to the literature on crisis informatics, digital resilience, and inclusive sustainability policy. Full article

With the emergence of AI technology, its adoption in higher education has become an interesting field for researchers. The present study explores the acceptance of AI for learning the English language by Pakistani EFL students using the UTAUT-2 and Metacognition theory. The UTAUT-2 questionnaire was adapted with minor changes to make it suitable for the EFL context. Data were collected from the English departments of the top ten general universities in Pakistan to make the findings generalizable. Another step taken to ensure generalizability was the sampling of 611 students randomly from both undergraduate (BS and ADP) and postgraduate (MPhil and PhD) programs studying in different semesters. PLS-SEM was employed for data analysis. In the first step, the PLS algorithm was run for the measurement model, which confirmed the reliability, validity, and fitness of the model. Second, the bootstrapping method was used for hypothesis testing. The findings reveal that six of the ten hypotheses for direct relationships are supported. Habit (0.489) was found to be the strongest contributor to BI, followed by PE (0.141), SI (0.100), and FC (0.093). Moreover, actual use behaviour was predicted by habit (0.325) instead of BI and FC. These findings are supported by metacognition theory, as the habit of AI seems to shape the metacognitive knowledge of EFL learners in place of traditional learning methods, and other factors seem to reinforce the metacognitive experience of using AI language. The study suggests implications for EFL experts, academia, and policymakers to strategically integrate AI into language learning by informing them of its potential benefits and risks. Full article

Healthcare disparities in cancer care remain pervasive, driven by intersecting socioeconomic, racial, and insurance-related inequities. These disparities manifest in various forms such as limited access to medical resources, underrepresentation in clinical trials, and worse cancer outcomes for marginalized groups, including low-income individuals, racial minorities, and those with inadequate insurance coverage, who face significant barriers in accessing comprehensive cancer care. This manuscript explores the multifaceted nature of these disparities, examining the roles of socioeconomic status, race, ethnicity, and insurance status in influencing cancer care access and outcomes. Historical and contemporary data highlight that minority racial status correlates with reduced clinical trial participation and increased cancer-related mortality. Barriers such as insurance coverage, health literacy, and language further hinder access to cancer treatments. Addressing these disparities requires a systemic approach that includes regulatory reforms, policy changes, educational initiatives, and innovative trial and treatment designs. This manuscript emphasizes the need for comprehensive interventions targeting biomedicine, socio-demographics, and social characteristics to mitigate these inequities. By understanding the underlying causes and implementing targeted strategies, we can work towards a more equitable healthcare system. This involves improving access to high-quality care, increasing participation in research, and addressing social determinants of health. This manuscript concludes with policy recommendations and future directions to achieve health equity in cancer care, ensuring optimal outcomes for all patients. Full article

Latvia after the collapse of the Soviet Union regained its independence in 1991. Since then, many political and social reforms have been introduced, minority education among them. Latvia began gradually abandoning the use of minority languages as mediums of instruction and switching to teaching exclusively in Latvian as the sole state language. This caused protests by minority groups, especially by Russians—the largest minority group in Latvia. The article examines 77 online news articles by Latvian, Russian, and European media covering protests against minority education reform in Latvia between 2004 and 2024. Each news article used at least one photograph/video of placard(s) with written information from the protests. The aim of the article is to understand how different media represent the linguistic landscape of protests against minority education reform and what are the main discourses they create and maintain regarding to the linguistic landscape of such protests in Latvia. The description of the linguistic landscapes shows three main trends: (1) only journalists (most often anonymous) describe the written information expressed at the protests, (2) emphasis is on the number of placard holders at the protests, their age and affiliation with minority support organizations and political parties, (3) author(s) quote individual slogans, more often demonstrated from one protest to another, without disclosing in which language they were originally written and what problems (within and behind the language education) they highlight or conceal. The main narratives that are reinforced through the descriptions of the linguistic landscapes included in the articles are two: (1) the Russian community is united and persistent in the fight against the ethnolinguistically unjust education policy pursued by the government, and (2) students, parents, and the Russian community should have the right to choose which educational program to study at school. Full article

Understanding the dynamics of public opinion formation in digital environments is crucial for managing technological communications effectively. This study investigates stance interactions and opinion reversal phenomena in technical discourse through analysis of the Manus AI controversy that generated approximately 36,932 social media interactions during March 2025. Employing an integrated methodology combining Large Language Model (LLM)-enhanced stance detection with agent-based modeling (ABM), we reveal distinctive patterns challenging traditional public opinion theories. Our cross-correlation analysis identifies significant lagged interaction effects between skeptical and supportive stances, demonstrating how critical expressions trigger amplified counter-responses rather than inducing silence. Unlike prior conceptualizations of counter-silencing that emphasize ideological resistance or echo chambers, our notion of the “counter-spiral of silence” specifically highlights lagged emotional responses and reactive amplification triggered by minority expressions in digital technical discourse. We delineate its boundary conditions as arising under high emotional salience, asymmetrical expertise, and platform structures that enable real-time feedback. The agent-based simulation reproduces empirical patterns, revealing how emotional contagion and network clustering mechanisms generate “counter-spiral of silence” phenomena where challenges to dominant positions ultimately strengthen rather than weaken those positions. These findings illuminate how cognitive asymmetries between public expectations and industry realities create distinctive discourse patterns in technical contexts, offering insights for managing technology communication and predicting public response trajectories in rapidly evolving digital environments. Full article

Clinical documentation, particularly the hospital discharge report (HDR), is essential for ensuring continuity of care, yet its preparation is time-consuming and places a considerable clinical and administrative burden on healthcare professionals. Recent advancements in Generative Artificial Intelligence (GenAI) and the use of prompt engineering in large language models (LLMs) offer opportunities to automate parts of this process, improving efficiency and documentation quality while reducing administrative workload. This study aims to design a digital system based on LLMs capable of automatically generating HDRs using information from clinical course notes and emergency care reports. The system was developed through iterative cycles, integrating various instruction flows and evaluating five different LLMs combined with prompt engineering strategies and agent-based architectures. Throughout the development, more than 60 discharge reports were generated and assessed, leading to continuous system refinement. In the production phase, 40 pneumology discharge reports were produced, receiving positive feedback from physicians, with an average score of 2.9 out of 4, indicating the system’s usefulness, with only minor edits needed in most cases. The ongoing expansion of the system to additional services and its integration within a hospital electronic system highlights the potential of LLMs, when combined with effective prompt engineering and agent-based architectures, to generate high-quality medical content and provide meaningful support to healthcare professionals. Hospital discharge reports (HDRs) are pivotal for continuity of care but consume substantial clinician time. Generative AI systems based on large language models (LLMs) could streamline this process, provided they deliver accurate, multilingual, and workflow-compatible outputs. We pursued a three-stage, design-science approach. Proof-of-concept: five state-of-the-art LLMs were benchmarked with multi-agent prompting to produce sample HDRs and define the optimal agent structure. Prototype: 60 HDRs spanning six specialties were generated and compared with clinician originals using ROUGE with average scores compatible with specialized news summarizing models in Spanish and Catalan (lower scores). A qualitative audit of 27 HDR pairs showed recurrent divergences in medication dose (56%) and social context (52%). Pilot deployment: The AI-HDR service was embedded in the hospital’s electronic health record. In the pilot, 47 HDRs were autogenerated in real-world settings and reviewed by attending physicians. Missing information and factual errors were flagged in 53% and 47% of drafts, respectively, while written assessments diminished the importance of these errors. An LLM-driven, agent-orchestrated pipeline can safely draft real-world HDRs, cutting administrative overhead while achieving clinician-acceptable quality, not without errors that require human supervision. Future work should refine specialty-specific prompts to curb omissions, add temporal consistency checks to prevent outdated data propagation, and validate time savings and clinical impact in multi-center trials. Full article