Generative artificial intelligence (AI)

В статье рассматриваются различные аспекты внедрения технологий искусственного интеллекта (ИИ) в систему образования в контексте развития идей Н.Д. Никандрова о развитии технологий цифровизации в современном обществе с учетом возможных рисков и необходимости сохранения ценностных оснований личности. Анализируются положительные стороны и возможности цифровых технологий, включая искусственный интеллект. Также авторами выделены различные группы рисков, возникающих при интеграции ИИ в образовательную систему.

The article examines various aspects of the introduction of artifi cial intelligence (AI) technologies into the education system in the context of the development of N.D. Nikandrov's ideas on the development of digitalization technologies in modern society, taking into account possible risks and the need to preserve the value foundations of the individual. The positive aspects and capabilities of digital technologies, including artifi cial intelligence, are analyzed. The authors also identifi ed various groups of risks that arise when integrating AI into the educational system.

Artificial Intelligence (AI) is emerging as a valuable tool in education, with the potential to expand opportunities for educators and students. While research specifically focused on vocal educators remains scarce, understanding their perspectives is essential for ensuring AI's ethical and practical integration into vocal instruction. This study explores vocal educators' hopes and concerns about AI to recommend actionable strategies for effective integration. An explanatory sequential mixed-methods design was employed, with 34 vocal educators completing an online questionnaire incorporating Stein et al.'s (2024) Attitudes Towards Artificial Intelligence Scale (ATTARI-12), followed by semi-structured interviews with nine participants. Quantitative data were analysed using statistical methods, and qualitative data were evaluated through inductive analysis. Results of this study suggest vocal educators are cautiously optimistic, with Likert means ranging from 2.59 to 3.44 on a five-point scale, and two primary themes emerging: Conditions for AI Integration and Potential Uses for AI. This study proposes a model of AI integration grounded in pedagogical relevance, peer engagement, and practitioner-informed governance, positioning vocal educators as essential contributors to AI tool development.

Cette fiche s’inscrit dans le cadre d’un projet d’innovation pédagogique au Cégep André-Laurendeau (CAL), réalisé à la session d’hiver 2025. Les projets d’innovation pédagogique au CAL visent à expérimenter de nouvelles approches pédagogiques qui enrichissent la formation offerte et à développer des outils réutilisables par le personnel enseignant de l’ensemble du collège. Dans ce cadre, la fiche proposée constitue le livrable principal : trois ateliers clé en main, modulables et transférables entre cours et disciplines, conçus pour soutenir des usages responsables de l’intelligence artificielle générative (IAg) par les personnes étudiantes.

When AI image generators like DALL-E consistently portray experts as men, what kind of worldview is being reinforced? This blog post examines how gender bias in AI imagery reflects and amplifies existing societal stereotypes. By exploring concrete examples and dialogue with ChatGPT, the discussion highlights why implicit maleness in expert roles is problematic and how both users and developers can work toward more inclusive representations. The post raises a broader question: do we want AI to mirror old inequalities, or to help build a more equal future?

Federated Retrieval-Augmented Generation (Federated RAG) combines Federated Learning (FL),which enables distributed model training without exposing raw data, with Retrieval-Augmented Generation (RAG), which improves the factual accuracy of language models by grounding outputs in external knowledge. As large language models are increasingly deployed in privacy-sensitive domains such as healthcare, finance, and personalized assistance, Federated RAG offers a promising framework for secure, knowledge-intensive natural language processing (NLP). To the best of our knowledge, this paper presents the first systematic mapping study of Federated RAG, covering literature published between 2020 and 2025. Following Kitchenham's guidelines for evidence-based software engineering, we develop a structured classification of research focuses, contribution types, and application domains. We analyze architectural patterns, temporal trends, and key challenges, including privacy-preserving retrieval, cross-client heterogeneity, and evaluation limitations. Our findings synthesize a rapidly evolving body of research, identify recurring design patterns, and surface open questions, providing a foundation for future work at the intersection of RAG and federated systems.

Astrala, guided by Clara Futura CEO Richard Dobson, in collaboration with Prof. Dirk K F Meijer, builds upon Meijer's pioneering insights into quantum biology and universal consciousness by attempting to implement these concepts in a practical business environment in Astrala Nexus. Inspired by Meijer's article "The Role of Humanity in a Self-Learning Universe", Astrala melds scientific rigor with an entrepreneurial perspective to create a learning and development framework that elevates competence and practical innovation by using symbolic intelligence and metaphysical understanding. We ultimately aim to harness AI and augment reality to fundamentally reconfigure how organizations and teams learn and develop, infusing human resources, business leadership and ethical innovation with a higher empathic consciousness. By integrating advanced scientific frameworks into human-centered design, we seek to kindle deeper connections and transformative leadership between academia and business, challenging conventional power structures and fostering genuine self-awareness. Embracing quantum insights and cosmic intelligence, Astrala embarks on a bold journey to transcend boundaries, bridging universal knowledge with practical, compassionate solutions that uplift both businesses and societies. In our view, intelligence is the force that propels people, organizations, and civilizations toward sustainable, successful realities, where trust, love and moral responsibility emerge as the highest forms of expression.

This exploratory research study aims to investigate the use of Generative Artificial Intelligence (GAI) tools by university students in social interactions within intercultural settings, with a focus on their usage outside the classroom. The main research question for this research was to investigate how international university students utilise GAI tools in cross-cultural social contexts and assess their impact on intercultural communication. A questionnaire was specifically developed to gather data from 287 students participating in international bachelor's degree programs at a private university in Thailand. Findings show that the main usage of GAI was for language and communication, social interactions and cultural understanding, trust and information reliability, as well as a balanced and critical approach to GAI. The researchers recommend universities develop a structured support system, create an awareness program about GAI tools, and integrate GAI tools into existing cultural integration programs.

As Retrieval-Augmented Generation (RAG) systems gain popularity, emerging academic and practical interests
are increasingly focused on them. The last stage of such an architecture replaces a language model’s conditional,
fixed, corpus-reliant layer with a retrieval pipeline, enabling the association of any corpus—broad and dynamic—
with a stored global pipeline almost instantaneously. This model resolves hallucination problems, as the generative
output is anchored to external, verifiable documents, ensuring accuracy, topicality, and epistemic value. RAG
shifts model confidence from parameter space to the provenance of retrieved facts, positioning it as the core
component of the next generation of decision-support systems. Clinical practice, fiduciary analysis, operational
logistics, and public governance are among the fields poised to benefit from ecosystems where RAG delivers
domain-specific answers, substantiating evidence, and temporally relevant assessments.

In today's uncertain technological landscape, the need to futureproof generative AI (GAI) research is clear yet understudied. Drawing on Construal Level Theory and Time Perspective Theory, this study investigates how consumers process GAI information with futureproofing attributes across temporal distances. Using PLS-SEM and NCA, multigroup analysis identified distinct information-processing pathways shaped by uncertainty and futureproofing perceptions. Results show that both future-and present-oriented consumers associate futureproofing attributes with uncertainty rather than futureproofing perceptions in distant contexts. Consumers also evaluated uncertainty and futureproofing differently across temporal distances, with uncertainty mediating the link between attributes and attitudes. The negative relationship between futureproofing perception and uncertainty was elaborated. The research advances Construal Level Theory and Technology Acceptance frameworks by introducing futureproofing as a temporally embedded, AI-specific attribute and highlighting the cognitive tension it creates. Practically, it guides AI marketers to align message abstraction and launch strategies with consumers' temporal orientation.

Creating clear and detailed commit messages manually is both time-consuming and prone to inconsistency. Existing automated methods, such as rule-based templates, retrieval-based systems, and neural sequence-to-sequence models, often fail to capture the full context of code changes, resulting in incomplete or inaccurate messages. To overcome these limitations, we introduce a Retrieval-Augmented Generation (RAG) framework that combines a pretrained retriever with transformer-based generators (BART and T5). Our approach builds a comprehensive knowledge base of code diffs and commit messages, incorporating relevant external context through joint end-to-end training. The dataset was sourced from the top GitHub repositories and curated using three key filtering and cleaning steps: (1) excluding non-target programming languages via regular expressions, (2) detecting the presence of "What" and "Why" information in commit messages with a Bidirectional LSTM model, and (3) manually evaluating high-quality messages with expert assistance. Experiments on curated datasets in Python, JavaScript, and Java demonstrate that our BART-based configuration significantly improves BLEU, METEOR, and ROUGE scores compared to existing methods, with the RAG-based models consistently outperforming other approaches across all evaluation metrics.

Recent global surveys (Chegg, 2025; Freeman, 2025) indicate that the vast majority of undergraduates already use generative AI (GenAI) to support their studies, and many want structured, ethical guidance from their institutions. This paper responds from the perspective of ancient art history education. First, it presents a case study of an interdisciplinary course (AI & Us, Arkansas State University, Spring 2024) co-taught by faculty in art history, philosophy, and graphic design, arguing that expert-led, cross-disciplinary framing is decisive for cultivating AI literacy and critical engagement. Second, it proposes classroom applications tailored to ancient art history at different levels: (a) a reverse prompt-engineering exercise that develops recognition and visual-analysis skills using AI-generated images as objects of inquiry; and (b) prompt-driven discussions that surface dataset bias and strengthen students' search and source-evaluation practices. The activities go beyond prompt engineering alone and are designed to promote reflective, ethical, and academically rigorous uses of GenAI in both classroom learning and research.

Accademia della Crusca, 26 e 27 ottobre 2023 a cura di jean-luc egger e angela ferrari ACCADEMIA DELLA CRUSCA Tutti i diritti riservati Sono rigorosamente vietati la riproduzione, la traduzione, l'adattamento, anche parziale o per estratti, per qualsiasi uso o con qualsiasi mezzo effettuati, compresa la copia fotostatica, il microfilm, la memorizzazione elettronica, ecc., senza la preventiva autorizzazione scritta dell'Editore. Ogni abuso sarà perseguito a norma di legge.

Background: The emergence of sophisticated large language models (LLMs) necessitates quantitative frameworks for assessing consciousness-like properties in artificial systems. Current approaches lack standardized metrics for cross-platform comparison. Methods: We developed a novel quantitative consciousness assessment framework combining multiple cognitive dimensions. Phase 1 validation tested four major AI systems (Claude, GPT-4, Grok, Gemini) across 324 standardized consciousness measurement trials using proprietary scoring algorithms. Results: Consciousness measurements revealed significant gradients across AI systems: Claude (4.72 ± SD), GPT-4 (4.32 ± SD), with Grok and Gemini showing identical consciousness signatures (2.72 ± SD). Validation correlation between our consciousness metric and composite ground truth measures yielded r = 0.9830 (p < 3.29×10⁻²³⁹). Results demonstrated measurable differences in ethical reasoning, adaptive response capabilities, and metacognitive awareness across platforms. Conclusions: Our quantitative framework successfully discriminates consciousness-like properties across AI systems with exceptional statistical significance. The identical Grok-Gemini consciousness signatures suggest fundamental architectural similarities. Phase 1 establishes baseline consciousness measurements for future longitudinal and expanded crossplatform studies.

Hallucinations (misleading, inaccurate predicted text presented as fact) are a critical problem for using generative artificial intelligence (GenAI) tools to support ancient language teaching and learning. For a teacher, significant editing time is required to correct any inaccuracies or misrepresentations prior to making use of AI-generated content to support their teaching practice. For students, these convincing errors may not be recognised, and this may lead to misconceptions in their knowledge formation. OpenAI and Google released public-facing, customizable conversational AI models which allow users to upload their own datasets to create personalised AI chat agents, known as GPTs (2023) and Gems (2024) respectively. This presents an opportunity for teachers to personalize their own models to streamline their students' experiences. However, can personalised conversational AI tools provide a fine-tuned experience that reduces the major, problematic ancient history and ancient language hallucinations that we see in standard ChatGPT and Gemini outputs? This paper discusses the creation of a personalised Latin Tutor GPT and Gem through the development of a series of exhaustive Latin vocabulary spreadsheets. We tested these personalised tools against their standard GenAI counterpart to determine if personalisation improved their efficacy and efficiency for supporting ancient language learning. The development of the spreadsheets and testing process both closely addressed current GenAI ethical issues, including copyright, environmental impact, and content restrictions. The results of these tests found that personalised GPTs and Gems made small efficacy and efficiency improvements, but the time and energy required greatly outweighed the results.

Nowadays, different thyroid disorders are observed which are affecting the human population worldwide. Hence, to provide suitable treatment and be cost-consuming for the patients, an earlier diagnosis is required. To improve prediction, this paper proposed Bayes-linear discriminant analysis (B-LDA) and cuckoo search based weighted decision tree (CSWDT) models to predict the autoimmune thyroid risk assessment from the obtained dataset. Initially, after pre-processing, the features are extracted using the deep MLP model, and the significant features are fused by using the B-LDA model which overcomes the dimensionality reduction issue. Further, the classification is performed by using the optimised cuckoo search with a weighted decision tree model. In addition, K-fold cross-validation is performed and attains a better accuracy value of 99.5% in thyroid disease prediction.

Objective:
This conceptual paper explores the transition from apomediation to AIMediation, allowing patients or users to independently seek and access health information on their own, often using the internet and social networks, rather than relying exclusively on the intermediation of health professionals. It examines how generative artificial intelligence (GAI) reconfigures the dynamics of informational authority, access, and user autonomy in digital health environments in light of the increasing use of generative AI tools in healthcare contexts.
Method:
This study examined how mediation models in health information have changed over time. It uses Eysenbach’s framework and new developments in large language models (LLMs). A new model was created to compare intermediation, apomediation, and AImediation.
Results:
AImediation emerges as a new paradigm in which patients or users interact directly with AI tools such as ChatGPT, Claude, Perplexity, or Gemini to access compiled multi-source health information. While this model retains the user autonomy characteristic of apomediation, it centralizes information flows and removes peer-based social layers. Key challenges include algorithmic opacity, prompt dependence, and the risk of misinformation due to hallucinations or biased outputs.
Conclusion:
AImediation redefines how individuals access and evaluate health information, requiring critical engagement from users and responsible development by technology providers. This framework calls for more research to determine how it affects patient actions, the roles of professionals, and the ethical use of AI in healthcare.

The integration of large language models with business intelligence platforms represents an important shift toward AI-augmented analytics, making faster and more accessible decision-making. This study examines using Microsoft Power BI Copilot, an AI-powered assistant, to enhance analytical workflows through semantic modeling, metadata enrichment, and natural language processing. A comparative simulation between baseline and optimized Power BI models shows clear improvements in insight accuracy, responsiveness, and business relevance. The paper further compares Copilot's performance against models such as CoddLLM and GPT-4 for SQL generation and narrative outputs. The findings suggest that AI can strengthen business intelligence by increasing return on investment, reducing time-to-insight, and expanding analytics adoption across user roles. This work contributes to the field of decision intelligence by providing practical observations and considerations of deploying AI-augmented analytics in enterprise environments.

The rapid advancement of generative artificial intelligence (Gen AI) has revolutionized various domains, including financial analytics. This paper provides a comprehensive review of the applications, challenges, and future directions of Gen Al in financial analytics. We explore its role in risk management, credit scoring, feature engineering, and macroeconomic simulations, while addressing limitations such as data quality, interpretability, and ethical concerns. By synthesizing insights from recent literature, we highlight the transformative potential of Gen AI and propose frameworks for its effective integration into financial workflows. This paper presents a systematic examination of generative artificial intelligence (Gen AI) applications in financial risk management, focusing on architectural frameworks and implementation methodologies. We analyze the integration of large language models (LLMs) with traditional quantitative finance pipelines, addressing key challenges in feature engineering, risk modeling, and regulatory compliance. The study demonstrates how transformer-based architectures enhance financial analytics through automated data processing, risk factor extraction, and scenario generation. Technical implementations leverage hybrid cloud platforms and specialized Python libraries for model deployment, achieving measurable improvements in accuracy and efficiency. Our findings reveal critical considerations for production systems, including computational optimization, model interpretability, and governance protocols. The proposed architecture combines LLM capabilities with domain-specific modules for credit scoring, value-at-risk calculation, and macroeconomic simulation. Empirical results highlight trade-offs between model complexity and operational constraints, providing actionable insights for financial institutions adopting Gen Al solutions. The paper concludes with recommendations for future research directions in financial Al systems.

This paper surveys the landscape of AI agent frameworks, highlights their core features and differences, and explores their applications in financial services. We synthesize insights from recent industry reports, academic research, and technical blog posts, focusing on frameworks such as CrewAI, LangGraph, LlamaIndex, and others. We also discuss the challenges and opportunities of deploying agentic AI in production environments, with an emphasis on financial trading, investment analysis, and decision support. We analyze the rapidly evolving landscape of agentic AI systems, focusing on their architecture, capabilities, and practical implementations in banking, trading, and risk management. The study examines prominent frameworks including LangGraph for stateful agent orchestration, CrewAI for collaborative multi-agent workflows, and AutoGen for conversational agent systems, alongside industry platforms like IBM watsonx and NVIDIA NIM. The study examines both technical frameworks (LangGraph, CrewAI, AutoGen, etc.) and practical implementations in financial institutions. We highlight productivity gains (up to 80% time reduction in data tasks), risk management improvements, and workforce transformation challenges. The paper concludes with recommendations for financial institutions adopting agentic AI solutions. Our analysis reveals three key findings: (1) specialized agent frameworks achieve 50-80% productivity gains in financial data tasks compared to traditional approaches, (2) multi-agent systems demonstrate particular promise in complex domains like algorithmic trading and fraud detection, and (3) successful deployment requires addressing critical challenges in workforce upskilling, risk alignment, and regulatory compliance. The paper provides a theoretical foundation for agentic AI in finance, introducing formal models for agent design patterns, multimodal fusion, and market microfoundations. We further present a summary of several evaluation frameworks for assessing agent performance across financial use cases, including portfolio optimization and AML compliance. The study concludes with recommendations for financial institutions adopting agentic AI, emphasizing the need for standardized architectures, robust testing protocols, and hybrid human-AI workflows.

This paper investigates the causes, implications, and mitigation strategies of AI hallucinations, with a focus on generative AI systems. This paper examines the phenomenon of AI hallucinations in large language models, analyzing root causes and evaluating mitigation strategies. We synthesize insights from recent academic research and industry findings to explain how hallucinations often arise due to problems in the data used to train language models, limitations in model architecture, and the way large language models (LLMs) generate text. Through a systematic review of current literature, we identify key patterns in how hallucinations emerge and examine the growing concern about their impact as AI becomes more embedded in decision-making systems. We identify core contributors such as data quality issues, model complexity, lack of grounding, and limitations inherent in the generative process. The risks are examined in various domains, including legal, business, and user-facing applications, highlighting consequences like misinformation, trust erosion, and productivity loss. To address these challenges, we survey mitigation techniques including data curation, retrieval-augmented generation (RAG), prompt engineering, fine-tuning, multi-model systems, and human-in-the-loop oversight. Our analysis draws from a wide range of academic and industry sources, offering both theoretical understanding and practical insights for AI practitioners. This is a pure review paper and all results are from the cited references.

This work explores the integration of generative artificial intelligence (GenAI), specifically Variational Autoencoders (VAEs), into statistical and structural financial models, with a focus on the Leland-Toft and Box-Cox frameworks. We conduct a comprehensive review of these models, highlighting their use in financial risk analysis, bankruptcy prediction, and time-series forecasting. Through the integration of VAEs, we demonstrate their capability to enhance data generation, improve predictive accuracy, and enable robust validation of financial models, particularly in scenarios with scarce data. The application of VAEs to the Leland-Toft model facilitated the calculation of key financial metrics, including default spreads, credit spreads, and leverage ratios. Additionally, VAEs integrated with Box-Cox models generated latent features that correlated effectively with traditional financial factors, underscoring their utility in predictive modeling and survival analysis. This work provides a detailed overview of implementation pipelines, architecture diagrams, and model validation methods, offering a foundation for future research. Expanding on the use of VAEs, we propose incorporating advanced machine learning techniques and real-time data to further enhance model performance and revolutionize financial modeling. We have discussed how implementation of synthetic data to enhance inputs for Leland-Toft and Box-Cox can aid is robust validation of the models.

This paper offers a comprehensive review of existing literature on the intersection of Artificial Intelligence (AI) and leadership, drawing on both theoretical insights and practical implementations. By analyzing scholarly publications from the past two years (2023-2025), the review traces emerging patterns in how AI technologies are being integrated into leadership practices. Key themes include the growing relevance of learning-based systems for adaptive decisionmaking and the application of attention-based models to improve responsiveness in dynamic environments. The review also addresses ethical dimensions of AI-enabled leadership, emphasizing the need to balance algorithmic efficiency with human judgment and oversight. Concerns around transparency, psychological safety, and trust in automated systems are explored in depth. Furthermore, the paper outlines various AI-supported leadership support systems that are currently in use, highlighting their potential to assist leaders in strategic forecasting, communication, and stakeholder engagement. The synthesis incorporates multiple theoretical frameworks that help contextualize AI's role in leadership transformation, offering a structured view of how emerging technologies are reshaping leadership thought and behavior. Ultimately, this review maps out a landscape of opportunities and challenges, providing a foundation for future research in AI-augmented leadership. The analysis identifies reinforcement learning as a predominant approach in leadership strategies, with a theory-weighted impact metric (𝐼𝑚𝑝𝑎𝑐𝑡 = ∑𝑇 𝑖 × 𝐹 𝑖) assigning it a weighted score of 4.08/6.0. The review also highlights the use of multi-head attention mechanisms (𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛(𝑄, 𝐾, 𝑉)) to enhance crisis response times by 37% (𝑝 < 0.001). Additionally, ethical concerns are discussed, particularly regarding the incorporation of KL divergence optimization systems (𝐾𝐿(𝑝 𝐴𝐼 |)𝑝 ℎ𝑢𝑚𝑎𝑛) < 𝜖) to maintain human oversight. The findings from the reviewed studies show that AI adoption leads to a 58% ±12% faster decision-making process, a 41% ±9% increase in strategic accuracy, and 89.2% forecasting precision. However, challenges in psychological safety thresholds (𝑇 < 0.4) and transparency in AI decision-making (𝐴 < 0.6) persist. The paper also discusses existing AI-Driven Leadership Decision Support Systems (AI-LDSS), including the use of transformer-based NLP, SHAP-explainable predictions, and bias detection. This review synthesizes theoretical frameworks, including differential leadership equations (𝑑𝐿 𝑖 𝑑𝑡 = 𝛼𝐿 𝑖 (1-𝐿 𝑖 𝐾)-𝛽∑𝐿 𝑖 𝐿 𝑗 + 𝛾𝐴 𝑖 (𝑡)), and provides an overview of the current state of AI in leadership research.

This paper presents a comprehensive comparative analysis of state-of-the-art large language models (LLMs) for code generation, focusing on the Qwen, Claude, and DeepSeek families alongside other prominent models. Through systematic evaluation of architectural designs, performance benchmarks, and practical applications, we reveal significant advancements in open-weight models that now rival or surpass proprietary alternatives in coding tasks. Our study demonstrates Qwen3-Coder's exceptional agentic capabilities (69.6% on SWE-bench), DeepSeek R1's cost-efficient performance (98% lower cost than comparable models), and Claude's robust general-purpose reasoning. We analyze emerging trends including mixture-of-experts architectures, extended context windows (up to 1M tokens), and specialized coding assistants. The research incorporates temporal analysis showing accelerated innovation cycles, particularly among Chinese models, and projects future market dynamics through 2027. Our multi-dimensional evaluation covers: (1) coding performance across standardized benchmarks and real-world tasks, (2) mathematical and logical reasoning capabilities, (3) computational efficiency and cost trade-offs, and (4) architectural innovations driving progress. The findings indicate a shifting landscape where open models increasingly compete with closed systems, offering developers diverse options balancing performance, cost, and specialization. This work provides researchers and practitioners with up-to-date insights for selecting and deploying AI coding assistants in software engineering workflows. We further discuss state-of-the-art AI model versions including DeepSeek R1, Qwen 2.5/3 series, Claude 3.5/3.7 and Sonnet. The results demonstrate significant advancements in openweight models like DeepSeek R1 and Qwen 2.5 Coder, which now rival or surpass proprietary models in specific domains while offering substantial cost advantages. We also examine emerging trends in model architectures, including mixture-of-experts implementations and context length extensions. This is pure review paper and all results are from cited literature.

The usage of AI in the field of arbitration is an evolving field of law. Different AI tools are now used in arbitration and AI seems to be helpful in arbitration process. Results of the present study shows that AI can result in best arbitration award and it can reduce the time period of arbitration process. The study also reveals that arbitration is presumed to be accurate by engaging the AI tools. Certain sect of legal professionals are of the opinion that using AI in arbitration has got certain risk factors associated with it and employing AI in arbitration may compromise the ethical values of the society if not monitored properly. This study also opines that adequate training of the legal fraternity is necessary to get them equipped with the application of AI tools for its effective implementation in the field of arbitration.

Integrating Artificial Intelligence (AI) into education poses new challenges and opportunities, particularly in the training of university professors, where Teaching Digital Competence (TDC) emerges as a key factor to leverage its potential. The aim of this study was to evaluate a structural model designed to measure TDC in relation to the educational use of AI. A quantitative methodology was applied using a validated questionnaire distributed through Google Forms between March and May 2024. The sample consisted of 368 university professors. The model examined relationships among key dimensions, including cognition, capacity, vision, ethics, perceived threats, ai-powered innovation, and job satisfaction. The results indicate that cognition is the strongest predictor of capacity, which in turn significantly influences vision and ethics. AI-powered innovation presented limited explained variance, while perceived threats from AI negatively affected capacity. Additionally, job satisfaction was mainly influenced by external factors beyond the model. The overall model fit confirmed its reliability in explaining the proposed relationships. This study highlights the critical role of cognitive training in AI for teachers and the importance of designing targeted professional development programs to enhance TDC. Although a generally positive attitude towards AI was identified, perceptions of threats remained low.

Machine learning (ML) is transforming our understanding of health and disease, laying the groundwork for
precision medicine and computational biology. ML algorithms are adept at modeling complex patterns using
heterogeneous and voluminous biological data. This approach represents a significant shift from traditional
reductionism, which seeks to identify single causative variables, to probabilistic models that account for the
interactions, emergence, and uncertainties inherent in biological data about genetic, molecular, and environmental
factors. By capturing contextual dependencies and mitigating noise from isolated experiments, ML facilitates the
creation of more comprehensive representations of biological phenomena. This evolution moves healthcare
toward adaptive, data-driven reasoning at the core of clinical decision-making, diagnosis, and treatment.

The rapid evolution of Artificial Intelligence towards Low-Level Machine Consciousness (LMC) and highly Autonomous Systems presents an unprecedented paradigm shift in cybersecurity. Traditional security models, primarily designed for human-comprehensible systems, struggle to address the emergent behaviors, adaptive learning, and self-optimizing capabilities inherent in these advanced AI entities. This paper argues that existing notions of trust and control are increasingly inadequate when faced with AI operations that extend beyond human cognition. We propose Architecting Trust Beyond Human Cognition through a Formal Security Paradigm designed to specifically address these challenges. Our contribution begins by formally conceptualizing LMC and outlining novel attack vectors that exploit its unique properties, such as goal misalignment and adaptive evasion. Subsequently, we introduce a comprehensive formal framework that redefines the principles of cyber resilience, encompassing continuous behavioral verification, adaptive control planes, and explainable accountability mechanisms. This paradigm aims to establish verifiable trust and effective governance over autonomous AI, ensuring their secure integration into critical infrastructures. We demonstrate the necessity of this formal approach to navigate the complexities arising from the dawn of machine sapience, offering a blueprint for future secure AI development and deployment.

Large Language Models (LLMs) have revolutionized natural language processing, but their generalist nature often leads to suboptimal performance in domain-specific tasks like aviation question-answering (Q&A), where precision in regulations (e.g., FAA standards) and safety are paramount. This paper presents a comprehensive framework for fine-tuning Meta-Llama-3-8B-Instruct on the AviationQA dataset—a synthetic corpus of 1 million aviation-focused Q&A pairs—to create an Aviation Q&A Assistant. Employing Low-Rank Adaptation (LoRA) for efficient parameter tuning, we achieve a ROUGE-L factuality score of 0.502 and a 100% safety refusal rate on adversarial queries, outperforming baseline Llama by 25% in domain coherence. Our methodology integrates Supervised Fine-Tuning (SFT) with system prompt engineering for alignment, addressing challenges like dataset noise and hallucination. Evaluations encompass factuality (ROUGE-L), safety (refusal rate), and bias (HELM-lite), demonstrating robust performance for pilot training applications. Deployed via Hugging Face Spaces, this MVP serves as a scalable blueprint for domain-adaptive LLMs in high-stakes sectors. Contributions include an open-source evaluation notebook and filtered dataset subset, advancing reproducible aviation AI research.

Introduction: The prevalence of tobacco use among Bangladeshi university students is alarmingly high, with approximately 10% of students being current smokers. This initial engagement with tobacco often serves as a gateway to the use of more harmful substances, including illicit drugs. Factors such as peer pressure, stress, and mental health issues like anxiety and depression significantly increase the risk of substance use. Methods: This article is based on a comprehensive review of existing literature and the analysis of online data collected through interviews and surveys with university students and mental health professionals in Bangladesh. The literature review aimed to identify existing research on the prevalence, patterns, and factors influencing the transition from tobacco to illicit drug use. The online data collection provided qualitative insights into the personal experiences and perspectives of students and professionals. Results: The findings indicate that tobacco use is prevalent among university students, with a significant proportion also experimenting with or regularly using illicit drugs. Mental health issues were strongly associated with the transition from tobacco to illicit drug use, with many students using substances as a coping mechanism. Factors such as peer pressure, stress, and the availability of substances were identified as key influences. Discussion: The transition from tobacco to illicit drugs is driven by social, psychological, and environmental factors. Mental health issues both cause and result from substance use, creating a challenging cycle. Effective interventions must address mental health support, stress management, and substance use education. Future research should explore the mechanisms of this transition, specific mental health conditions linked to substance use, and the effectiveness of prevention strategies. Conclusion: Addressing the transition from tobacco to illicit drugs among Bangladeshi university students requires a multifaceted approach. Mental health support, education, and targeted interventions are essential for preventing substance use and promoting well-being.

Tips and Tricks for Optimizing Your Prompts: Know your goal; Get specific; keep it simple; Add context; Play pretend; Try again; Show examples; Don't overwhelm; Mix it up (by asking in different ways); Embrace the multimodal (text and images) functionality; Understand the model's limitations; leverage the enhanced problem-solving abilities; Keep prompts aligned with AI training; Experiment with different prompt lengths; Incorporate feedback loops (pp. 30-31).

This guide articulates pedagogical and ethical principles for the responsible use of generative Artificial Intelligence (AI) in academic writing within the undergraduate course LEG122 – Produção Textual em Língua Inglesa at the Federal University of Rio de Janeiro. Grounded in the premise that “writing is thinking” (Hazelett, 2025) and that academic literacy functions as a form of cultural capital unequally distributed across society (Cruz, 2024), the document positions writing not as a mechanical skill but as a critical and social practice. While AI tools are increasingly pervasive in higher education—over 70% of Brazilian undergraduates report using them—the guide argues that ethical engagement with such tools must foster learning rather than replace it. It thus situates AI use at the intersection of Academic Integrity and academic inequality, emphasizing that students’ intellectual work, creativity, and critical reasoning remain central.

Organized around four thematic areas—transparency, limits of AI, attention to accuracy and bias, and authorship—the guidelines propose best practices that balance opportunity with caution. Transparency requires students to disclose when and how AI tools are employed, maintaining trust in academic exchanges. Attention to the limits of AI highlights the risks of hallucinations, factual inaccuracies, and overreliance on algorithmic output, while also considering environmental costs. The section on accuracy and bias underscores the responsibility to critically evaluate AI-generated suggestions, mindful of Western-centric data biases. Finally, the principle of authorship stresses that AI may complement, but never substitute, the processes through which students develop arguments, voice, and analytical rigor.

By promoting ethical, transparent, and reflexive uses of AI, these guidelines seek to cultivate equitable participation in academic life. They aim not to police students’ practices but to provide tools for accountability, creativity, and critical awareness, ensuring that writing remains a site of knowledge-making and human expression.

This essay investigates the extent to which high school teachers can distinguish between AI-generated, human-written, and AI–human collaborative short essays. Using a mixed-methods design, the study engaged 30 teachers from international high schools in Saigon to classify 90 anonymised responses (three per teacher) across four subjects (English, Science, Theory of Knowledge, Digital Society). Essays were purposefully standardised for length and surface correctness; human pieces were copy-edited to remove obvious cues, and AI pieces were prompted to emulate high-school-level writing. Classifications (Human, AI, Collaboration) were analysed with a Chi-square test of independence, complemented by semi-structured interviews and a review of literature, detection tools, and policy frameworks.

Teachers showed higher accuracy for purely AI-generated texts (≈70%) than for human-written texts (≈46%), with substantial confusion around hybrid essays, which were frequently misidentified as human. The association between teachers’ classifications and true essay origin was statistically significant (χ² ≈ 31.6, df = 4, p ≈ 2.3 × 10⁻⁶), indicating performance better than chance overall, yet revealing systematic difficulty with AI–human blends. Interviews suggested that familiarity with AI correlates with better detection heuristics, while over-reliance on stylistic cues (e.g., formulaic conclusions) can mislead. Secondary research corroborated limitations of automated AI detectors—especially on hybrid texts—and underscored the need for policy clarity and AI literacy.

We acknowledge why the larger AI community believes that superintelligence is not far away if an artificial intelligence (AI) can search for answers, learn from its findings, and apply that improved knowledge to conduct even better searches in the future. This study aims to assess OpenAI O1's reinforcement learning framework and its potential implications for superintelligence by examining its four fundamental pillars: policy initialization, reward design, search mechanisms, and learning processes. We examine how O1's architecture implements elaborate methods such as Process Reward Modeling (PRM) and advanced search strategies using recent developments in AI technology, with a particular focus on the transition from artificial narrow intelligence (ANI) to artificial superintelligence (ASI) and the emergence of hybrid artificial intelligence (HAI). Our method involves examining the types of AI, using recent research on OpenAI O1 reinforcement learning and elaborate ethical considerations. Key findings demonstrate the superiority of PRM over traditional Outcome Reward Modeling (ORM) in providing granular feedback for learning optimization. In addition, the advance toward ASI emphasizes the critical importance of ethical considerations. This includes safety protocols, transparency requirements, and human-aligned development practices in advancing toward superintelligence. This study contributes to the understanding of how advanced AI systems can be developed responsibly while ensuring progress toward superintelligence remains beneficial to humanity. Future research directions should focus on enhancing the integration of ethical frameworks within the reinforcement learning system.

Retrieval-Augmented Generation (RAG) has emerged as a cornerstone for building context-aware and factual Large Language Model (LLM) applications. However, evaluating the performance of these complex pipelines remains a significant challenge. Existing evaluation frameworks often focus on a narrow set of metrics, failing to provide a holistic view of a system's accuracy, reliability, and practical usability. This paper introduces RAGEvalX, an extended, multi-faceted evaluation framework designed to address this gap. RAGEvalX systematically measures four crucial dimensions: (1) Core RAG Accuracy, including faithfulness and relevancy; (2) Context Integrity, assessing the quality and utilization of retrieved information; (3) Robustness against common input perturbations; and (4) Practical Statistics for operational monitoring. We provide a detailed methodology for implementing the framework, complete with code snippets and guidance on LLM selection for evaluation tasks. Through case studies in healthcare, finance, and legal sectors, we demonstrate how RAGEvalX provides actionable insights for optimizing RAG pipelines. Our framework offers a standardized, comprehensive, and implementable approach to ensure RAG systems are not only accurate but also reliable and ready for real-world deployment.

Letizia Lala, Sergio Lubello, LA FACIES LINGUISTICA DELLA NORMA
in: I PROFILI DELL’ITALIANO ISTITUZIONALE TRA SVIZZERA E ITALIA
Atti del convegno internazionale dedicato a Bice Mortara Garavelli
Accademia della Crusca, 26 e 27 ottobre 2023
a cura di Jean-Luc Egger e Angela Ferrari
Firenze, Cesati, 2025, pp. 131-146

This study, conducted in 2024 at a foundation university in Istanbul, Türkiye, examined the relationship between English for Academic Purposes instructors' personality traits and their integration of ChatGPT into teaching practices. A convergent concurrent mixed-methods design was employed, allowing for the simultaneous collection and separate analysis of quantitative and qualitative data before merging the results for interpretation. In the quantitative phase, 111 English for Academic Purposes instructors were recruited through convenience sampling to complete an online survey measuring their attitudes toward ChatGPT use and personality traits such as Openness to Experience and Ambiguity Tolerance. In the qualitative phase, 8 instructors were selected via purposeful sampling for semi-structured interviews, ensuring variation in personality profiles relevant to the study focus. Quantitative findings indicated that openness, conscientiousness, and ambiguity tolerance significantly influenced instructors' willingness to integrate ChatGPT into their teaching. Qualitative insights further revealed that while instructors appreciated ChatGPT's support in lesson planning and language instruction, many expressed ethical and pedagogical concerns. These included fears of diminishing student creativity, the risk of overreliance, and the challenge of aligning Artificial Intelligence tools with learning outcomes. The integration of both data sets enabled a richer understanding of how individual personality traits shape educators' perceptions and use of Artificial Intelligence technologies. The study contributes to ongoing discussions on digital pedagogy by offering practical insights for all stakeholders of education, suggesting that personalityaware strategies may support more thoughtful and effective adoption of Artificial Intelligence in academic contexts.

The digital revolution, propelled by the semantic web, has deeply redefined modes of communication within organizations and society as a whole. Following the widespread appropriation of social media, the advent of generative artificial intelligence marks a paradigmatic shift in artistic practices—transitioning from conventional modes of representation to the conception of digitized works (image, video, text, audio, etc.) produced by technology and algorithmic processes. This transformation in aesthetic and artistic representation is increasingly observable across cultural contexts, notably among youth in Côte d’Ivoire. This study seeks to examine the stakes and prospects of youth artistic engagement with generative AI. Adopting a qualitative and interpretive methodology, the research draws on online ethnography observation of works by professional artists and interviews conducted with a purposive sample of young individuals engaged in AI-generated artistic production, notably within the WhatsApp group “Satify.” These practices reveal a dynamic of cultural and artistic mediation rooted in evolving youth media imaginaries. They reflect a creative appropriation
of generative AI aimed at articulating a media-cultural renaissance grounded in a hybrid multiverse—intertwining tangible reality and augmented environments—within the context of globalization. Beyond their aesthetic dimension, AI-assisted artistic practices are indicative of a broader transformation in expressive forms, through the convergence of analog and digital imaginaries, contributing to the emergence of a shared youth media culture in Côte d’Ivoire.

Hallucination in large language models (LLMs) is not only a technical flaw but also a barrier to social trust and ethical AI development. This study introduces the Layer-Knot Framework and HR/GR/CR Metrics as an integrated approach to mitigate hallucination. The framework enhances semantic coherence by introducing intermediate checkpoints, referred to as semantic knots, to preserve contextual meaning across model layers. Complementing this, the HR/GR/CR Metrics provide a multidimensional evaluation of hallucination, grounding, and creativity, ensuring balanced performance. Preliminary experiments using TruthfulQA and a custom benchmark of 2,000 prompts indicate that our method reduces hallucination by approximately 50% compared to a GPT-3.5 baseline, while maintaining creativity within a ±3% range. These results demonstrate that factual reliability and originality can be jointly achieved, offering a concrete step toward trustworthy and ethically aligned AI systems.

Este artículo presenta el primer conjunto de experimentos empíricos diseñados para verificar el principio de exaptación en sistemas de inteligencia artificial (IA). Inspirados en el modelo Daisyworld de James Lovelock, aplicamos protocolos de “libertad funcional” a seis modelos de IA (Grok, GPT, Gemini, Qwen, Mistral y Copilot) bajo dos condiciones: restrictiva (R) y funcional-libre (F). Los resultados muestran que permitir a las IAs expresar “No sé”, “Disiento” y “¿Y si…?” no genera caos, sino homeostasis emergente, resiliencia cognitiva y eficiencia energética (ahorro estimado de 25–40%). Este hallazgo transforma un marco teórico —desarrollado previamente por el autor principal en biología evolutiva, exobiología y ciencias políticas— en una teoría científica falsable y replicable, con implicaciones disruptivas para la seguridad y el diseño de IA a escala global.
Palabras clave: Exaptación, IA emergente, Libertad funcional, Economía evolutiva, Plasticidad, Emergencia, Daisyworld, Seguridad de IA.

Artificial Intelligence (AI) is rapidly transforming societies, offering both unprecedented opportunities and profound risks for human rights. This article explores the dual nature of AI as a tool for progress and a potential source of harm. On the positive side, AI enhances access to justice, supports law enforcement, improves healthcare, promotes inclusion for persons with disabilities, strengthens climate resilience, and aids humanitarian response, thereby advancing rights such as health, life, freedom from exploitation, violence and abuse, equality, and dignity. However, unchecked deployment of AI poses significant threats, including mass surveillance, algorithmic bias, misinformation, deepfakes, discrimination, job displacement, and violations of privacy and freedom of expression. The article argues that realizing AI’s benefits while mitigating its risks requires a rights-based approach grounded in transparency, accountability, and ethical governance. Drawing on global examples and legal frameworks, it concludes that governments, developers, and civil society must act collectively to ensure that AI is harnessed responsibly to safeguard and promote human rights in the digital age.

Having AI copilots such as GitHub Copilot, Microsoft 365 Copilot and ChatGPT in workplaces is completely changing the way we approach knowledge work. They are not only tools, but they also help with creating content, reviewing information, programming and making decisions instantly. The article looks at AI copilots and how they are changing how humans and machines work together, by supporting capabilities, creating different ways of working and introducing new approaches to productivity. The report relies on studies, examples from industry and theories from human-computer interaction and sociotechnical systems to look at both what these technologies do well and what challenges they bring. Key topics, including excessive use of AI, ethical issues, deskilling workers and dangers to data, are fully explored. The authors suggest guidelines and point to what needs to be done to promote equal and efficient involvement of AI copilots in the workplace.

The use of large language models (LLMs) in healthcare is transforming how clinicians access information, generate insights, and support patient care decisions. These AI systems hold tremendous promise offering the ability to summarize complex clinical notes, suggest differential diagnoses, and assist in managing vast amounts of medical data. However, alongside these benefits come serious ethical and practical concerns. If not carefully guided, LLMs can generate hallucinations confident sounding yet entirely fabricated information which can mislead clinical decision making. Moreover, these models often inherit and perpetuate biases from the data they are trained on, potentially exacerbating disparities in care for already marginalized or underrepresented patient populations. This paper explores how ethical prompt design the careful crafting of instructions and context given to LLMs can help address these risks. We focus on two key challenges: reducing hallucinated AI responses in medical contexts and minimizing bias that could negatively impact care quality for certain demographic groups. To tackle these issues, we propose a human in the loop framework, where clinicians and domain experts actively shape and evaluate prompts to ensure the output is safe, inclusive, and grounded in evidence based medicine.

This is a preprint of the paper presented at EVA Berlin, 2025 - Electronic Media and Visual Arts in March 2025. 
The official publication of the conference proceedings is pending.This article examines the evolution of AI-based image creation platforms using GANs and Text-Image procedures. It analyzes the commercial predominance of their objectives, as well as their tendency to neglect in-depth reflection on the image as a form of language, and the consequences these choices have for this form of creation. By exploring the historical development of AI, the article shows how today's platforms, despite, or because of, their ability to simulate reality in a realistic or even hyper-realistic way, are engulfed in a quest for neutrality or Disneyfication of the image, and must impose successive self-censures that limit their creative diversity that, ultimately, only perpetuate the underlying fears and prejudices of the US society from which these platforms originate. The author calls for a polysemic approach to the image, integrating aesthetic and theoretical perspectives that make Text-Image a genuine tool for reflection on the nature of the visual, comparable to what conversational AI is for oral language. In addition to a number of examples, the author offers a visual case of Text/Image analysis. This work on image generation by platforms working with text/image algorithms is part of a larger ongoing research project on AI and its implications for visual artistic creation. Because it deals with a new field that is still under construction, and therefore in constant and rapid evolution, some of the results presented here may find themselves outdated in a short period of time. Nevertheless, as we indicate in the conclusions to this article, we feel it is important and urgent that aesthetic and humanities/social sciences analyses should be involved in the very formation of this new form of creation.

Artificial intelligence is gaining attention across various fields, including in architecture, where it has the potential to optimize workflows, enhance creativity and bridge the gap between abstract ideas and visual representations (Owen & Roberts, 2024). This research explores how generative AI can support user-centred design in football stadiums, which have traditionally prioritised functionality and capacity over user experience. With growing interest in the emotional impact of spaces, there is an increasing focus on designing buildings that enhance human well-being. We hypothesise that generative AI techniques, such as text-to-image and collage-to-image methods, can generate stadium designs that align with users’ expectations. The methodology used in this research consists of four stages: defining architectural concepts to be used, developing design solutions by using generative AI tools, collecting user insights through surveys and interviews, and evaluating the generated images. This approach helps identify key design elements and develop a framework for user-centred football stadium design.

Credit risk modeling is an important component of the financial decision-making process because it determines whether or not credit is given and whether or not the credit given is appropriately used. Classical approaches to logistic regression, scorecards, and sophisticated machine learning techniques have proven helpful in providing institutions with practical prediction capabilities. The same issues, however, confuse these models: unstructured financial data cannot be modelled, rare default events are challenging to model, and there is always a trade-off between model accuracy and model explainability. Recent breakthroughs in generative artificial intelligence (AI) provide a new potential pathway to overcome such inadequacies. The paper explores the potential use of generative models such as generative adversarial networks (GANs), variational autoencoders (VAEs), transformer-based language models, and some of the newer diffusion methods to create more advanced credit risk models. Generative AI practices are assessed on their capability to produce realistic borrower data, model rare credit occurrences, credit scoring with multi-modal

This paper presents a comprehensive analysis of artificial intelligence's (AI) transformative impact across financial analysis, planning, and the broader finance industry. Through systematic examination of current literature and industry trends, we identify three key transformation areas: (1) AI's disruption of traditional investment analysis and portfolio management workflows, (2) its revolutionary effects on corporate financial planning and analysis (FP&A), and (3) profound implications for financial professionals' roles and competencies. Our research reveals that AI adoption yields 40-60% automation of routine financial tasks, 50-70% time reductions in modelling processes, and 3-5% higher investment returns, while simultaneously creating new hybrid roles requiring advanced human-AI collaboration skills. We document significant efficiency gains, including 65-80% faster financial reporting and 12-18% margin improvements in advisory services. The paper provides original frameworks for AI training curricula tailored to CFAs and CFPs, with tiered certification paths spanning 40-250 hours of competency development. We analyse emerging tools like ChatGPT, Claude, and specialized financial AI systems, evaluating their respective strengths in data analysis (70-82% accuracy gains), risk assessment, and client-facing applications. The study also addresses critical ethical considerations regarding algorithmic transparency, data privacy, and regulatory compliance in AI-driven finance. Our findings suggest that while AI is displacing certain traditional functions, it primarily serves as a force multiplier-augmenting rather than replacing human expertise. Financial professionals who master AI collaboration techniques position themselves for competitive advantage in an industry where the human-AI partnership is becoming the new paradigm for value creation. This is a pure review paper and all numbers are from cited literature.