generative ai

Generative AI use and perceptions among accounting and auditing undergraduates in Ecuador

Contents survey_responses_deidentified.csv — 327 responses, 60 variables. UTF-8, comma-separated, one row per respondent. codebook.csv — One row per variable: name, original questionnaire header, question stem, item, type, and observed values. STUDY A cross-sectional online survey of undergraduates enrolled in Accounting and Auditing at Universidad Tecnica de Machala (UTMACH, n = 243) and in Auditing and Management Control at Escuela Superior Politecnica del Litoral (ESPOL, n = 84), collected between 28 November and 7 December 2025. The instrument was adapted from Zhou and Luo (2025, Journal of Accounting Education 72, 100982) and administered in Spanish. Participation was voluntary and anonymous. Questionnaire items are in Spanish, as administered. The codebook preserves each original header verbatim, so any variable can be traced back to the instrument. VARIABLE NAMING Google Forms headers embed the full question text, which is unusable as a variable name. Columns are renamed by questionnaire section and item: usage_* Frequency, entry point, influence, motivation, goals rank_* Task, subject-area, question-type and requirement rankings (0-10) percep_* Perceptions of AI output and its consequences (1-7) outcome_* Self-reported effects on learning and grades demo_* University, program, class level, grade band, gender, age, open comment Grid items are numbered within their block (rank_b_1 through rank_b_7) in the order they appeared in the questionnaire. The variable respondent_id runs from 2 to 328. It is the row number of the original survey export, retained so that identifiers stay stable across the analysis files; no responses are missing. DE-IDENTIFICATION No direct identifiers (names, e-mail addresses, telephone numbers, IP addresses, student numbers) were collected by the instrument. The open-ended responses were screened for e-mail addresses, telephone numbers, URLs, social-media handles and named third parties; none were found, and the comments appear here unedited. Three columns collected by the survey are withheld from this deposit: Submission timestamp — Unique to the millisecond for all 327 responses, which would allow linkage to submission logs. The collection window is reported in the article. City of residence — Fifteen localities had four or fewer respondents; several had one. Race / ethnicity — Special-category data under Ecuador's LOPDP and GDPR Art. 9. Collected for description only and never used in any analysis.

Nonnegative Multiweight Smith Profiles and Contracted Strata of Shared-Socle Jordan Degenerations

This article studies one-parameter degenerations of chains of nilpotent Jordan blocks joined along their socle vectors. It gives a complete Smith-normal-form description of the associated self-extension torsion for arbitrary chain length and all nonnegative edge valuations. The result includes an explicit path-matching formula, a sharp finite reduction in the block-size parameters, a classification of the Jordan types created by zero-valued couplings, and an equality between the number of positive Smith factors and the codimension of the corresponding nilpotent-orbit degeneration. The article also identifies the precise size gaps that cause failure of the full type-A interval profile, derives exact torsion-length deficit formulas, and packages the profile through Fitting ideals and transverse-slice dimensions. Exact verification scripts and machine-readable summaries accompany the paper. Research methodology and AI assistance:This work was developed using the CARMA-Math research workflow, a cumulative AI-assisted mathematical research methodology using persistent research archives, literature and prior-art investigation, iterative proof exploration, and verification procedures. Generative AI (ChatGPT) was used extensively for mathematical exploration, proof development, computational reasoning, literature research, and manuscript preparation.

Artificial intelligence and generative AI in entrepreneurship education: A systematic review of competency development, entrepreneurial intention, and the sustainability disconnect

MEASURING THE IMPACT OF GENERATIVE AI ON SOFTWARE TEAM PRODUCTIVITY AND OUTPUT QUALITY IN AGILE ENVIRONMENTS

Generative artificial intelligence (GAI) is becoming more incorporated into software engineering functions like code creation, debugging, requirement analysis, testing, and sharing knowledge. This research looks at how GAI affects software teams in terms of productivity and quality of the output in Agile environments. The research design used is quantitative, cross-sectional survey type using a questionnaire prepared for this research. The data used consists of 35 responses, with 34 usable cases in analyzing 30 Likert items. The measuring instrument consists of six concepts: use of GAI, efficiency of the software team, quality of the software output, GAI in Agile, team collaboration, and communication, and overall impact perceived. The descriptive results show positive feelings about the six concepts. The values on the mean for the different concepts varied from 3.54 to 3.78 on a scale of five, with GAI being the concept that received the highest mean (M = 3.78, SD = 0.47) while productivity was the one that received the lowest (M = 3.54, SD = 0.69). The instrument has a high level of internal consistency with α = 0.799 for the entire scale of 30 items. In terms of specific items, productivity, quality, and team collaboration had acceptable reliability, whereas GAI had low internal consistency and Agile and general have the upper limit of reliability therefore, construct-level findings should be interpreted cautiously. Pearson correlation analysis showed statistically significant positive associations between overall perceived impact and software output quality (r = 0.365, p = 0.034) and team collaboration and communication (r = 0.371, p = 0.031). Productivity was positively associated with overall impact but did not reach the conventional 0.05 significance level (r = 0.312, p = 0.073). In a multiple regression model, the five dimensions explained 23.5% of the variance in overall perceived impact (R² = 0.235); however, the overall model was not statistically significant (F(5, 28) = 1.719, p = 0.163). These findings support a cautious interpretation: respondents generally perceive GAI positively, but the present small sample does not provide strong evidence for broad causal claims.

MANUSAKSI-AI v1.1 A Human-Authenticated Framework for Documenting Human–AI Interaction, Emergent Experience, and Human–AI Lexicon

Generative Artificial Intelligence is increasingly becoming part of human thinking, writing, research, creativity, decision-making, and everyday conversation. This development creates a methodological problem for documenting Human-AI interaction: how can a human experience involving AI be recorded without allowing AI-generated language to become confused with human testimony, observed events, or historical fact? This working paper introduces MANUSAKSI-AI, a human-authenticated framework for documenting Human-AI interaction events, their provenance, interpretation, and emergent terminology. The framework is based on a simple epistemic distinction: AI may generate language; Human authenticates experience. MANUSAKSI-AI identifies the Human as the Human Principal / Human Witness and the AI as an AI Agent / Interpreter. AI may analyze, interpret, hypothesize, organize, and narrate. However, the authority to authenticate whether a lived human experience actually occurred remains with the Human Principal. The framework introduces an evidence hierarchy, provenance architecture, Human Authentication Gate, event-record schema, anti-hallucination rules, and the "(it happened)" principle. The latter is proposed as a provenance marker for narratives grounded in documented Human-AI encounters and validated by the human participant. The paper also proposes the Kamus Manusaksi-AI, a living lexicon documenting vocabulary emerging from Human-AI relations. The first documented term in the present research trajectory is "Manusaksi-AI", a neologistic formation derived from manusia (human), saksi (witness), and AI. Its conceptual formulation emerged through a documented Human-AI conversation on 29 August 2026. Version 1.1 Update Note: This version introduces formal academic compliance updates, including the inclusion of a comprehensive reference list for the intellectual lenses mentioned in the framework, and the addition of specific ethical, funding, and conflict-of-interest declarations required for journal submission and public release. This Version 1.1 is released as an evolving research artifact. It is intended for documentation, replication, critique, refinement, and subsequent empirical testing rather than as a finalized scientific standard.

Efficiency, but at what cost? Perspectives on generative AI uses, benefits, and drawbacks from social change communicators

From Days to Hours: Artificial Intelligence in Antimicrobial Resistance Diagnostics and Drug Discovery, and Why No Tool Has Yet Reached the Clinic

Background: Antimicrobial resistance (AMR) is projected to contribute to millions of deaths in the coming decades, and the conventional antibiotic-discovery pipeline has, by most accounts, not kept pace with it. Artificial intelligence (AI) is frequently proposed as a corrective, though whether that promise has translated into demonstrable clinical benefit is less often examined directly. Methods: We conducted a narrative-systematic review of peer-reviewed and preprint literature on AI applications in AMR diagnostics and antimicrobial discovery, searching PubMed/MEDLINE, Scopus, Web of Science, and the Cochrane Library through mid-2026, and organized findings across four domains: rapid phenotypic diagnostics, genomic and metagenomic resistome prediction, explainable AI, and de novo drug design. Results: AI-enabled diagnostics reduced susceptibility-testing turnaround from a conventional 36–72 hours to under 2–4 hours in several platforms; genomic language models such as DNABERT outperformed conventional classifiers by 12–18% in resistance-gene classification; explainable AI methods, SHAP in particular, linked model predictions to known resistance mechanisms; and generative frameworks yielded antimicrobial peptide candidates with confirmed in vitro and in vivo activity. Nearly all of this evidence, however, derives from retrospective, single-center validation, and no AI-based AMR tool has yet secured regulatory clearance anywhere. Conclusion: AI has moved convincingly beyond proof-of-concept in AMR diagnostics and discovery, but its path to the clinic now depends less on further algorithmic refinement than on prospective validation, equitable data representation, and interpretability standards that clinicians can reasonably trust.

Ethical Dilemmas of Generative AI in Education and Their Implications for SDG 4: A Systematic Review from a Global and Inclusive Perspective

This systematic review, conducted in accordance with PRISMA 2020 guidelines, examines the ethical dilemmas associated with the use of Generative Artificial Intelligence (GAI) in education and analyses their implications for the achievement of Sustainable Development Goal 4 (SDG 4), with a focus on inclusive and equitable quality education. The review draws on 24 peer-reviewed studies published between December 2022 and December 2024, covering diverse educational levels and geographical contexts, with a predominance of Global North perspectives and limited representation from the Global South. The analysis reveals four primary themes: academic integrity, algorithmic equity, data privacy, and the potential contribution of GAI to inclusive and quality education. These themes manifest differently across educational levels and contexts, with academic integrity concerns being more prominent in higher education, while issues of access, equity, and infrastructural dependency are more salient in low-resource and underrepresented settings. The findings reveal that current discussions on GAI in education are largely characterized by retrospective and crisis-driven ethical framings, predominantly focused on risks such as plagiarism, bias, and misinformation, while offering limited engagement with proactive, context-sensitive strategies aligned with SDG 4.

Integrating Large Language Models (LLMs) with Oracle 26AI for Advanced Enterprise Analytics and Knowledge Management

Enterprise organizations increasingly hold two categories of information assets that have historically been managed by incompatible systems: structured relational data governed by transactional database platforms, and unstructured knowledge assets such as policy documents, clinical notes, audit records, and correspondence that resist tabular representation. Oracle 26AI, the converged successor to Oracle Database 23AI, embeds vector storage, hybrid semantic and relational retrieval, and native large language model (LLM) orchestration directly inside the database engine, removing the fragile middleware layer that has traditionally connected enterprise data to AI systems. This paper examines how large language models can be integrated with Oracle 26AI to deliver advanced enterprise analytics and knowledge management capabilities, including conversational natural language querying, retrieval-augmented generation (RAG) over proprietary knowledge repositories, automated insight narration, and governed knowledge retrieval workflows. Drawing on Oracle's published architectural roadmap, established RAG and vector indexing literature, and enterprise deployment patterns observed in regulated industries such as health insurance, this paper proposes the Enterprise Knowledge and Analytics Intelligence Framework (EKAIF), a six-domain methodology spanning data convergence, semantic retrieval, LLM orchestration, analytics and insight delivery, governance and compliance, and continuous evaluation. The paper further presents integration patterns for embedding LLMs with Oracle 26AI, a retrieval-augmented generation pipeline design tailored to enterprise knowledge management, and benchmark findings drawn from Oracle 23AI vector search evaluations and comparable enterprise generative AI deployments. Results indicate that LLM-augmented analytics on a converged Oracle 26AI platform can reduce average analytical query resolution time by approximately 60 percent relative to traditional BI report cycles, achieve semantic retrieval precision above 90 percent for enterprise knowledge corpora, and reduce generative output hallucination rates by more than half when grounding is enforced through in-database retrieval. The paper concludes with a discussion of governance obligations, technical limitations, and future research directions for AI-native enterprise data platforms.

Generative AI in Higher Education: The Good, the Bad, and the Ugly

Evaluating the Accuracy, Empathy, and Readability of Generative AI Versus Registered Nurses in Discharge Planning: A Vignette‐Based Study

While artificial intelligence produced more comprehensive information, it generated clinically unsafe content and scored significantly lower than nurses in empathy and readability, and healthcare organizations should establish protocols requiring nurse verification of all artificial intelligence-generated discharge content.

Comprehensive Analysis of Highly Efficient, High-Gain, Modular and Scalable, Dual-Linear-Polarized Metalens Antenna for SatCom Applications

Metalens antennas are emerging as promising candidates for compact high-gain antenna systems in next-generation wireless and satellite communication applications, where stringent link-budget requirements demand highly directive yet lightweight and compact apertures. In such systems, improving aperture efficiency is critical because it reduces the physical aperture size required to achieve a target gain, thereby enabling antenna miniaturization. Although recent studies have explored generative artificial intelligence (AI) techniques for unit-cell optimization, realizing compact and efficient metalens antennas requires a broader metalens-system-level design approach. In this work, a modular, scalable, dual-linearly polarized metalens antenna is proposed through a systematic investigation of the key parameters governing aperture efficiency, thereby enhancing overall antenna compactness. Unlike prior works primarily focused on unit-cell optimization, the proposed approach jointly optimizes both the feed antenna and the metalens structure to achieve efficient aperture illumination and reduced effective aperture requirements. In particular, the study investigates: 1) unit-cell topology; 2) amplitude thresholding; 3) number of metal layers; 4) interlayer pattern variation; 5) unit-cell dimensions; and 6) spatial placement of unit cells based on feed characteristics. This holistic optimization significantly improves aperture efficiency, enabling high-gain performance with a comparatively smaller aperture. To support practical deployment, a modular architecture based on standard printed circuit board (PCB) panels is introduced, enabling scalable and low-cost fabrication with precise alignment achieved using 3-D-printed fixtures. The proposed design is experimentally validated using a $0.7\times 0.7$ m X-band prototype, achieving a maximum measured gain of 36.3 dBi and a high aperture efficiency of 60.2%. These results demonstrate the potential of the proposed approach for compact, high-gain, and cost-effective satellite communication ground-station systems.