Quirino State University

This study explores the use of generative artificial intelligence (AI) in academic research writing by employing a qualitative descriptive methodology integrated with topic modeling. Semi-structured interviews and focus group discussions were conducted with students from various academic disciplines to capture diverse experiences and perceptions of AI-assisted writing. The qualitative data were analyzed using thematic analysis, while the textual corpus underwent preprocessing procedures, including stop-word removal and stemming, prior to topic extraction through Latent Dirichlet Allocation (LDA) using RapidMiner. The results reveal that generative AI is widely perceived as a useful support tool for idea generation, improving coherence, and enhancing the overall quality of academic writing. Nevertheless, participants also identified major concerns related to academic integrity, plagiarism, fairness, algorithmic bias, and excessive reliance on AI technologies. Although generative AI was recognized for its capacity to increase efficiency and productivity, participants emphasized the importance of maintaining ethical awareness and independent critical thinking in its use. The study concludes that educational institutions should implement comprehensive measures, including AI literacy programs, ethics-oriented discussions, and clearly defined guidelines for acceptable AI use. These efforts are necessary to maximize the educational value of generative AI while minimizing its potential risks in academic research writing.

Accidental drowning is a leading cause of death among children with Autism Spectrum Disorder (ASD), making aquatic interventions a public health priority. Although clinical benefits are well documented, the literature remains fragmented, often separating child-centered physiological and behavioral outcomes from the socioecological barriers faced by caregivers and instructors. To address this gap, this narrative review synthesizes 21 peer-reviewed empirical studies published between 2017 and 2025. Using a dual-lens framework, the review examines both clinical efficacy and community-based implementation. Quantitative evidence shows that structured aquatic programs improve water competence, gross motor skills, and executive functioning. Preliminary findings suggest that these gains may be linked to physiological mechanisms, including increased vagal tone and modulation of inflammatory cytokines [IL-6 and IL-10], which are associated with better sleep regulation and fewer stereotypic behaviors. In contrast, qualitative evidence reveals a serious implementation gap. Although families value aquatic participation as a meaningful activity that reduces parental stress, equitable access remains limited by sensorychallenging pool environments, high costs, and a shortage of autism-informed instructors. Bridging this efficacyimplementation divide requires scalable policy reforms, including task-sharing models in which therapists support mainstream instructors. Larger randomized controlled trials are also needed to strengthen the evidence base and help establish aquatic therapy as an equitably accessible standard of care.

When prices rise, it is often mothers who feel the strain first. In the Philippines, the implementation of the Tax Reform for Acceleration and Inclusion (TRAIN) Law increased the cost of basic goods, placing additional pressure on those who manage household finances. This study explores the resilience of mothers in Quirino Province through qualitative narrative inquiry using in-depth interviews. Findings reveal two central themes: the challenges brought by the TRAIN Law and the coping strategies mothers employ. Motherhood emerges as a core social identity that sustains resilience, with family serving as the primary source of motivation and strength. Six indicators: survival, inspiration, social identity, diversification, strategy, and optimism, show that resilience is a dynamic, everyday process. The study emphasizes the value of mothers' narratives in guiding policies and programs that support adaptive strategies and strengthen both individual and community resilience.

This study assessed faculty satisfaction with faculty development programs (FDPs) in a state university and identified factors affecting overall satisfaction and career mobility. An explanatory sequential mixed-methods design was used. Primary data were analyzed through descriptive statistics, t-test, one-way ANOVA, Pearson correlation, and multiple regression, while qualitative responses underwent thematic analysis. Secondary data were examined using Logistic Regression and Random Forest in Google Colab to predict occupational mobility. Results showed that faculty were generally satisfied with FDPs, especially in relevance and alignment, content and delivery quality, and perceived impact. Accessibility and institutional support received lower ratings. Multiple regression indicated that only institutional support and perceived impact significantly predicted overall satisfaction. Qualitative results showed that faculty valued relevant and practical training but suggested stronger administrative support, improved accessibility, more specialized workshops, and workload-sensitive implementation. For the predictive models, Logistic Regression achieved 90.87% accuracy, while Random Forest reached 100%. Both models identified job satisfaction, interest in career change, and salary as the strongest predictors of occupational mobility, with Random Forest ranking job satisfaction as the most influential. The study concludes that FDPs are most effective when relevant, impactful, and backed by strong institutional support. It also suggests that faculty satisfaction is a key link between professional development and career stability. Integrating primary institutional and secondary predictive data offers a broader view of how faculty development shapes satisfaction and long-term professional outcomes.

Guava disease detection remains an important task in precision agriculture because visual symptoms on leaves may vary in color, texture, shape, and severity, making manual diagnosis prone to delay and subjectivity. This study developed a dual-backbone attention-enhanced deep learning framework for guava disease detection using leaf image classification. The proposed model, GuavaXAF-Net, integrates MobileNetV3Small and EfficientNetV2B0 as complementary feature-extraction branches. To improve disease-specific representation learning, the framework incorporates Efficient Channel Attention, spatial attention, cross-backbone feature interaction, and self-adaptive feature fusion. The model was trained using a three-stage fine-tuning strategy consisting of classifier head training, MobileNetV3Small fine-tuning, and EfficientNetV2B0 finetuning. The dataset consisted of three guava leaf classes: Anthracnose, fruit fly, and healthy guava. Experimental results showed that the proposed model achieved a test accuracy of 99.74% and a test loss of 0.0033 on 382 independent test images. The classification report produced near-perfect precision, recall, and F1-score values across all classes. The confusion matrix showed that all Anthracnose and fruit fly samples were correctly classified, while only one healthy guava image was misclassified as a fruit fly. In terms of efficiency, the model contained $7,401,877$ parameters, had a file size of 51.59 MB, and achieved an inference time of 3.19 ms per image, equivalent to 313.52 frames per second. These findings indicate that the proposed dual-backbone attentionenhanced framework is accurate and computationally efficient for guava disease classification. The model has potential applications in automated crop monitoring, mobilebased plant disease diagnosis, and intelligent agricultural decision-support systems.