
Dyslexia is a specific learning difficulty that affects reading accuracy, fluency, and comprehension, as well as writing skills; thus, it negatively impacts academic performance and educational inclusion. In this context, educational mobile applications have emerged as a technological alternative to support literacy processes by providing interactive environments adaptable to students' needs. The objective of this study was to compare the performance and usability of three mobile applications designed to strengthen literacy skills among students with dyslexia: ABCDinos, Reader Pro, and Galexia. Research was conducted using an iterative methodological approach, which enabled the progressive evaluation of criteria related to accessibility, usability, motivation, and support for the development of reading and writing skills. Results show that these three applications positively contribute to literacy development. ABCDinos stands out for its playful approach, Reader Pro for its level of personalization, and Galexia for the integration of visual and auditory resources. It is concluded that educational mobile applications are a relevant support tool for the inclusion of students with dyslexia in digital learning environments.
The present article addresses the control and organizational structures prevalent in Micro, Small, and Medium Enterprises (MSMEs) within the Colombian construction sector. The PRISMA methodology was implemented, which involved a rigorous information search and systematization process. This process aimed to identify and analyze research and literature related to corporate control and organization, progressing from the macro- to the micro-level, specifically focusing on the construction sector's MSMEs in the country. A systematization matrix was developed to categorize documents containing relevant information on control and organizational forms across various sectors, ultimately pinpointing the specific management of MSMEs in construction, particularly concerning personnel and budget aspects. Finally, a data analysis was performed, leading to the conclusion that the implementation of control and management strategies is critically important for optimizing internal processes. These strategies must be guided by technological advancements and emerging national regulatory frameworks.
The Mesa de Los Santos in Santander, Colombia, is a region affected by water scarcity and supply limitations, where groundwater has become a key alternative resource. Groundwater exploration in the area has been primarily conducted using electrical resistivity tomography (ERT), a geophysical method aimed at characterizing subsurface resistivity contrasts associated with hydrogeological features. More recently, machine-learning approaches, particularly neural networks, have been used as complementary tools to predict porosity values from geophysical and geological variables and support the assessment of hydrogeological characteristics of the Mesa de Los Santos. This work presents porosity estimates for the region obtained via two approaches: (i) from ERT; (ii) from rock-sample data to produce new porosity maps. Both approaches include a bootstrapping analysis to compute mean and standard deviation values, allowing us to identify areas where the models show greater or lesser agreement. To train the neural-network models, existing porosity and electrical-resistivity measurements from 358 core plugs were used. The dataset was balanced by lithology, and additive noise (5%) was applied to the measurements. The implementation was carried out with the scikit-learn library in the Google Colab environment. A total of 100 neural network models, with a hidden-layer architecture of 28-14-7-3, were trained to estimate porosity from electrical resistivity values. The process required approximately 24 min for achieving a minimum training score of 0.7 and an average test score of 0.28. For the generation of porosity maps in the MLS, 200 neural network models were trained using two architectures: 39-26-13-7-3 and 26-13-7-3, with a total training time of approximately 58 min. In both approaches, the analysis of standard deviations provided an additional criterion to assess the reliability of porosity estimates. Furthermore, results are consistent with the geological context and enable identifying potential hydrogeological target areas. The integration of neural-network results with previous studies made it possible to identify and categorize hydrogeologically significant areas, thus providing useful input for decision-making and groundwater management in the region.
The transnational deployment of autonomous vehicles faces serious obstacles when applied to the high-entropy traffic ecosystems of Latin America. Standard algorithmic models, based on the structured predictability of the Global North, falter in the face of regional infrastructure deterioration and heterogeneous driver behavior. Through a systematic approach guided by PRISMA, this research reviews existing literature to delineate the necessary behavioral adaptation parameters. Thirteen primary studies out of an initial corpus of 254 documents met the inclusion and exclusion criteria and passed quality assessment using the Parsifal tool, thus revealing a scarcity of empirical evidence specific to the region. Analysis of these studies suggests that maneuvers traditionally classified in the global literature as "reckless driving" constitute, in practice, "functional aggressiveness"-where adaptation heuristics and space negotiation are necessary to navigate congested and non-standardized roads. Consequently, it is determined that imported navigation systems run the risk of misclassifying these local maneuvers as safety anomalies or false positives caused by road deterioration, introducing latent operational risks and paralyzing the flow of mixed traffic. The study concludes that the direct and uncritical transfer of technology is operationally and socially unsustainable. Instead, it advocates for a strategy of algorithmic "tropicalization" grounded in data sovereignty, which requires on-site data collection through ubiquitous mobile sensing, recalibrating multi-agent simulations to traditional infrastructure, and redesigning human-machine interactions. This multidimensional, context-aware approach is a prerequisite for designing a technically safe, efficient autonomous mobility framework, socially integrated into the complex realities of Latin American cities.
The increasing complexity of machine learning models, such as deep decision trees, has limited their applicability in scenarios where transparency is crucial, particularly in multiclass classification problems. However, evidence on how XAI facilitates interpretability in deep multiclass trees remains fragmented, and evaluation practices are not consistently standardized. This study presents a systematic mapping of the literature published between 2020 and 2025 on Explainable Artificial Intelligence (XAI) approaches applied to decision trees, with an emphasis on visualization and class decomposition strategies. Using a rigorous methodological protocol, 43 relevant articles were analyzed, and a predominance of techniques such as SHAP and LIME, which rely on visualizations to explain model decisions, was identified. Strategies focused on the class approach can improve the extraction of understandable rules. However, their impact on cognitive time still requires further research. A diverse set of quantitative and qualitative metrics for validating interpretability was also identified. In general, mapping synthesizes the dominant techniques, evaluation metrics, and gaps for more user-centered evaluations to promote more transparent and useful models for decision-making in different domains.