Discover our new board game!
Empowering education through Data Science
Data Science Lab for Teaching Innovation (DSLAB-TI) is a teaching innovation group whose main mission is to promote the use of Data Science techniques and methods to enhance educational practices. Using data science methodologies, student data is collected through online questionnaires to build machine learning models capable of accurately predicting students’ future evaluations based on their partial results. These models help identify specific student profiles, enabling targeted actions to improve both teaching and learning processes. The results and models are presented through customized dashboards tailored to the needs of the end user. Models will be deployed in production by both instructors and students in courses following those used for training.
The proposed innovation actions are fully replicable, as the techniques and tools developed can be applied across all undergraduate and graduate courses offered at URJC. Instructors would only need to design course-specific questionnaires and access results and predictions via dedicated dashboards. These actions require a broad timeframe (at least two academic years) for full implementation.
Strategic objectives
Key action lines
Teaching innovation projects
ViLT: virtual intelligent tutor for supporting URJC students
2024 — 2025
ViLT proposes the development of an intelligent tutoring system based on AI and LLMs to support students at URJC. It integrates a dynamic chatbot, generation of embeddings from course materials, and a feedback module to detect conceptual drift. The system uses a modular architecture via a REST API, follows agile development, and includes dashboards for instructors. Key technologies include NLP, document processing, and interactive visual analytics.
Explore the code Get the INTED2025 slidesDSExams: massive and automated generation of multipurpose randomized questionnaires
2022 — 2023
DSExams aims to build an automated platform for creating randomized quizzes using the R
package {exams}, formatted with LaTeX. It allows infinite variations of
questions using probabilistic methods and provides detailed feedback for self-assessment.
The quizzes integrate with Moodle and are managed through a private GitHub repo. A public R
package is planned, supporting filters like student profile, topic, and language.
Featured publications
Using an LLM-based framework to analyze student performance
A. Fernández-Isabel, C. Lancho, I. Martín de Diego, Á. Udías, A. Alonso-Ayuso, C. Alfaro, E. L. Cano, F. Ortega, J. Gómez, J. M. Moguerza, M. J. Algar
The work explores a framework powered by LLMs to support educational mentoring and tutoring, especially for students less likely to seek face-to-face help. The system is designed to offer contextualized, AI-based academic support.
Read articleChatGPT’s performance in university admissions tests in mathematics
Á. Udías, A. Alonso-Ayuso, C. Alfaro, M. J. Algar, M. Cuesta, A. Fernández-Isabel, J. Gómez, C. Lancho, E. L. Cano, I. Martín de Diego, F. Ortega
This study evaluates ChatGPT-4.0 on Spanish university entrance exams in math. Results show strengths in probability and statistics, while identifying weaknesses in algebra. The paper suggests careful integration of LLMs in educational assessment.
Read articleTutor virtual inteligente basado en modelos generativos del lenguaje
A. Fernández-Isabel, I. Martín de Diego, E. L. Cano, M. Cuesta, C. Lancho
This system introduces a fine-tuned LLM-based tutor that provides interactive support to students via a web interface. By adapting the model to specific subjects using instructor-provided materials, it delivers personalized explanations, exercises, and continuous academic assistance.
Book details Download bookDSExams: generación masiva y automatizada de cuestionarios aleatorios multipropósito
E. L. Cano, M. Cuesta, C. Lancho, C. Alfaro, M. J. Algar, A. Alonso-Ayuso, A. Fernández-Isabel, J. Gómez, I. Martín de Diego, J. M. Moguerza, F. Ortega, Á. Udías
DSExams is a framework for generating dynamic, randomized questionnaires to enhance assessment and self-evaluation in Data Science education. Built with R and LaTeX, it enables scalable, customizable exercises for subjects like Statistics and AI, fully compatible with platforms like Moodle.
Book details Download bookEl potencial de los modelos grandes de lenguaje para mejorar el aprendizaje de probabilidad: un estudio sobre ChatGPT3.5 y estudiantes de primer año de ingeniería informática
Á. Udías, A. Alonso Ayuso, I. Sánchez, S. Hernández, M. E. Castellanos, R. M. Diez, E. L. Cano
This study analyzes ChatGPT's ability to solve probability exam problems, showing it outperforms average students in reasoning and clarity. Despite struggles with basic math, its explanations and R-script outputs prove useful in educational contexts.
Book details Download bookEmpowering Academic Performance: Data-Driven Mentoring for Personalized Education
M. Cuesta, C. Lancho, I. Martín de Diego, A. Fernández-Isabel, E. L. Cano, J. M. Moguerza
This paper introduces a mentoring framework powered by learning analytics, using student questionnaires and machine learning to identify mentoring needs and optimize teaching strategies based on personalized profiles and performance forecasts.
Read articleDSExams: Massive and automated generation of randomized multipurpose questionnaires
E. L. Cano, M. Cuesta, C. Lancho, C. Alfaro, M. J. Algar, A. Alonso-Ayuso, A. Fernández-Isabel, J. Gómez, I. Martín de Diego, J. M. Moguerza, F. Ortega, Á. Udías
DSExams proposes a scalable solution to generate automated questionnaires for assessment and self-evaluation in Data Science education, leveraging randomization and digital tools to enhance learning outcomes and reduce repetitive study strategies.
Read article Browse the slidesDSGame Kids: Learning Data Science projects through a storytelling board game
A. Fernández-Isabel, I. Martín de Diego, M. Cuesta, C. Lancho, J. M. Moguerza
DSGame Kids is a physical board game designed to teach the data science project lifecycle to computer science students in a collaborative and interactive way. It replaces digital isolation with teamwork, using storytelling and mission-based progression.
Read articleConcurso de Monty Hall: una aplicación interactiva con R para explicar probabilidad
E. L. Cano
Through a playful approach, this Shiny application explains conditional probability by simulating the Monty Hall problem. Students explore outcomes through interaction and gain better insights into probabilistic reasoning and decision-making.
Launch appMejorando la comprensión de conceptos estadísticos mediante aplicaciones interactivas innovadoras
E. L. Cano, María Jesús Algar, A. Alonso-Ayuso, J. M. Moguerza, F. Ortega
This work presents innovative Shiny apps to help students grasp abstract statistical concepts by visualizing simulations, enhancing engagement and motivation in STEM and social sciences. It highlights the use of R and Shiny for creating browser-based interactive content to replace traditional teaching methods.
Launch app