research lines
Some information about our work in the research lines.
Artificial Intelligence in Education
Artificial Intelligence in Education (AIED) is an interdisciplinary field that leverages AI technologies to enhance teaching and learning processes. It integrates principles from computer science, cognitive science, education, and artificial intelligence to design tools and systems that support students and educators in diverse educational contexts. Some key aspects of the AIED field explored in our research include:
Personalized Learning: Adaptive learning environments that adjust to the needs of individual students. These systems analyze data from students’ interactions and performance to customize content and learning paths, ensuring that each learner receives the appropriate level of challenge and support.
Automated Assessment and Feedback: AI can assist in grading and providing real-time feedback to students, thus reducing the burden on teachers and delivering faster, more personalized responses to learners. Additionally, Automated Question Generation (AQG) can be employed to improve educational assessments by generating relevant questions for evaluation.
Learning Analytics: AI-driven analysis of student data (e.g., test scores, question responses) to predict future performance, identify at-risk students, detect problematic questions, and provide valuable insights to improve teaching strategies and student outcomes.
Generative AI for Hybrid Intelligence: The application of generative AI systems, such as large language models, to support teaching and learning to enhance human capabilities. This line of research explores how AI can collaborate with teachers and students to co-create personalized content, generate engaging learning materials, facilitate creative problem-solving, and foster hybrid intelligence, where human and generative artificial intelligence work together to maximize educational outcomes.
Data Mining and Machine Learning
One of the main approaches to extract information from large and complex datasets using artificial intelligence is the application of data mining processes and machine learning methods. Traditional analytical tools often fail to uncover underlying patterns, relationships and coherences. Data mining refers to the systematic process of discovering meaningful structures, trends, and correlations within large volumes of data. Within this process, machine learning plays a central role by providing algorithms capable of learning from data, identifying patterns, and making predictions or decisions without being explicitly programmed. Together, these techniques enable intelligent analysis that adapts and improves over time as new data becomes available.
The team has been actively involved in these fields since before the year 2000. In addition to applying state-of-the-art machine learning techniques—such as ensemble methods, online learning, decision trees, or Markov models—we also develop our own algorithms tailored to specific challenges. These approaches are combined with efficient data mining processes and applied to real-world problems across a variety of domains. Current applications include urban sustainability, nature-based solutions, renewable energy systems, and the detection of suicidal behavior, among others. All of these initiatives are supported by a robust data infrastructure and advanced engineering methodologies that enable the effective extraction, transformation, and integration of data from heterogeneous sources.
Keywords: Learning Algorithms, Supervised and Unsupervised Learning, Predictive Modeling, Data-driven Decision Making
Agentic AI, Multiagent Systems and Intelligent Autonomous Agents
In the field of Artificial Intelligence, one of the strategies for solving complex problems is based on the use of distributed components endowed with intelligence, called agents, which in turn interact with a common environment to achieve common or conflicting objectives, giving rise to multi-agent systems. These systems, whose origins date back to the 1970s, decompose complex problems into smaller, more specific parts that perform specific tasks autonomously and intelligently. In recent years, the advent of language modeling (LLM) has transformed classical multi-agent systems into what is known as Agentic AI, where agents become more sophisticated, autonomous and strategic in their decision making thanks to this type of models.
Our group has been involved in research on multi-agent systems since the 1990s. In this line, we are currently working on projects to implement digital twins for crime, mobility and sustainability in the urban environment, combining Agentic AI with other lines of AI such as multi-agent reinforcement learning or deep learning models, and with data engineering techniques for intelligent data extraction and fusion.
Keywords: Distributed AI, Agent-Based Modeling and Simulation, Multiagent Reinforcement Learning, LLM Agents, Large Action Models
Intelligent Analysis of Socio-economic and Sports Data
Smart data analysis, also known as advanced analytics, is an essential tool for transforming large volumes of information into actionable knowledge. Intelligent data analysis allows moving from data to knowledge, and from knowledge to informed action. In an environment with limited resources, high competitiveness or uncertainty (such as in economics or sport), this capability makes the difference between intuition and strategy. This quantitative approach makes it possible to identify patterns that are not obvious to the naked eye and make evidence-based decisions to improve equity, emotion and overall performance.
But even more important for any field of activity, is to integrate and link data to social, economic or geographic characteristics. This is where the integration of socio-economic data brings essential value to the interpretation and decision support of the knowledge extraction process.
By applying optimization, data mining or simulation techniques on complex structures, a deep and explanatory vision of the analyzed phenomenon is obtained. This allows not only to describe what happens, but to understand why it happens and to anticipate what might happen under other scenarios.
The study of sports data incorporates the modeling of areas with data from a discrete space, such as sports competitions - such as allocations, classifications or binary results - with increasingly complex structures, where the progressive incorporation of information on social impact, audiences or bets must be highlighted and the use of combinatorial models, integer programming and heuristic methods allows complex configurations to be explored with high mathematical rigor.