Conference Program

The paper explores, from a theoretical and design-oriented perspective, the potential of Generative Artificial Intelligence (GAI) to support formative assessment (assessment as learning), with the aim of developing pathways that integrate teaching, assessment, and students’ self-regulation processes. Within this framework, assessment tasks are conceived not merely as tools for measuring learning outcomes but as intentionally designed opportunities for deep cognitive activation (retrieval, restructuring, and transfer of schemas) and for the development of metacognition and evaluative agency.

The proposed learning pathways are structured as sequences of progressive challenges, in which learners advance to the next task only after mastering the conceptual tools underpinning the previous one—a condition referred to as learning readiness—through forms of deliberate practice. Each stage of the training process entails the gradual construction of cognitive, strategic, and dispositional prerequisites through experimentation, debriefing, conceptualization, and the automatization and transfer of acquired knowledge and skills. The expected outcome of this pedagogical approach is the cultivation of learners capable of mobilizing increasingly complex cognitive schemas with mastery and autonomy.

The paper begins by examining the potential of classical Intelligent Tutoring Systems (ITS) and extends their possibilities in light of current GAI developments, with the goal of defining design guidelines for their application within formative assessment pathways. Specifically, it discusses how GAI systems can enable adaptive cycles of interaction between student and machine, whereby the latter proposes sequences of challenging tasks and monitors, in real time, the evolution of students’ knowledge, skills, attitudes, and strategies. These tasks are designed to engage learners in a wide range of cognitive processes—including the identification of key concepts and inconsistencies, comparison and classification, synthesis, and graphical representation—with instructional guidance provided by the system itself. Learner reflection is fostered through process narration, error analysis, the definition of quality criteria, and reasoned self-assessment of performance.

The proposal outlines three levels of integration:

1. Knowledge base design that codifies not only disciplinary content but also performance quality constraints, evaluative criteria, and observational indicators of progression in knowledge, skills, attitudes, and competences.

2. Dynamic learner modeling that tracks not only outcomes but also the evolution of underlying cognitive processes and mental dispositions relevant to learning readiness.

3. Tutoring strategies that adapt task difficulty, feedback type and timing, forms of metacognitive scaffolding, and opportunities for collaborative engagement, transforming each assessment moment into an opportunity for cognitive restructuring and internalization of self-assessment and self-regulation criteria.

In conclusion, the paper highlights key challenges and open issues, such as modeling complexity, fine-grained personalization, and ethical and privacy concerns, to guide future research toward a pedagogically grounded development of GAI-based intelligent tutoring systems.