This contribution investigates the role of generative Artificial Intelligence in formative feedback processes in mathematics education, with particular reference to pre-service teacher education. The study interprets an Artificial Agent, developed through ChatGPT 5.2, as a possible dynamic semiotic mediator in light of the Theory of Semiotic Mediation and the paradigm of Formative Assessment. The experimentation was carried out with 300 students enrolled in the Primary Education Sciences degree programme at the University of Bari Aldo Moro, who worked in pairs on the solution of an open-ended mathematical problem concerning the equi-extension of parts of a rectangle. The activity was structured into three phases: the production of an initial solution, interaction with the Artificial Agent through a structured prompt aimed at generating non-solution-providing feedback, and a subsequent collective discussion. The qualitative analysis of 150 protocols made it possible to observe the evolution of students’ responses before and after the interaction with AI, with particular attention to solution strategies, the quality of argumentation, and processes of cognitive and metacognitive self-regulation. The results show a tendency to move from local, incomplete, or poorly argued responses towards broader, more structured, and more generalised responses, especially in cases characterised by extended dialogic interactions, reconsideration, or contestation. At the same time, some critical issues emerge, related to possible misleading feedback, overly verbose outputs, and the risk of a mechanical use of the tool. These aspects, when appropriately discussed and mediated by the teacher, may become formative opportunities for developing a critical and informed use of AI. The study therefore highlights that the Artificial Agent cannot replace the role of the teacher, but may constitute a significant resource for supporting revision, argumentation, and the shared construction of mathematical meanings.
Formative feedback and artificial intelligence: new perspectives for pre-service teacher education
Fiorentino M. G.;Montone A.;Romito R.;Tosques C.
2026-01-01
Abstract
This contribution investigates the role of generative Artificial Intelligence in formative feedback processes in mathematics education, with particular reference to pre-service teacher education. The study interprets an Artificial Agent, developed through ChatGPT 5.2, as a possible dynamic semiotic mediator in light of the Theory of Semiotic Mediation and the paradigm of Formative Assessment. The experimentation was carried out with 300 students enrolled in the Primary Education Sciences degree programme at the University of Bari Aldo Moro, who worked in pairs on the solution of an open-ended mathematical problem concerning the equi-extension of parts of a rectangle. The activity was structured into three phases: the production of an initial solution, interaction with the Artificial Agent through a structured prompt aimed at generating non-solution-providing feedback, and a subsequent collective discussion. The qualitative analysis of 150 protocols made it possible to observe the evolution of students’ responses before and after the interaction with AI, with particular attention to solution strategies, the quality of argumentation, and processes of cognitive and metacognitive self-regulation. The results show a tendency to move from local, incomplete, or poorly argued responses towards broader, more structured, and more generalised responses, especially in cases characterised by extended dialogic interactions, reconsideration, or contestation. At the same time, some critical issues emerge, related to possible misleading feedback, overly verbose outputs, and the risk of a mechanical use of the tool. These aspects, when appropriately discussed and mediated by the teacher, may become formative opportunities for developing a critical and informed use of AI. The study therefore highlights that the Artificial Agent cannot replace the role of the teacher, but may constitute a significant resource for supporting revision, argumentation, and the shared construction of mathematical meanings.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
