Machine learning predictive models: limits, condition and application possibilities in educational context

According to G. Floridi (2023), the spread of artificial intelligence in everyday life contexts determines, on a general level, a process of "wrapping" the real world around the virtual world. Based on this process "è l’ambiente ad essere progettato in modo da essere compatibile con i robot, non il contrario…avvolgiamo microambienti attorno a robot semplici per adattarli ad essi” (Floridi, 2023, p. 56). Machine Learning represents the form of artificial intelligence specifically oriented towards the management of learning processes, and its dynamisms (Binary Classification, Multiple Classification, Clusterization, etc.) can be considered as the anchor point around which teaching contexts are increasingly "wrapped". The purpose of this reflection is to place under observation the ML procedures with the greatest predictive impact, with the aim of verifying their conditions, limits and adaptive possibilities in order to the variables constituting the teaching-learning processes.