Conference Program

Within the platform society, the silent risk arising from the delegation of human decision-making to algorithmic systems manifests itself in a progressive erosion of human agency in relation to the automatism of algorithmic agency. In this context, it becomes essential to cultivate, from early childhood onward, a critical competence oriented toward understanding how algorithms function. Educational robotics, AI platforms, AI-powered toys, and conversational agents represent, from this perspective, privileged pedagogical devices, as they enable children to recognize the algorithmic influences that anticipate and shape action and decision-making processes. This, in turn, supports the protection of those “micro-spaces of attentional autonomy” (Lazzaroli, 2025), which are fundamental for counteracting the risk of algorithmic discrimination. Such a risk may undermine collective memory, which is grounded in democratic ideals and values, thereby contributing to the perpetuation of anti-democratic political and cultural orientations, as well as racist and gender stereotypes. These mechanisms influence the presentation of online information and intersect with anti-democratic and nationalist dynamics aimed at erasing the identities and sovereignty of peoples and nations.

Several studies have shown that the data used to train algorithms are shaped by historical constructions and deeply rooted social practices, which contribute to the systematic reproduction of inequalities (Buolamwini & Gebru, 2018; Noble, 2018). Applications such as ChatGPT, for instance, do not allow users to clearly identify the criteria employed in the training process of algorithms, nor to reconstruct the sources from which responses are generated. This limited transparency may foster the reproduction of stereotypes and biases, as well as the phenomenon of so-called “hallucinations”, namely the generation of responses based on an incorrect interpretation of the user’s request. Consequently, the opacity of the processes through which responses are produced makes it difficult to assess their reliability and limits users’ ability to exercise critical oversight over the generated content (Bordogna & Rubbia, 2024). An additional risk is represented by deepfakes—manipulated images and videos that can be used to distort reality and influence public opinion (Balafrej & Dahmane, 2024). Their detection is particularly challenging due to the high degree of realism with which they reproduce faces and voices, thereby constituting a significant threat to information integrity and democratic processes.

Adopting a pedagogical approach grounded in experiential and cooperative learning, and acknowledging the educational value of play in early childhood development (Montessori, 1953), this paper aims to critically examine these processes. In doing so, it conceptualizes AI-based educational technologies both as tools for the pedagogical “debugging” of algorithmic automatism and as formative levers for strengthening critical agency within contemporary educational contexts.