Conference Program

The growing presence of digital technologies in educational contexts raises critical questions about the pedagogical conditions through which they are meaningfully integrated into classroom practices. In particular, the potential introduction of AI-based conversational agents in school settings calls for reflection on how their use can be structured so as to support students’ comprehension processes and strategic engagement with text without reducing learning to exclusively individual interactions.

This contribution presents an instructional intervention developed across multiple fourth-grade primary school classrooms aimed at supporting students’ text comprehension through a structured pedagogical cycle combining individual work on the text, collective discussion, and metacognitive reflection. The intervention draws on research on reading comprehension and on instructional approaches that emphasise guided mediation of students’ understanding during reading (Kintsch, 1998; Palincsar & Brown, 1984; Lumbelli, 2009).

Within this framework, the pedagogical cycle was implemented across all experimental conditions. In one of the conditions, interaction with an AI-based conversational agent was introduced as a specific modality of mediation during the initial phase of individual work on the text. Through guided prompts, students were supported in making explicit use of comprehension strategies such as prediction, clarification, questioning, inference, and summarisation. The outcomes of individual work were subsequently re-examined in whole-class discussions, in which students’ interpretations were treated as shared resources for collective meaning-making. This process was intended to situate individual comprehension within a dialogic learning context, allowing students to compare perspectives and progressively refine their understanding.

A further component of the instructional cycle involved structured metacognitive reflection, supported by a self-reflection tool designed to foster awareness of reading strategies and regulation of comprehension processes.

The intervention is framed within a quasi-experimental design involving four instructional conditions implemented in parallel classroom contexts, including a control group following regular instructional practices. This contribution explores the pedagogical conditions under which the integration of AI-based educational tools can be meaningfully situated within instructional practices that sustain the social dimension of learning. In particular, it examines how individual interaction with AI-based tools, when embedded in mediated instructional practices, may be articulated with collective interpretive processes in classroom contexts.