Our proposal is based on the assumption that the pervasive presence of artificial intelligence in the daily lives of younger generations may lead to a sort of “naturalization” and “humanization” of automated communicative phenomena, whether productive or reproductive. This phenomenon risks eroding essential spaces for critical data processing, primarily due to the difficulty in deciphering the underlying processes behind AI-generated outputs, especially when they present themselves in linguistic forms that are formally identical to those used in human interaction. In this regard, the relationship between AI and social sciences can be framed along two key directions: on the one hand, the usefulness of integrating generative AI models within the teaching of historical and social disciplines; on the other, the contribution that social sciences can provide in helping students develop a critical approach and conscious use of such models.

The second direction guided an experimental initiative conducted during the 2024/25 academic year with students from seven high school classes at Liceo Scientifico P. S. Mancini in Avellino (Italy). This initiative was part of an interdisciplinary project integrating philosophical and mathematical thinking.

Objectives of the Experiment:
The core idea was to compare human intelligence with artificial intelligence to highlight how seemingly identical phenomena result from radically different processes. It was necessary to explore the cognitive processes underlying human behaviors that can be defined as “intelligent” and the algorithmic mechanisms that generate similar responses in Large Language Models (LLMs).
The chosen method was a semi-structured lesson, which, given the educational goals, allowed for the emergence and subsequent problematization of students’ implicit conceptions regarding the functioning of different AI models. Additionally, this method helped address misconceptions or anthropomorphic perceptions students may have about AI. The aim was to explicitly highlight the various and often divergent mechanisms underlying human and artificial data processing and response generation.

Methods and Tools:
The interdisciplinary approach was fundamental in this experiment as integrating different scientific perspectives to interpret complex phenomena and, in an epistemological sense, as an exploration of the conceptual and procedural structures through which disciplines construct their respective objects of study.

The disciplines involved included:
- Psychology and Pedagogy: To analyze different conceptions of intelligence and learning and define human cognitive processes.
- Argumentative Logic: To identify fallacies in specific reasoning patterns.
- Linguistics: To examine the difference between the syntactic and semantic levels of communication.

The AI models used primarily included ChatGPT (free version) and other applications commonly used autonomously by students to assist comprehension and task execution. The interactions with ChatGPT were strictly verbal, avoiding multimedia content generation. Additionally, the study examined the impact of recommendation systems, which influence users through social media, media streaming platforms, and online consumer markets, often without their explicit awareness.

Keywords: Education, Natural Intelligence and Artificial Intelligence.