hal-04977331 Understanding how to assess the impact of Artificial Intelligence on learning : a (…)

Since late 2022, access to Generative Artificial Intelligence (Gen-AI) such as ChatGPT, Copilot, or Gemini has become more widespread. This new phenomenon raises questions about the real impact of Gen-AI, including Large Language Models (LLM), especially on learning. Although we are facing a quite new phenomenon, a better understanding of the impact of Gen-AI on learning can be suggested by considering how other forms of Artificial Intelligence (AI) have been studied. Based on a systematic review approach, our study shows that most existing researches focus on the evaluation of AI that is dedicated to education (AIED), and do not consider AI devices designed for broader applications. Above all, the systematic review identifies more than forty indicators used in the assessments of the impact of AI on learning; indicators that can be grouped into six key dimensions: technological, learning achievement, pedagogical, interaction, affective, metacognitive. This provides a relevant framework and a solid basis for future evaluations of the impact of Gen-AI on learning.