Generative Artificial Intelligence and Videogame Analysis for Game-Based Learning: A Comparison between Automated Assessment and Expert Evaluation

Abstract

This study focuses specifically on Game-Based Learning (GBL), with the aim of examining how generative AI can assist teachers in designing educational activities that incorporate the pedagogical use of videogames not originally intended for educational purposes. Within a more general model which includes a game analysis from a pedagogical point of view and, subsequently, the design of learning activities, in this paper we compare the results of a human-based analysis whit an AI-based one, both referring to a specific framework involving three main dimensions: directivity, sociality and modifiability. We chose 8 games varying by notoriety, type and recency; we analyzed them according to our framework and then made them analyze by generative AI, trained with the same framework, and then compared them one with each other. We found that in most renowned games, the AI adopts a more technical or slang language. AI-generated answers also tend to go beyond the focus led by the prompt giving suggestion about instructional design. The study also helped to refine our theoretical framework by comparing its human interpretation with another one, AI-generated, namely about the dimension of modifiability.