A Method to Evaluate Virtual Opponent Based on IA for an Accessible Educational Digital Game: a Case of Study with Dinobase

Abstract

The aim of this article is to present a method to evaluate the adequacy of a virtual opponent in a digital game. The motivation for this research arose from the lack of methodologies to evaluate a game in terms of its artificial intelligence (IA) applied to a virtual opponent when confronted with its players. The methodology for this research consisted of the case study applied with 17 children - 8 with intellectual disability and 2 with hearing impairment - in a basic school in play activities with Dinobase digital game. Dinobase is a digital game developed to help children to learn about mathematics, and it was created in a universal design approach, by considering children with disabilities. The virtual opponent must identify the level of ability from children and adapted its behavior. On previous research, the virtual opponent was evaluated with personas, on this research, we use data and analysis triangulation that involved log records, observations, interviews, and video. The results show that the methodology adopted allowed to identify the needs of improvement to the virtual opponent, that were not shown just using personas.