CheMate: Anthropomorphic-cues-Mediated Experiential Learning Game Using Generative AI

Abstract

Anthropomorphic design cues (ADCs) refer to giving non-human objects human-like characteristics, widely used in human-computer interaction (HCI) and education. However, the current interactive modes of anthropomorphized knowledge roles (KRs) are relatively monotonous, with insufficient educational content integration, which may weaken user interaction and learning experiences. In this paper, we propose the Anthropomorphic-cues-Mediated Experiential Learning Game (AMELG) framework and implement a chemistry puzzle game called CheMate. In the framework, ADCs serve as intermediary signs, and KRs engage in interactions, providing feedback on player’s actions. Subsequently, players reflectively observe the experiential process, associate abstract conceptualization with existing knowledge, understand new knowledge conveyed through ADCs, actively apply the knowledge, and thereby better memorize, understand, and master it. In the game, atoms KRs with different personalities can gather into molecules. These KRs are driven by large language models (LLMs) for dialogues. First, players talk with various KRs, accumulate intimacy, and gain the ability to move them. Then, players engage in searching, reasoning, and chemical reactions. The game dramatizes scenarios of chemical reactions. Players improve conditions or mediate through KR dialogues to drive the reactions. We set each reaction as a process of atoms’ growth and provide the player with a retrospective review of it. A pilot study was conducted to validate the effectiveness of the framework, indicating its potential to enhance interactive experiences and learning outcomes. The article briefly summarizes and analyzes effective ADCs design methods in the game, and provides insights into the HCI modes of ADCs and their integration with educational content.