Emotion-Related Pedagogical Agent: A Systematic Literature Review

Abstract

The purpose of this systematic literature review (SLR) is to examine more deeply the various learning interventions provided by pedagogical agents and their relation to emotional aspects. A pedagogical agent is a software agent that provides guidance, feedback, or intervention to learners in digital environments. Pedagogical agents have the potential to address issues in computer-based learning, particularly online learning, which often neglects affective aspects, such as the emotions of its users. Most research on pedagogical agents has focused on their visual design, types of feedback, and other empirical aspects. However, this research has not examined what underlies these agents' ability to provide interventions personalized to learner's emotions. This paper explores the extent to which pedagogical agents have addressed learners' emotional needs. It also identifies opportunities and challenges for further research on interventions by pedagogical agents personalized to learners' emotional states. The study's research questions include: 1) To what extent does research exist on models, frameworks, or architectures for pedagogical agents, especially those related to emotions? 2) How are pedagogical agents represented, what types of interventions do they use, and how do these interventions affect learners' emotions? 3) What kinds of inputs are used to activate the pedagogical agent's functions? This SLR applied the Kitchenham method to select reference sources from 2013 to 2023 and was indexed by Scopus in the Q1 to Q4 range. Our review revealed the absence of a specific model for mapping out interventions tailored to the learner's emotional needs. Most existing pedagogical agents provide learning interventions that are less adaptive and personalized based on the learner's emotional state and are applied to asynchronous learning systems such as e-learning. There are still very few pedagogical agents that use real-time input technology by utilizing artificial intelligence to recognize the emotional state of learners so that they can trigger an adaptive and personalized intervention.