Designing Virtual Agents for Simulation-Based Learning in Virtual Reality

Introduction

Defined by their core functions, virtual agents in a multimedia or virtual learning environment are typically: (a) pedagogical agents—graphical characters presented on a computer screen to support or guide student learning of the presented material (Heidig & Clarebout, 2011; Makransky, Wismer, & Mayer, 2019; Wang, Li, Mayer, & Liu, 2018); (b) embodied conversational agents—human-like characters that are specifically conversational (both verbal and non-verbal) in their behaviors and present the similar properties as humans in face-to-face conversations (Cassell, Sullivan, Churchill, & Prevost, 2000); and (c) interactive simulators that simulate not only physical behaviors but also cognitive functions (e.g., reasoning, decision-making, and learning) of a subject or person that is interacted with, such as a virtual patient- or a student-simulator (Kleinert et al., 2015; Matsuda, Cohen, & Koedinger, 2015).

Motivation of using and designing virtual agents, sometimes called virtual humans, in a digital learning environment is to present an active, cost-efficient, and personalized teaching/learning experience. Specifically, virtual agents have been used to (a) motivate and scaffold the processing, retention, and application of the presented information, (b) enable computer-assisted personalized learning, (c) simulate the behaviors of a system in reaction to user inputs to frame a situated learning experience or to elicit and foster targeted cognitive functions, and (d) facilitate the authoring or assessment of a competency model. Prior research examining the effects and conditions of using virtual agents indicated a great variety of agents used and hence inconclusive findings about their effectiveness, even though recent meta-analyses started to show a small but significant effect of using a pedagogical agent on learning (Heidig & Clarebout, 2011; Schroeder, Adesope, & Gilbert, 2013; Wang et al., 2018). Rather, the previous studies and reviews of virtual agents argue that it is more meaningful to question the design and targeted function of virtual agents in relation to the properties or characteristics of learning, learner, and implementation contexts or conditions.

In this chapter, we aims to review and explore propositions, approaches, and core facets of designing virtual agents that contextualize and scaffold simulation-based active learning. Virtual agents can be considered as a major pedagogical design element in the virtual learning environment. They can be used to improve learning interactivity and engagement by providing adaptive learning feedback and other personalized learning services (e.g., mentoring and guidance). Virtual agents can be the middle point between the learner and the virtual environment for achieving personalization, by providing personalized learning services based on a stored learner or user profile. They are also considered effective to improve the sense of presence in the immersive virtual learning environments and improve motivation. They are the key to creating a continuous virtual presence for users within a virtual world. Prior research indicated that the feeling of co-presence fostered by virtual agents can complement user-controlled avatars to achieve an ever-present virtual presence (Gerhard, Moore, & Hobbs, 2001).

Despite the increasing interest and existing studies on the usage and design of virtual agents, open questions remain governing the dynamic interaction and convergence among the targeted agent-based learning, learner characteristics, affordances of the technological platform, and the implementation contexts. Conceptual or design research outlining the heuristics of embodied and potentially intelligent agents in the virtual environment is still lacking. The constraints and opportunities presented by the current virtual environments for agent-based learning, such as virtual reality and game-based learning, warrant a critical and contextualized review and discussion.