Paper Summary
Share...

Direct link:

Investigating the Influence of Pedagogical Agents on Learners’ Motivation: A Comprehensive Meta-Analysis

Thu, April 24, 9:50 to 11:20am MDT (9:50 to 11:20am MDT), The Colorado Convention Center, Floor: Terrace Level, Bluebird Ballroom Room 2C

Abstract

Objectives and Framework
Educational technology is rapidly evolving, with increasing attention on virtual characters to enhance digital learning environments. Although these characters can play many roles in a learning environment, one type that has been widely researched are pedagogical agents (PAs). PAs can encourage educational conversations, provide instruction or feedback, and motivate learners (Siegle et al., 2023). Previous research shows PAs can aid learning across various domains and age groups (Castro-Alonso et al., 2021; Guo & Goh, 2015; Schroeder et al., 2013; Wang et al., 2023), but their impacts on learner motivation are less clear (Heidig & Clarebout, 2011; Schroeder & Adesope, 2014). As PAs become more integrated into learning systems, it is necessary to build an in-depth and theory-driven understanding of how PAs may influence learners’ motivation. We used social cognitive theory (Bandura, 1997), situated expectancy-value theory (Eccles & Wigfield, 2020), interest theory (Hidi & Renninger, 2006), and self-determination theory (Ryan & Deci, 2000) to guide our examination of the impact of PAs on motivation constructs.
Methods and Materials
On 09/15/2023, we conducted our literature search in eight popular databases from various professional fields (e.g., APA PsychInfo, Education Research Complete; see Table 1). Our search string was designed to capture studies investigating the impacts of PAs on students’ motivation and consisted of the following: ("virtual human"* OR "embodied agent"* OR "virtual character"* OR "pedagogical agent"* OR "conversational agent"* OR "motivational agent"*) AND (motivat* OR self-efficacy OR self-confidence OR ability belief* OR self-concept OR interest* OR engag* OR value* OR util* OR “sense of belonging” OR belong*). There were 52 studies included in the review after screening abstracts and full texts for our inclusion criteria (e.g., must compare a PA to a non-PA condition, must measure some aspect of learners’ motivation; see Figure 1).
We used a three-level meta-analytic (3LMA) model that accounted for correlated and hierarchical effects (CHE) and used robust variance estimation (RVE). This approach allowed us to use more data than previous synthesis from each study, while providing precise estimates that account for the relationships between measures within studies. We used 3LMA with CHERVE to explore seven research questions that examined the effect of PAs on learners' broad motivation, self-efficacy expectations, value beliefs, utility beliefs, engagement, interest, and intrinsic motivation as well as possible moderators of these effects (due to space limitations moderators will not be discussed here).
Results and Scholarly Significance
Results revealed that PAs significantly influence learners’ self-efficacy expectations and interest and had non-significant effects on broad motivation, value beliefs, utility beliefs, engagement, and intrinsic motivation (see Table 2). The lack of significant findings for these constructs are indicative of a larger, long-standing problem in the field: there needs to be more cross-disciplinary collaborations to ensure better measurement and operationalizations of these motivation constructs. Our results show the importance of intentionality in PA design and implementation, and highlight that even though it has been noted in the literature for nearly 20 years, researchers in the field should use theoretically and psychometrically strong measures in their studies.

Authors