Assessing the Social Agency of Pedagogical Agents in Adaptive Training Systems

Abstract

Pedagogical agents (PAs) could be used to mimic a one-on-one human tutoring experience in adaptive training systems. Social agency theory (SAT) is one perspective that describes how humans learn from PAs. However, there are no measures to test the components of SAT and their effects on learning. Therefore, we discuss the development of a subjective measure, the Social Agency Theory Questionnaire (SATQ), to assess components of the social agency theory framework: social cues, the cooperation principle, and deep cognitive processing. Next, we present a study that investigates the effectiveness of a PA instructor who provides error-sensitive human-voiced feedback to learners in an adaptive training system (PA-Present group), compared against an instructionally equivalent text-based instructor system (PA-Absent group). We hypothesize that the PA-Present group will exhibit higher learning outcomes than the PA-Absent group. Additionally, we hypothesize that participants in the PA-Present group will rate the “instructor” higher on all three subscales of the SATQ than participants in the PA-Absent group. Data collection for this project is currently underway. Future analyses will examine between-group differences on performance and learning outcomes, and their association with subjective ratings on SATQ subscales. In addition, we will present psychometric analyses of the SATQ and proposed revisions to the scale.