How Do Contextual Emotions Differ for Self-Regulated Learning and Scientific Reasoning During Game-Based Learning?

• In Event: 35.031 - Multimodal Data During Learning With Advanced Learning Technologies: What Does Evidence Reveal About Self-Regulated Learning?

Abstract

Game-based learning environments (GBLEs) are designed to foster sustained engagement and motivation during learning (Plass et al., 2015). GBLEs foster different types of learning processes, such as self-regulated learning (SRL) and scientific reasoning. Studies have investigated emotions during game-based learning (e.g., Ocumpaugh et al., 2017; Sabourin & Lester, 2014), however they have not investigated emotions during different learning contexts. The objective of this study was to investigate contextualized emotions during in-game activities during learning and gameplay with CRYSTAL ISLAND.

We use theoretical models for each learning process. The information processing theory (Winne, 2018) views SRL as an event, and the model of scientific discovery as dual search (Klahr & Dunbar, 1988) segments scientific reasoning into a hypothesis space and an experimental space; but neither model focuses on emotions. Thus, we use the model of affective dynamics (D’Mello & Graesser, 2012), which states that confusion is a result of cognitive disequilibrium, and can lead back to engagement if resolved, or frustration if not resolved.

Participants were 61 undergraduate students (69% female) from a large North American university who played CRYSTAL ISLAND (Rowe et al., 2011). The goal of the game was to solve the mystery of what illness impacted inhabitants. To do so, students read books and tested food items as the transmission source (see Figure 1). Students completed the game when they submitted a correct diagnosis (illness type, transmission source, and treatment).

We analyzed log files to determine overall game score as a proxy for performance (see Table 1) and positive or negative action outcomes to contextualize in-game activities. We differentiated between relevant or irrelevant books to solving the mystery, positive or negative food scan outcomes as the transmission source, and a correct or incorrect diagnosis. We ran videos of facial expressions of emotions through FACET (facial recognition software) to obtain evidence scores of confusion, frustration, and joy during book reading and after scan outcomes and diagnosis submissions.

Hotelling T2 tests (see Table 2) revealed significant differences in durations of emotions after positive, compared to negative scan results. One sample paired differences t-tests revealed that evidence score of confusion was significantly higher after a negative, compared to a positive scan (see Table 3). There were no significant correlations between emotions and overall game score, however regressions (see Table 4) revealed that joy during reading relevant, (vs. irrelevant) books, and confusion after positive (vs. negative) scans, both positively predicted overall game score.

Results have implications for understanding emotions during different processes at different levels. For example, if we can determine when confusion has a positive impact on overall game score, we can design GBLEs to foster this emotion during specific instances of gameplay. Future directions should aim to develop adaptive advanced learning technologies that scaffold students based on their contextualized emotions, as well as their cognitive, metacognitive, and motivational SRL data using multichannel data, such as eye tracking and physiological sensors, to ensure effective learning.

Authors

• Michelle Taub, University of Central Florida

• Robert Sawyer, North Carolina State University

• James Lester, North Carolina State University

• Roger Azevedo, University of Central Florida