Antecedents and Effects of Emotions in Technology-Based Learning: Integrating Three Decades of Research

• In Event: 41.025 - Learning, Engagement, and Assessment in Technology-Rich Learning Environments

Abstract

Rationale: Studies examining the role of anxiety in computer-assisted learning (Spielberger, 1970) or of ‘technophobia’ in the context of educational-vocational technology literacy development (Moreno, 2012) demonstrate that emotions emerged as key players in the education technology revolution from the beginning of this movement. With advances in the development of intelligent educational technologies and the growing recognition of the importance of emotions for learning more generally (Harley, 2016; Pekrun & Linnenbrink-Garcia, 2014), a large body of empirical findings addressing antecedents and effects of emotions in technology-based learning (TBL) has accumulated. To date, however, literature reviews have solely focused on the relative incidence of different emotions (D’Mello, 2013) or emotion detection within TBL (see Harley, 2016, for a summary).
Methods: To address this gap, we employed an inductive (Wilson, 2009) and control-value theory-based (CVT; Pekrun, 2006) meta-analytic approach to integrate evidence for causes and effects of learners’ emotions within TBL settings in nonclinical samples. As outlined in Figure 1, we conducted a literature search covering databases, conference proceedings and edited books to identify English-language peer-reviewed studies in which learners’ emotions were directly measured and/or experimentally induced within settings that involved hands-on interaction with technology for educational purposes (i.e., acquisition of technology skills or content knowledge such as mathematics). In a multi-stage review process, candidate studies were screened by two independent coders for their adherence to the above-mentioned criteria. Additionally, studies were required to have investigated discrete emotions as defined by the CVT (e.g., Loderer et al., under review) and to provide sufficient data to calculate effect sizes for at least one emotion-correlate relationship of interest (see Figure 2 for the research model and coding scheme). Both self-report indicators of antecedents (e.g., control-value appraisals) and consequents of emotions (e.g., learning strategy use) as well as experimental investigations of the effects of different TBL environments on emotions (e.g., TBL with versus without cognitive support) were included. For comparative purposes, effect sizes were extracted in the form of correlations following Wilson (2016; Morris & DeShon, 2002), using random number imputation (see Murayama, Miyatsu, Buchli, & Storm, 2014) and correcting for artificial dichotomization were necessary, and analyzed using different R packages (see Tables 1-3 for details).
Results and Significance: Overall, effect sizes were extracted from 185 reports. As expected, sample mean effect size estimates (see Tables 1-3) correspond to the patterns proposed by the CVT and previous empirical findings for non-TBL settings (Pekrun & Perry, 2014), indicating that emotions are closely linked to control-value appraisals as well as learning outcomes. Initial moderator analyses currently conducted suggest that methodological variables such as state- versus trait-level assessment of emotions may influence the strength and directions of the relations observed (Bieg, Goetz, & Hubbard, 2013). Furthermore, descriptive analyses document that research on emotions in TBL has evolved both quantitatively and qualitatively in recent years (see Figures 3-4). In sum, this study provides integrative evidence that can help inform the design of emotionally sound TBL (Lester et al., 2014) and highlights open questions for future research.

Authors

• Kristina Loderer, University of Munich

• Reinhard Pekrun, University of Essex

• James Lester, North Carolina State University