Does Skin Conductance Response Indicate Metacognitive Processes?

• In Event: 34.030 - Physiology of Emotion in Education: Theory, Methodology, and Insight

Abstract

Objectives: Metacognitive monitoring during complex learning with ALTs is extremely challenging for students (Azevedo et al., in press). Adaptive feedback and scaffolding may be given if we can identify moments of effective and ineffective metacognitive processes, which can be done using multichannel data. Here, we focused on SCR events, given that they are temporally contextualized, reactive responses. Our present study allows us to explore SCRs as students engage in STEM learning. We aim to unravel the complexities of timing, duration, and context of SCR events as indicators of CAM processes.

Theoretical Frameworks: We draw upon knowledge of EDA to interpret SCR events (Boucsein, 2012). EDA has been shown to differentiate among positive and negative affect in learning contexts and everyday life (Picard et al., 2016; Vail et al., 2016). We conceptualize these moments of physiological reactivity within appraisal frameworks of emotion (Scherer, 2009), and couple these notions of discrete, contextualized affective events during learning with CAM processes (Winne & Hadwin, 2008) within the domain of multimedia learning (Mayer, 2014).

Method, Data, and Results: We collected multichannel trace data (EDA, eye tracking, face video), self-report, and behavioral data from 8 undergraduates who participated in a 3x3x2 within-subjects design. Each participant was exposed to three agent facial expressions: neutral (neutral expression), congruent (expression congruent with content relevancy, i.e., joy), and incongruent (expression incongruent with content relevancy, i.e., confusion); and each type of relevancy: fully relevant (text and diagram relevant to the question), text somewhat relevant (but diagram fully relevant), and diagram somewhat relevant (but text fully relevant). Two types of questions were posed: function (regarding the function of a body system) and malfunction (regarding the malfunction of a body system).
EDA recordings were collected using a Shimmer GSR+. SCR events were extracted using the Ledalab Matlab plugin, which applies continuous decomposition analysis to extract tonic and phasic EDA activity. SCR events were identified from the momentary changes of the phasic EDA signal. We standardized the amplitudes of SCR events to identify the largest SCR events, which were counted and standardized within each page of each trial to determine the number of SCR events per second across pages for each student.
We examined the standardized counts for the text and diagram content page with agent expression (Figures 1 & 2), the multiple-choice content question page, and the justification page in which students explained their answer. Our data show a dramatic increase in SCR events during the justification page (Figure 3), fewer SCR events during the content page, and very few SCRs during the multiple-choice question page (Figure 4). The temporal occurrence of SCR events may be further investigated at finer granularity.

Significance: Prior research has demonstrated the importance of multichannel data, and EDA in particular, to investigate real-time CAM processes (Azevedo et al., in press; Picard et al., 2016; Vail et al., 2016). Our approach has the potential to identify how SCR events correspond to metacognitive monitoring, which may inform ALTs that leverage physiological data to drive real-time feedback and scaffolding.

Authors

• Joseph Grafsgaard, North Carolina State University

• Roger Azevedo, North Carolina State University

• Nicholas Vincent Mudrick, North Carolina State University

• Michelle Taub, North Carolina State University

• Garrett C. Millar, North Carolina State University