Collaborative Learning Quality Classification Through Physiological Synchrony Recorded by Wearable Biosensors

Interpersonal physiological synchrony has been consistently found during collaborative tasks. However, few studies have applied synchrony to predict collaborative learning quality in real classroom. To explore the relationship between interpersonal physiological synchrony and collaborative learning activities, this study collected electrodermal activity (EDA) and heart rate (HR) during naturalistic class sessions and compared the physiological synchrony between independent task and group discussion task. The students were recruited from a renowned university in China. Since each student learn differently and not everyone prefers collaborative learning, participants were sorted into collaboration and independent dyads based on their collaborative behaviors before data analysis. The result showed that, during group discussions, high collaboration pairs produced significantly higher synchrony than low collaboration dyads (p = 0.010). Given the equivalent engagement level during independent and collaborative tasks, the difference of physiological synchrony between high and low collaboration dyads was triggered by collaboration quality. Building upon this result, the classification analysis was conducted, indicating that EDA synchrony can identify different levels of collaboration quality (AUC = 0.767 and p = 0.015).

Introduction and Related Literatures

In a world that is deeply connected, collaborative learning is believed to be the most important way of learning, shared knowledge construction, decision-making, critical thinking, and problem solving (Bruffee, 1999; Dillenbourg, 1999; Van Kleef et al., 2010; Gokhale, 2012). Scholars and practitioners advocating collaborative learning believe that learning is inherently active, constructive, and social. Successful collaboration benefits the whole group by immersing the students in an active learning condition to increase engagement and joint attention, relearn through retrieval, negotiate multiple perspectives, increase working memory resources, to name a few (Johnson and Johnson, 1985; Barron, 2003; Roediger III and Karpicke, 2006; Kirschner et al., 2009; Kuhn and Crowell, 2011). Broader education goals, such as involvement, cooperation and teamwork, and civic responsibility, are also believed to be achieved by collaborative learning (Smith and MacGregor, 1992).

Dillenbourg (1999) provided a general definition of collaborative learning as “a situation in which two or more people learn or attempt to learn something together.” In this definition, “learn something” was broadly interpreted as activities, including “follow a course,” “study course material,” “perform learning activities such as problem solving,” “learn from lifelong work practice,” and “together,” was interpreted as different forms of interaction. In fact, individual interaction is crucial in successful collaborative learning (Soller et al., 1998; Hiltz and Turoff, 2002; Kreijns et al., 2003), and thus serves as a key component of collaboration quality.

Given the importance of collaborative learning, the measurement of its quality is, however, very complex and challenging. Existing approaches can be categorized into subjective and objective measurements. Subjective measurement mainly relies on self-report data, including both interview (Salovaara and Järvelä, 2003; Sidani and Reese, 2018) and scales (Orchard et al., 2012; Organization for Economic Co-operation and Development (OECD), 2013); while objective measurement mainly relies on explicit observational data, which captures verbal communication and non-verbal behaviors (Odom and Ogawa, 1992; Marty and Carron, 2011; Mehl, 2017; Chi et al., 2018). The main defect of self-report data is the subject perspective that could be manipulated by the participants. For instance, the social desirability bias is a famous potential threat (Fisher, 1993; Holbrook and Krosnick, 2009).

Analyzing the verbal content of interactions is the most straightforward approach for analyzing the quality of interaction, in both face-to-face and computer supported contexts and has been commonly used in educational and psychological studies (Grau and Whitebread, 2012; Kent et al., 2016; Vuopala et al., 2016). The limitations of content analysis, include labor intensive and difficult, to provide instant feedback to either students or teachers. Thanks to the rapid development of computing capability and machine learning algorithm, non-verbal interactions, such as eye contacts, facial expression, and body movement, are also possible to be captured and analyzed (Montague et al., 2013; Tunçgenç and Cohen, 2018) and has been proved to be sufficient indicators of collaborative learning quality in face-to-face classroom or in online courses (Bronack, 2011; Al Tawil, 2019). These qualitative methodologies have shown rich effectiveness in the interpretation of human behaviors (Hsieh and Shannon, 2005; Elo and Kyngäs, 2008). However, there are still issues on trustworthiness of both the content analysis and the interpretation of explicit behaviors when using these methods alone (Elo et al., 2014). The implicit factors, such as emotional contagion, and affect infusion among individuals may be more crucial to cognition (Okon-Singer et al., 2015), but is far from being fully investigated due to challenges in measurement and data constraint (Fujiki et al., 2002).

As quite a few factors may affect the quality of collaboration, emotion is one of the most significant and moderates human behaviors in observable patterns (Balters and Steinert, 2017). Emotion regulation abilities are highly related to the success of interpersonal interactions, especially when the individuals collaborate with peers or in the workplace (Lopes et al., 2005; Eligio et al., 2012).

Educational activities require intensive interpersonal interactions. Thus, emotion plays an important role in education, especially in collaborative tasks (Schutz and Pekrun, 2007; Järvenoja and Järvelä, 2009). The effect can be either positive or negative (Imai, 2010). Comparing to verbal content, emotional state is directly detectable though quantitative measurements. Building upon the theory on human automatic nerves system (ANS), the important components of collaborative learning, such as cognitive load and emotional state, are believed to be monitorable through neurophysiological signals including EEG, fNIRs, ECG, and EDA. Arousal and valence can be evoked and detected in specific situations (Agrafioti et al., 2011; Boucsein, 2012; Ramirez and Vamvakousis, 2012; Dawson et al., 2017). The effects of individual interactions on emotion can also be measured through multimodal physiological signals (Heaphy and Dutton, 2008; Mønster et al., 2016).

Neurophysiological signals have been considered as promising measurements of emotional characteristics and can capture students’ learning process that go beyond acquisition of knowledge (Léger et al., 2014; Ochoa and Worsley, 2016). Positive evidences on the correlation between interpersonal neurophysiological synchrony and interaction are consistently reported in recent years. Using various hyperscanning technologies, inter-brain synchrony has been identified during face-to-face communication or interactive decision-making (Jiang et al., 2012; Dikker et al., 2017; Hu et al., 2018), suggesting special neural processes recruited by interaction. Interpersonal physiological synchrony has also been consistently found during collaborative tasks and used as indicators of effective collaboration. Higher level of synchrony is associated with better task performance and learning gains in collaborative tasks (Ahonen et al., 2016; Pijeira-Díaz et al., 2016; Dich et al., 2018).

However, brain hyperscanning devices are usually not easily to use in real classroom learning, because of people’s concern on health safety issues of brain hyperscanning, and their visual interruption to both students and teachers. Data quality might also be a problem in naturalistic settings (Matusz et al., 2019). Physiological signal, on the other hand, can be easily and steadily recorded at distal sites, such as fingers and wrists (Boucsein, 2012), and thus easily accepted by parents and students. The majority of existing studies that measure physiological synchrony in collaborative learning are basically lab-based experiments (Pijeira-Díaz et al., 2016; Dich et al., 2018; Dindar et al., 2019). The tasks include open-ending problem-based learning topics, such as designing breakfast for marathoners, (Haataja et al., 2018) or pair-programing task with restricted solutions (Xie et al., 2018). Although a few of them collected data in real classroom (Ahonen et al., 2018), they only reported correlation between synchrony and collaboration, but did not further explore the practical potential of categorizing collaboration quality through synchrony in naturalistic scenarios.

Naturalistic scenario, instead of laboratory setting, is crucial in the research of collaborative learning. First, it is social in nature and cannot be simulated in fully controlled, isolated environment. Second, it is constructive and targets at high-level cognitive skills, such as problem-solving, knowledge construction, and collaboration, and cannot be substituted by simple cognitive tasks, which are frequently used in laboratory investigations. Third, even applying ethologically relevant stimuli in a laboratory context, participants may react differently in both behaviors and neurocognitive signals (de Heer et al., 2017; Qu et al., 2020). In the real classroom context, the interaction behavior varies across students, which is very different from laboratory settings, where participants will try their best to comply to the research design. In fact, naturalistic real-world research is believed to be necessary to understanding human behaviors in neuroscience (Matusz et al., 2019). Classroom learning can serve as an ideal semi-structured scenario to bridge the laboratory based research and the real world.

The rapid development of wearable biosensing technologies makes it possible to record neurophysiological signals during naturalistic classroom learning. Recently, researchers used portable EDA and EEG sensors in classroom to record the neurophysiological signals of teachers and students in variance of learning conditions including lectures, discussion, movie viewing, and real exam (Dikker et al., 2017; Poulsen et al., 2017; Zhang et al., 2018; Qu et al., 2020). By recording physiological signal in fully real-world collaborative learning, the present study attempts to apply physiological synchrony to predict interaction quality in real collaborative learning. Although collaborative activities can range from classroom discussions to team research that covers a whole semester or year, this study focuses on classroom discussion as it is the simplest and most general scenario among diverse collaborative learning approaches.

In the current study, the researchers collected physiological and behavioral data from naturalistic class sessions and analyzed interpersonal synchrony during individual and collaborative tasks. Students were naturally divided into two kinds of collaborative dyads (CDs) according to their different learning styles as captured by behaviors, i.e., CDs and IDs. The result showed a significant difference in interpersonal physiological synchrony between CDs and IDs, i.e., high and low interactive levels. Following classification analysis confirmed the potential of applying physiological synchrony as an indicator of collaboration quality. This finding is promising in future applications of evaluation in student learning style.

Materials and Methods

Participants

Participants were recruited from an undergraduate level elective course that requires no prerequisite domain knowledge. All participants were Chinese nationals and full-time university students. The 16 students who registered in this course were from 12 different departments and programs across natural sciences, pharmaceutical science, engineering, social sciences, and humanities. Data collection lasted for two class sessions in 2 consecutive weeks. Fifteen out of 16 students (M = 21.61, SD = 2.43, eight females) signed the informed consent form at the beginning of the first data collection session. One student quit at the second session due to health conditions.

Sixteen students formed four three-people and one four-people discussion groups at the beginning of the semester, resulted in 16 dyads in the first data collection session. The grouping was simply decided by their seat location and most of them were strangers to each other at the before taking this course. In the second session, the grouping remained the same. There were 14 dyads since one person quit from a three-people group, making a total of 30 dyads. Dyad no. 15 was eliminated from all analyses and no. 13 and no. 15 were eliminated from analyses on collaborative learning, because one student in this group shared laptop screen with their teammate and discussed (pair no. 15) during the independent task (IT), violating the experiment requirement (Figure 1).

FIGURE 1

Figure 1. Classroom setting and intra-group dyads. The gray figure did not participate in the study at all. The yellow figure and nodes were excluded from the first data collection session because of the violation of experiment requirement. The green figure quit the study in the second-round data collection session because of health conditions. Thus, a total of 28 dyads were used in the analyses.

Experimental Tasks and Materials

Each course sessions had two main parts. In the second half of each class session, after the lecture, the instructor assigned an open-ended problem to the class. Students were required to solve the problem independently first (IT), followed immediately by a group collaborative discussion (interaction analyzed in pairs, PT) on the same problem. The problems in both steps were the same, except that participants were asked to solve the problem alone or with group members. In the first class session, the students were asked to review course materials and sort out a list of key knowledge by its significance, then discuss with their group members to forge a comprehensive agreement on the list. In the second class session, the students were asked to explore new approaches for engagement measurements alone and then discuss with their group members to form a comprehensive approach. The group would share their final solution to the whole class.

A short survey was used to evaluate participants’ engagement level and emotional state during IT and PT, respectively. The engagement level was self-reported by the participants with a 5-point Likert scale. The emotional state was measured with a five-scale Self-Assessment Manikin (SAM) to rate the affective dimensions of valence, arousal, and dominance (Bradley and Lang, 1994).

Settings and Apparatus

The settings followed the naturalistic class settings of this course. Each student had their own chair desk with rolling wheels. Skin conductance and heart rate (HR) was collected from each participant using the unobtrusive Huixin Psychorus wristband (Beijing Huixin Technology, 2021), capturing data at a sampling rate of 40 Hz for EDA and 1 Hz for HR. Each group was videotaped during both IT and collaborative discussion.

Procedures

Before the beginning of the first data collection session, researchers collected the signed informed consent and helped the students to wear the wristbands properly to ensure good data quality.

There was a 3-min close-eye baseline session and a 2-min open-eye baseline session before the IT. After the baseline sessions, all instructions were given by the instructor. The independent step last for 7–10 min and the group collaborative learning task last for 12–17 min. The short survey was collected immediately after IT and PT (Figure 2). Two to five minutes were cut from the beginning of the PT sessions to eliminate any continued effect from the IT sessions in data analysis. Students’ own physiological data reports were provided to them after the data collection to appreciate their participation.

FIGURE 2

Figure 2. Procedure of the class data collection.

Physiological Data Preprocessing

Skin conductance was collected using galvanic skin response (GSR). GSR records the changes of the electrical activity on the skin. The more general name of the GSR is electrodermal activity also known as EDA (Boucsein et al., 2012). The term EDA will be used to refer to the skin conductance signal in the following part of the article. The visual inspection was performed to control the quality of the raw EDA signals. Samples of the EDA signal clips during independent and paired tasks were randomly selected for visual inspection and artifacts manual removal. The cleaned data were used for further calculations. The results of the new dataset were then compared with the results from the original data set. Since the difference between two datasets was not significant, manual artifacts removal was skipped for the whole sample to minimize the influence on the raw data. The signals were then smoothed using the Gaussian smoothing algorithm and was down sampled from 40 to 10 Hz in a MATLAB-based EDA analysis software (Ledalab 3.4.9; Benedek and Kaernbach, 2010).¹

Analysis and Results

Ground Truth and Procedure of Analysis

In naturalistic setting without artificial design, ground truth should first be defined before physiological data analysis. The ground truth is that students learn differently and not everyone prefers/suites collaborative learning. In the real class sessions, students acted in their own learning style, which refers to the stable trait which decides how learners perceive and respond to learning environments (Keefe, 1979). The Felder and Silverman mode categorizes students’ learning style into five dimensions: active/reflective; sensing/intuitive; visual/verbal; sequential/global; and inductive/deductive (Felder and Silverman, 1988; Alfonseca et al., 2006). Active learners prefer to internalize information from external environments, and they are more likely to share opinions with peers frequently during collaboration; however, reflective learners tend to examine and process information by themselves (Felder and Silverman, 1988). That is, even though the students were equally engaged in the tasks, they may generate different perspectives and preferences for collaborative learning mode, and use different cognitive strategies during interaction (Cabrera et al., 1998; Kayes, 2005). The homogeneity of learning styles within a group will also affect the interactive effect, and the higher homogeneity group members may have better collaboration quality (Alfonseca et al., 2006).

Thus, participants were sorted into CDs and IDs based on their collaborative behaviors. Table 1 presents the collaboration events and their descriptions. CDs are expected to have higher quality interaction during collaborative task and generate higher physiological synchrony.

TABLE 1

Table 1. Collaboration events and descriptions.

A series of hypothesis test were conducted to compare groups of participants along different dimensions. The tests were performed by t-test. Before performing the t-tests, tests of normality (Kolmogorov-Smirnov and Shapiro-Wilk) were applied to the EDA synchrony, engagement, and emotional state datasets for IT and PT, all the tests did not show significant departure from normality. Therefore, the EDA synchrony, engagement, and emotional state variables for IT and PT were normally distributed. Classification analysis was then applied to see if physiological synchrony can identify collaborative learning quality. Figure 3 indicates the logic of the whole analysis.

FIGURE 3

Figure 3. Participants were categorized into collaborative dyads (in red) and independent dyads (in blue) according to the behavioral indicators during collaborative learning task. A hypothesis test was performed to explore the relationship between collaboration behavior and physiological synchrony during collaborative learning. At last, the classification analysis was applied to test the effectiveness of physiological synchrony as an identifier of collaborative learning quality.

Define Collaborative and Independent Dyads

In the current study, two raters watched the videos of the group discussion and categorized the dyads into high or low collaboration style without being aware of the physiological data analysis. Effective interaction includes verbal and non-verbal interactions such as tight conversation, eye contact, and joint attention on course material. As discussed in the previous section, better collaboration behavior quality is strongly correlated with collaborative learning quality including mutual gaze and joint attention (Schneider and Pea, 2013).

The two raters recorded the time range of key interaction events in the video for each pair of students. For groups with three people, when the total time of interaction events exceeds 1/3 of the total time of PT, this pair is then categorized to CDs, otherwise to IDs. For the one group with four people, the threshold is set to 1/6 since there are six different pairs of students sharing the total interaction time. Fifteen dyads were sorted to CD and 14 dyads to ID.

Engagement and Emotional Statement

It is important to check the validity of experimental settings in naturalistic classroom. First, students’ engagement levels should be the same across the two sessions to ensure that any identified differences are not due to engagement differences. Second, students’ self-report on their subjective experience during IT and PT should be compared to check if the IT and PT did mean different learning strategies to them.

According to Figure 4, engagement difference was not found between IT and PT. The participants were equally and highly engaged in both the IT (M = 2.76, SD = 0.951) and group discussion task (PT, M = 3.00, SD = 1.035), t(28) = 1.565, p = 0.258. This result suggests that participants were equally and highly engaged in both IT and PT, making engagement less likely to be the possible confounding factor for the additional synchrony during collaboration.

FIGURE 4

Figure 4. Engagement and emotional state between IT and PT. ***p < 0.001.

Same analysis was also conducted on the three affective dimensions. During group discussions, participants were more aroused (PT: M = 2.03, SD = 0.944; IT: M = 1.14, SD = 0.833, t(28) = 3.455, p = 0.002) and experienced higher positive emotion (PT: M = 2.93, SD = 0.813; IT: M = 2.11, SD = 0.875, t(27) = 5.037, p < 0.001). In the independent sessions, the participants reported to be more in control to their situation (PT: M = 1.90, SD = 0.673; IT: M = 2.79, SD = 1.013, t(28) = −5.363, p < 0.001). Higher score on arousal and valence indicated pleasant and excited discussion atmosphere during the collaborative learning task. Lower score in dominance is reasonable during PT since the process of multi-personal discussion came with negotiation and compromise (Figure 4).

Physiological Synchrony

The algorithm for the computation of EDA synchrony was adopted from Marci and Orr (2006) and calculated the moment-by-moment physiological concordance named as single session index (SSI). Same algorithm was also implemented on HR.

It should be noted that the EDA signal used in the analysis was the overall EDA instead of plain skin conductance level (SCL) or skin conductance responses (SCRs). The SCL represents the tonic level of electrical conductivity of the skin, relating to the slow and background change of EDA. The SCRs represent the phasic changes of electrical conductivity of the skin, reflecting the rapid and event-related changes of EDA (Braithwaite et al., 2013; Posada-Quintero and Chon, 2020). This study did not focus on the SCRs of the physiological signal and paid more attention on the overall changing trend of the EDA. But to keep the high ecological validity of this naturalistic experiment, the researchers chose to not eliminate the possible influence of SCRs for the authenticity of the study.

First, the 10 Hz signal was further down sampled by averaging the 10 numbers in each second. The moment-by-moment slope of the 1 Hz data for each signal was then calculated using a 5-s window with a regression model at a 1-s roll-rate. Next, Pearson correlations were conducted on the slope for each pair of data with a 15 s window rolling at the rate of 1 s, reflecting a moment-by-moment synchrony in the last 15 s. The SSI is an index that shows the synchrony over a time period instead of discrete time points. It is the natural logarithm of the ratio of the sum of positive correlation coefficients divided by the absolute value of the sum of negative correlation coefficients over a given period of time.

$\mathrm{SSI}=\ln \frac{{{\displaystyle \sum }}_{i=0}^n\left(r_i|r>0\right)}{\left|{{\displaystyle \sum }}_{i=0}^n\left(r_i|r<0\right)\right|}$

A sample of each step of the signal processing and synchronization calculation for IT and PT is shown in Figure 5.

FIGURE 5

Figure 5. A sample of electrodermal activity (EDA) and heart rate (HR) data processing in the EDA-IT figure, (a) is the raw EDA signals of pair No. 10 in IT; (b) is the trajectory of slope for a 5-second window, rolling on the rate of 1 second; (c) shows the moment-by-moment correlation coefficients on a 15-second window. The three panels are the same as in the EDA-PT figure. This is an example of the same pair in the collaborative task. The synchrony (SSI) was higher than that of in IT. The figures in the third and fourth row are examples of HR data.

In the EDA-IT figure, (a) is the raw EDA signals of pair no. 10 in IT; (b) is the trajectory of slope for a 5-s window, rolling on the rate of 1 s; and (c) shows the moment-by-moment correlation coefficients on a 15-s window. The three panels are the same as in the EDA-PT figure. This is an example of the same pair in the collaborative task. The synchrony (SSI) was higher than that of in IT. The figures in the third and fourth row are examples of HR data.

It is interesting to find that synchrony on EDA during IT and PT reflects different styles of learners. When doing the IT, the synchrony level between ID (M = 0.322, SD = 0.449) and their CD peers (M = 0.009, SD = 0.485) was not significant [t(27) = 1.800, p = 0.084]; while during the group discussion, the synchrony among ID (M = −0.170, SD = 0.396) was significantly lower than the CD (M = 0.231, SD = 0.380), t(27) = 2.781, p = 0.010 (Figure 6).

FIGURE 6

Figure 6. Single session index (SSI) among CD is significantly higher than that in ID during collaborative learning tasks. ^**p < 0.01.

The results showed in Figure 6 also showed that SSI was significantly lower for ID during PT, as compared to IT [t(27) = 3.070, p = 0.005].

Same analysis was conducted to explore the difference of synchrony of HR. When doing IT, ID (M = 0.171, SD = 0.462) and CD (M = −0.158, SD = 0.637) showed insignificant difference, t(26) = 1.157, p = 0.129; and during PT, there was also no significant difference between the synchrony among ID (M = −0.076, SD = 0.446) and the CD (M = −0.106, SD = 0.327), t(26) = 0.203, p = 0.841 (Figure 7).

FIGURE 7

Figure 7. HR synchronization did not show significant difference between ID and CD during IT and PT.

Physiological Synchrony as a Classifier of Collaborative Learning Quality

Since there is a strong correlation between the interaction level and EDA synchrony, a receiver operating characteristic (ROC) analysis was performed to test the accuracy of synchrony as a classifier of the collaborative learning behaviors in both IT and PT. Results showed that SSI is an acceptable indicator to identify interaction levels for collaborative task (AUC = 0.767, p = 0.015). Synchrony did not discriminate different collaboration styles during IT (AUC = 0.343, p = 0.15), which is good since there was no collaborative behaviors and no significant difference between CD and ID during IT. The results of IT and PT together verified the robustness of synchrony as the predictor for collaborative learning quality (see Figure 8).

FIGURE 8

Figure 8. The ROC curve of EDA for PT (A) and IT (B) using synchrony as classifier for collaborative learning quality.

Same analysis was also applied on HR data. As shown in Figure 9, the synchrony of HR exhibited low accuracy in classifying collaboration style (AUC = 0.454, p = 0.679) during PT. This is consistent with the low HR synchrony and undifferentiated HR synchronization level across collaborative learning quality.

FIGURE 9

Figure 9. The receiver operating characteristic (ROC) curve of HR for PT (A) and IT (B) using synchrony as predictor for collaborative learning quality.

Conclusion and Discussion

The aim of the present study is to explore the potentials of using physiological synchrony to classify collaboration quality in realistic educational settings, based on consistently identified synchrony during interpersonal interaction by previous studies. Existing studies show that learners are diverse in learning style and collaborative learning can manifest this diversity while students take different roles in the learning process (Smith and MacGregor, 1992; Pashler et al., 2008).

In the current study, the participants were categorized into CDs and IDs according to their natural behaviors in the collaborative learning tasks, and this behavioral difference significantly correlated with EDA synchrony between the pairs of participants. The results showed that participants who were categorized as CD during group discussions were associated with higher EDA synchrony. However, there was no significant difference between CD and ID in collaborative tasks in their HR synchrony.

One possible explanation for these inconsistent results in EDA and HR may have to do with the fact that talking affects one’s cardiovascular system but not EDA. Talking, even without emotional expression, can increase the blood pressure of hypertension patients (Le Pailleur et al., 2001). Simple mental and verbal activities also affect HR variation through changes in respiratory frequency (Bernardi et al., 2000). On the other hand, no evidence was found for the correlation between EDA and free talking (Fowles et al., 2000). In the natural group discussion context, students focused on the same task, trying to forge an integrated answer. Among CD, two students were tuning their emotional state during the discussion, resulted in higher synchrony in EDA. But when two people talk, they talk in turns, not simultaneously, thus the asynchronous HR. Actually, when participants doing verbal and motor activities in unison, HR synchrony was significantly higher than during unsynchronized moments (Müller and Lindenberger, 2011; Noy et al., 2015). Therefore, even the cognitive and emotional elements that generated synchrony in EDA may also synchronize HR in cognitive tasks as reported in the laboratory based studies (Henning and Korbelak, 2005; Montague et al., 2014; Mitkidis et al., 2015), the effect could be mixed with that of talking on one’s HR. While on the other hand, EDA synchrony was identified during unstructured conversation (Silver and Parente, 2004).

The result also showed that during PT, the EDA synchrony of ID was significantly lower when during IT. It seemed counterfactual on first thought but it could be reasonable if learning style was brought into consideration. Learners differ in the preference for collaborative learning (Cabrera et al., 1998). As a result, different people would choose different learning strategies. Independent learners may prefer to learn by themselves and process information in a more implicit way. When doing IT, this kind of learners can spend more cognitive resources on their task, thus two learners may show a moderate physiological synchrony as shown in Figure 6. But when they were in collaborative learning context, they have to spare part of their cognitive resources to other people or the entire environment, or could be overwhelmed by the intense communication in the group. In this case, the ID participants may show an even lower physiological synchrony than during IT.

Classification analysis proved that physiological synchrony may serve as a good indicator for interpersonal interaction quality. Higher physiological synchrony is positively correlated with higher interaction level. That is, higher frequency and longer time of interaction behaviors. Similar approach can be found in the research of predicting communication behavior using neural or physiological synchronization (Henning et al., 2009; Jiang et al., 2012). This application can help to identify different collaborative learning quality of the learners. It can also give instructors feedback on course content. One student may be attracted to one topic or interaction scheme but disinclined to another. In such case, physiological synchrony can provide clues in teaching adjustment.

Our findings provide evidence for the potential application of biosensors in the real-world classroom. We focus on the connection between the bio-signals and human behaviors on which we believe is the advantage of this interdisciplinary research area. This project also suggests that future researches in the same realm place attention to the scope of appropriate assumptions and research questions so that the laboratory-based experiments and naturalistic setting studies can be good complement for each other.

Students’ immediate learning outcome was not evaluated as the tasks were open-ended class discussions. Next, we will choose class sessions that has planned quizzes as a measure of learning performance.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics Statement

The studies involving human participants were reviewed and approved by the Institute of Education, Tsinghua University. The patients/participants provided their written informed consent to participate in this study.

Author Contributions

YL, TW, and YZ designed the experiments and drafted the manuscript. YL, TW, and KW carried out the fieldwork and collected and analyzed the data. All authors contributed to the article and approved the submitted version.

Funding

This research was funded by the Research Grant of Institute of Future Education and Evaluation (IFEE), Tsinghua University, which is named Dynamic System Evaluation of Learning Science Teaching Strategies from the Interdisciplinary Perspective (2020IFEE003).

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Footnotes

1. ^www.ledalab.de

References

Agrafioti, F., Hatzinakos, D., and Anderson, A. K. (2011). ECG pattern analysis for emotion detection. IEEE Trans. Affect. Comput. 3, 102–115. doi: 10.1109/T-AFFC.2011.28

CrossRef Full Text | Google Scholar

Ahonen, L., Cowley, B. U., Hellas, A., and Puolamäki, K. (2018). Biosignals reflect pair-dynamics in collaborative work: EDA and ECG study of pair-programming in a classroom environment. Sci. Rep. 8, 1–16. doi: 10.1038/s41598-018-21518-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Ahonen, L., Cowley, B., Torniainen, J., Ukkonen, A., Vihavainen, A., and Puolamäki, K. (2016). Cognitive collaboration found in cardiac physiology: study in classroom environment. PLoS One 11:e0159178. doi: 10.1371/journal.pone.0159178

PubMed Abstract | CrossRef Full Text | Google Scholar

Alfonseca, E., Carro, R. M., Martín, E., Ortigosa, A., and Paredes, P. (2006). The impact of learning styles on student grouping for collaborative learning: a case study. User Model. User-Adap. Inter. 16, 377–401. doi: 10.1007/s11257-006-9012-7

CrossRef Full Text | Google Scholar

Al Tawil, R. (2019). Nonverbal communication in text-based, asynchronous online education. Int. Rev. Res. Open Dis. Learn. 20. doi: 10.19173/irrodl.v20i1.3705

CrossRef Full Text | Google Scholar

Balters, S., and Steinert, M. (2017). Capturing emotion reactivity through physiology measurement as a foundation for affective engineering in engineering design science and engineering practices. J. Intell. Manuf. 28, 1585–1607. doi: 10.1007/s10845-015-1145-2

CrossRef Full Text | Google Scholar

Barron, B. (2003). When smart groups fail. J. Learn. Sci. 12, 307–359. doi: 10.1207/S15327809JLS1203_1

CrossRef Full Text | Google Scholar

Beijing Huixin Technology (2021). Psychorus Wear. Available at: https://www.hxpsych.cn/kcdsjslclyws (Accessed April 5, 2021).

Google Scholar

Benedek, M., and Kaernbach, C. (2010). A continuous measure of phasic electrodermal activity. J. Neurosci. Methods 190, 80–91. doi: 10.1016/j.jneumeth.2010.04.028

PubMed Abstract | CrossRef Full Text | Google Scholar

Bernardi, L., Wdowczyk-Szulc, J., Valenti, C., Castoldi, S., Passino, C., Spadacini, G., et al. (2000). Effects of controlled breathing, mental activity and mental stress with or without verbalization on heart rate variability. J. Am. Coll. Cardiol. 35, 1462–1469. doi: 10.1016/S0735-1097(00)00595-7

PubMed Abstract | CrossRef Full Text | Google Scholar

Boucsein, W. (2012). Electrodermal Activity. Wuppertal, Germany: Springer Science & Business Media.

Google Scholar

Boucsein, W., Fowles, D. C., Grimnes, S., Ben-Shakhar, G., Roth, W. T., Dawson, M. E., et al. (2012). Publication recommendations for electrodermal measurements. Psychophysiology 49, 1017–1034. doi: 10.1111/j.1469-8986.2012.01384.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Bradley, M. M., and Lang, P. J. (1994). Measuring emotion: the self-assessment manikin and the semantic differential. J. Behav. Ther. Exp. Psychiatry 25, 49–59. doi: 10.1016/0005-7916(94)90063-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Braithwaite, J. J., Watson, D. G., Jones, R., and Rowe, M. (2013). A guide for analysing electrodermal activity (EDA) & skin conductance responses (SCRs) for psychological experiments. United Kingdom: Selective Attention & Awareness Laboratory (SAAL) Behavioural Brain Sciences Centre, University of Birmingham.

Google Scholar

Bronack, S. C. (2011). The role of immersive media in online education. J. Contin. High. Educ. 59, 113–117. doi: 10.1080/07377363.2011.583186

CrossRef Full Text | Google Scholar

Bruffee, K. A. (1999). Collaborative Learning: Higher Education, Interdependence, and the Authority of Knowledge. Baltimore, MD: Johns Hopkins University Press.

Google Scholar

Cabrera, A. F., Nora, A., Bernal, E. M., Terenzini, P. T., and Pascarella, E. T. (1998). “Collaborative learning: preferences, gains in cognitive and affective outcomes, and openness to diversity among college students” in Annual Meeting of Association for the Study of Higher Education; November 1998; Miami, FL.

Google Scholar

Chi, M. T., Adams, J., Bogusch, E. B., Bruchok, C., Kang, S., Lancaster, M., et al. (2018). Translating the ICAP theory of cognitive engagement into practice. Cogn. Sci. 42, 1777–1832. doi: 10.1111/cogs.12626

CrossRef Full Text | Google Scholar

Dawson, M. E., Schell, A. M., and Filion, D. L. (2017). “The electrodermal system” in Cambridge Handbooks in Psychology. Handbook of Psychophysiology. eds. J. T. Cacioppo, L. G. Tassinary, and G. G. Berntson (New York, USA: Cambridge University Press), 217–243.

Google Scholar

de Heer, W. A., Huth, A. G., Griffiths, T. L., Gallant, J. L., and Theunissen, F. E. (2017). The hierarchical cortical organization of human speech processing. J. Neurosci. 37, 6539–6557. doi: 10.1523/JNEUROSCI.3267-16.2017

PubMed Abstract | CrossRef Full Text | Google Scholar

Dich, Y., Reilly, J., and Schneider, B. (2018). “Using physiological synchrony as an indicator of collaboration quality, task performance and learning” in International Conference on Artificial Intelligence in Education; June 2018; Cham: Springer, 98–110.

Google Scholar

Dikker, S., Wan, L., Davidesco, I., Kaggen, L., Oostrik, M., McClintock, J., et al. (2017). Brain-to-brain synchrony tracks real-world dynamic group interactions in the classroom. Curr. Biol. 27, 1375–1380. doi: 10.1016/j.cub.2017.04.002

PubMed Abstract | CrossRef Full Text | Google Scholar

Dillenbourg, P. (1999). “What do you mean by collaborative learning?” in Collaborative-learning: Cognitive and Computational Approaches. ed. P. Dillenbourg (Oxford: Elsevier), 1–19.

Google Scholar

Dindar, M., Alikhani, I., Malmberg, J., Järvelä, S., and Seppänen, T. (2019). Examining shared monitoring in collaborative learning: a case of a recurrence quantification analysis approach. Comput. Hum. Behav. 100, 335–344. doi: 10.1016/j.chb.2019.03.004

CrossRef Full Text | Google Scholar

Eligio, U. X., Ainsworth, S. E., and Crook, C. K. (2012). Emotion understanding and performance during computer-supported collaboration. Comput. Hum. Behav. 28, 2046–2054. doi: 10.1016/j.chb.2012.06.001

CrossRef Full Text | Google Scholar

Elo, S., Kääriäinen, M., Kanste, O., Pölkki, T., Utriainen, K., and Kyngäs, H. (2014). Qualitative content analysis: a focus on trustworthiness. SAGE open 4, doi: 10.1177/2158244014522633

PubMed Abstract | CrossRef Full Text | Google Scholar

Elo, S., and Kyngäs, H. (2008). The qualitative content analysis process. J. Adv. Nurs. 62, 107–115. doi: 10.1111/j.1365-2648.2007.04569.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Felder, R. M., and Silverman, L. K. (1988). Learning and teaching styles in engineering education. J. Eng. Educ. 78, 674–681.

Google Scholar

Fisher, R. J. (1993). Social desirability bias and the validity of indirect questioning. J. Consum. Res. 20, 303–315. doi: 10.1086/209351

CrossRef Full Text | Google Scholar

Fowles, D. C., Kochanska, G., and Murray, K. (2000). Electrodermal activity and temperament in preschool children. Psychophysiology 37. 777–787. doi: 10.1111/1469-8986.3760777

PubMed Abstract | CrossRef Full Text | Google Scholar

Fujiki, M., Brinton, B., and Clarke, D. (2002). Emotion regulation in children with specific language impairment. Lang. Speech Hear. Serv. Sch. 33, 102–111. doi: 10.1044/0161-1461(2002/008)

PubMed Abstract | CrossRef Full Text | Google Scholar

Gokhale, A. A. (2012). “Collaborative learning and critical thinking” in Encyclopedia of the Sciences of Learning. ed. N. M. Seel (New York, USA: Springer), 634–636.

Google Scholar

Grau, V., and Whitebread, D. (2012). Self and social regulation of learning during collaborative activities in the classroom: the interplay of individual and group cognition. Learn. Instr. 22, 401–412. doi: 10.1016/j.learninstruc.2012.03.003

CrossRef Full Text | Google Scholar

Haataja, E., Malmberg, J., and Järvelä, S. (2018). Monitoring in collaborative learning: co-occurrence of observed behavior and physiological synchrony explored. Comput. Hum. Behav. 87, 337–347. doi: 10.1016/j.chb.2018.06.007

CrossRef Full Text | Google Scholar

Heaphy, E. D., and Dutton, J. E. (2008). Positive social interactions and the human body at work: linking organizations and physiology. Acad. Manag. Rev. 33, 137–162. doi: 10.5465/amr.2008.27749365

CrossRef Full Text | Google Scholar

Henning, R. A., Armstead, A. G., and Ferris, J. K. (2009). Social psychophysiological compliance in a four-person research team. Appl. Ergon. 40, 1004–1010. doi: 10.1016/j.apergo.2009.04.009

PubMed Abstract | CrossRef Full Text | Google Scholar

Henning, R. A., and Korbelak, K. T. (2005). Social-psychophysiological compliance as a predictor of future team performance. Psychologia 48, 84–92. doi: 10.2117/psysoc.2005.84

CrossRef Full Text | Google Scholar

Hiltz, S. R., and Turoff, M. (2002). What makes learning networks effective? Commun. ACM 45, 56–59. doi: 10.1145/505248.505273

CrossRef Full Text | Google Scholar

Holbrook, A. L., and Krosnick, J. A. (2009). Social desirability bias in voter turnout reports: tests using the item count technique. Public Opin. Q. 74, 37–67. doi: 10.1093/poq/nfp065

CrossRef Full Text | Google Scholar

Hsieh, H.-F., and Shannon, S. E. (2005). Three approaches to qualitative content analysis. Qual. Health Res. 15, 1277–1288. doi: 10.1177/1049732305276687

PubMed Abstract | CrossRef Full Text | Google Scholar

Hu, Y., Pan, Y., Shi, X., Cai, Q., Li, X., and Cheng, X. (2018). Inter-brain synchrony and cooperation context in interactive decision making. Biol. Psychol. 133, 54–62. doi: 10.1016/j.biopsycho.2017.12.005

PubMed Abstract | CrossRef Full Text | Google Scholar

Imai, Y. (2010). Emotions in SLA: new insights from collaborative learning for an EFL classroom. Mod. Lang. J. 94, 278–292. doi: 10.1111/j.1540-4781.2010.01021.x

CrossRef Full Text | Google Scholar

Järvenoja, H., and Järvelä, S. (2009). Emotion control in collaborative learning situations: do students regulate emotions evoked by social challenges. Br. J. Educ. Psychol. 79, 463–481. doi: 10.1348/000709909X402811

PubMed Abstract | CrossRef Full Text | Google Scholar

Jiang, J., Dai, B., Peng, D., Zhu, C., Liu, L., and Lu, C. (2012). Neural synchronization during face-to-face communication. J. Neurosci. 32, 16064–16069. doi: 10.1523/JNEUROSCI.2926-12.2012

PubMed Abstract | CrossRef Full Text | Google Scholar

Johnson, D. W., and Johnson, R. T. (1985). “The internal dynamics of cooperative learning groups” in Learning to Cooperate, Cooperating to Learn. eds. R. Slavin, S. Sharan, S. Kagan, R. Hertz-Lazarowitz, C. Webb, and R. Schmuck (Boston, MA: Springer), 103–124.

Google Scholar

Kayes, D. C. (2005). Internal validity and reliability of Kolb’s learning style inventory version 3 (1999). J. Bus. Psychol. 20, 249–257. doi: 10.1007/s10869-005-8262-4

CrossRef Full Text | Google Scholar

Keefe, J. W. (ed.) (1979). “Learning style: an overview” in Student Learning Styles: Diagnosing and Prescribing Programs. Vol. 1. Reston, VA: National Association of Secondary School Principals, 1–17.

Google Scholar

Kent, C., Laslo, E., and Rafaeli, S. (2016). Interactivity in online discussions and learning outcomes. Comput. Educ. 97, 116–128. doi: 10.1016/j.compedu.2016.03.002

CrossRef Full Text | Google Scholar

Kirschner, F., Paas, F., and Kirschner, P. A. (2009). A cognitive load approach to collaborative learning: united brains for complex tasks. Educ. Psychol. Rev. 21, 31–42. doi: 10.1007/s10648-008-9095-2

CrossRef Full Text | Google Scholar

Kreijns, K., Kirschner, P. A., and Jochems, W. (2003). Identifying the pitfalls for social interaction in computer-supported collaborative learning environments: a review of the research. Comput. Hum. Behav. 19, 335–353. doi: 10.1016/S0747-5632(02)00057-2

CrossRef Full Text | Google Scholar

Kuhn, D., and Crowell, A. (2011). Dialogic argumentation as a vehicle for developing young adolescents’ thinking. Psychol. Sci. 22, 545–552. doi: 10.1177/0956797611402512

PubMed Abstract | CrossRef Full Text | Google Scholar

Léger, P. M., Davis, F. D., Cronan, T. P., and Perret, J. (2014). Neurophysiological correlates of cognitive absorption in an enactive training context. Comput. Hum. Behav. 34, 273–283. doi: 10.1016/j.chb.2014.02.011

CrossRef Full Text | Google Scholar

Le Pailleur, C., Montgermont, P., Feder, J. M., Metzger, J. P., and Vacheron, A. (2001). Talking effect and “white coat” effect in hypertensive patients: physical effort or emotional content? Behav. Med. 26, 149–157. doi: 10.1080/08964280109595762

PubMed Abstract | CrossRef Full Text | Google Scholar

Lopes, P. N., Salovey, P., Côté, S., Beers, M., and Petty, R. E. (2005). Emotion regulation abilities and the quality of social interaction. Emotion 5:113. doi: 10.1037/1528-3542.5.1.113

PubMed Abstract | CrossRef Full Text | Google Scholar

Marci, C. D., and Orr, S. P. (2006). The effect of emotional distance on psychophysiologic concordance and perceived empathy between patient and interviewer. Appl. Psychophysiol. Biofeedback 31, 115–128. doi: 10.1007/s10484-006-9008-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Marty, J. C., and Carron, T. (2011). Observation of collaborative activities in a game-based learning platform. IEEE Trans. Learn. Technol. 4, 98–110. doi: 10.1109/TLT.2011.1

CrossRef Full Text | Google Scholar

Matusz, P. J., Dikker, S., Huth, A. G., and Perrodin, C. (2019). Are we ready for real-world neuroscience? J. Cogn. Neurosci. 31, 327–338. doi: 10.1162/jocn_e_01276

PubMed Abstract | CrossRef Full Text | Google Scholar

Mehl, M. R. (2017). The electronically activated recorder (EAR) a method for the naturalistic observation of daily social behavior. Curr. Dir. Psychol. Sci. 26, 184–190. doi: 10.1177/0963721416680611

PubMed Abstract | CrossRef Full Text | Google Scholar

Mitkidis, P., McGraw, J. J., Roepstorff, A., and Wallot, S. (2015). Building trust: heart rate synchrony and arousal during joint action increased by public goods game. Physiol. Behav. 149, 101–106. doi: 10.1016/j.physbeh.2015.05.033

PubMed Abstract | CrossRef Full Text | Google Scholar

Mønster, D., Håkonsson, D. D., Eskildsen, J. K., and Wallot, S. (2016). Physiological evidence of interpersonal dynamics in a cooperative production task. Physiol. Behav. 156, 24–34. doi: 10.1016/j.physbeh.2016.01.004

PubMed Abstract | CrossRef Full Text | Google Scholar

Montague, E., Chen, P. Y., Xu, J., Chewning, B., and Barrett, B. (2013). Nonverbal interpersonal interactions in clinical encounters and patient perceptions of empathy. J. Participat. Med. 5:e33.

Google Scholar

Montague, E., Xu, J., and Chiou, E. (2014). Shared experiences of technology and trust: an experimental study of physiological compliance between active and passive users in technology-mediated collaborative encounters. IEEE Trans. Hum. Mach. Syst. 44, 614–624. doi: 10.1109/THMS.2014.2325859

CrossRef Full Text | Google Scholar

Müller, V., and Lindenberger, U. (2011). Cardiac and respiratory patterns synchronize between persons during choir singing. PLoS One 6:e24893. doi: 10.1371/journal.pone.0024893

PubMed Abstract | CrossRef Full Text | Google Scholar

Noy, L., Levit-Binun, N., and Golland, Y. (2015). Being in the zone: physiological markers of togetherness in joint improvisation. Front. Hum. Neurosci. 9:187. doi: 10.3389/fnhum.2015.00187

PubMed Abstract | CrossRef Full Text | Google Scholar

Ochoa, X., and Worsley, M. (2016). Augmenting learning analytics with multimodal sensory data. J. Learn. Sci. 3, 213–219. doi: 10.18608/jla.2016.32.10

CrossRef Full Text | Google Scholar

Odom, S. L., and Ogawa, I. (1992). Direct observation of young children’s social interaction with peers: a review of methodology. Behav. Assess. 14, 407–441.

Google Scholar

Okon-Singer, H., Hendler, T., Pessoa, L., and Shackman, A. J. (2015). The neurobiology of emotion–cognition interactions: fundamental questions and strategies for future research. Front. Hum. Neurosci. 9:58. doi: 10.3389/fnhum.2015.00058

PubMed Abstract | CrossRef Full Text | Google Scholar

Orchard, C. A., King, G. A., KhaLCi, H., and Bezzina, M. B. (2012). Assessment of interprofessional team collaboration scale (AITCS): development and testing of the instrument. J. Contin. Educ. Health Prof. 32, 58–67. doi: 10.1002/chp.21123

PubMed Abstract | CrossRef Full Text | Google Scholar

Organization for Economic Co-operation and Development (OECD) (2013). PISA 2015: Draft collaborative problem solving framework.

Google Scholar

Pashler, H., McDaniel, M., Rohrer, D., and Bjork, R. (2008). Learning styles: concepts and evidence. Psychol. Sci. Public Interest 9, 105–119. doi: 10.1111/j.1539-6053.2009.01038.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Pijeira-Díaz, H. J., Drachsler, H., Järvelä, S., and Kirschner, P. A. (2016). “Investigating collaborative learning success with physiological coupling indices based on electrodermal activity” in Proceedings of the Sixth International Conference on Learning Analytics & Knowledge; April 2016; ACM, 64–73.

Google Scholar

Posada-Quintero, H. F., and Chon, K. H. (2020). Innovations in electrodermal activity data collection and signal processing: a systematic review. Sensors 20:479. doi: 10.3390/s20020479

PubMed Abstract | CrossRef Full Text | Google Scholar

Poulsen, A. T., Kamronn, S., Dmochowski, J., Parra, L. C., and Hansen, L. K. (2017). EEG in the classroom: synchronised neural recordings during video presentation. Sci. Rep. 7:43916. doi: 10.1038/srep43916

PubMed Abstract | CrossRef Full Text | Google Scholar

Qu, Z., Chen, J., Li, B., Tan, J., Zhang, D., and Zhang, Y. (2020). Measurement of high-school students’ trait math anxiety using neurophysiological recordings during math exam. IEEE Access 8, 57460–57471. doi: 10.1109/ACCESS.2020.2982198

CrossRef Full Text | Google Scholar

Ramirez, R., and Vamvakousis, Z. (2012). “Detecting emotion from EEG signals using the emotive epoc device” in International Conference on Brain Informatics. Lecture Notes in Computer Science. Vol. 7670. eds. F. M. Zanzotto, S. Tsumoto, N. Taatgen and Y. Yao; December 2012; Berlin, Heidelberg: Springer, 175–184.

Google Scholar

Roediger, H. L. III., and Karpicke, J. D. (2006). The power of testing memory: basic research and implications for educational practice. Perspect. Psychol. Sci. 1, 181–210. doi: 10.1111/j.1745-6916.2006.00012.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Salovaara, H., and Järvelä, S. (2003). Student strategic actions in computer-supported collaborative learning. Learn. Environ. Res. 6, 267–284. doi: 10.1023/A:1027379824485

CrossRef Full Text | Google Scholar

Schneider, B., and Pea, R. (2013). Real-time mutual gaze perception enhances collaborative learning and collaboration quality. Int. J. Comput.-Support. Collab. Learn. 8, 375–397. doi: 10.1007/s11412-013-9181-4

CrossRef Full Text | Google Scholar

Schutz, P. A., and Pekrun, R. (2007). “Introduction to emotion in education” in Emotion in Education. eds. P. A. Schutz, R. Pekrun, and G. D. Phye (San Diego, CA: Academic Press), 3–10.

Google Scholar

Sidani, Y., and Reese, S. (2018). A journey of collaborative learning organization research: interview with Victoria Marsick and Karen Watkins. Learn. Organ. 25, 199–209. doi: 10.1108/TLO-01-2018-0015

CrossRef Full Text | Google Scholar

Silver, R., and Parente, R. (2004). The psychological and physiological dynamics of a simple conversation. Soc. Behav. Pers. 32, 413–418. doi: 10.2224/sbp.2004.32.5.413

CrossRef Full Text | Google Scholar

Smith, B. L., and MacGregor, J. T. (1992). “What is collaborative learning” in Collaborative Learning: A Sourcebook for Higher Education. eds. A. Goodsell, M. Maher, and V. Tinto (University Park, PA: National Center on Postsecondary Teaching, Learning and Assessment, Pennsylvania State University).

Google Scholar

Soller, A., Goodman, B., Linton, F., and Gaimari, R. (1998). “Promoting effective peer interaction in an intelligent collaborative learning system” in International Conference on Intelligent Tutoring Systems. Lecture Notes in Computer Science. eds. B. P. Goettl, H. M. Halff, C. L. Redfield and V. J. Shute. August 1998; Berlin, Heidelberg: Springer, 186–195.

Google Scholar

Tunçgenç, B., and Cohen, E. (2018). Interpersonal movement synchrony facilitates pro-social behavior in children's peer-play. Dev. Sci. 21:e12505. doi: 10.1111/desc.12505

PubMed Abstract | CrossRef Full Text | Google Scholar

Van Kleef, G. A., De Dreu, C. K., and Manstead, A. S. (2010). “An interpersonal approach to emotion in social decision making: the emotions as social information model” in Advances in Experimental Social Psychology. Vol. 42. ed. M. P. Zanna (Burlington, MA: Academic Press), 45–96.

Google Scholar

Vuopala, E., Hyvönen, P., and Järvelä, S. (2016). Interaction forms in successful collaborative learning in virtual learning environments. Act. Learn. High. Educ. 17, 25–38. doi: 10.1177/1469787415616730

CrossRef Full Text | Google Scholar

Xie, B., Reilly, J. M., Dich, Y. L., and Schneider, B. (2018). “Augmenting qualitative analyses of collaborative learning groups through multi-modal sensing” Rethinking Learning in the Digital Age: Making the Learning Sciences Count, 13th International Conference of the Learning Sciences (ICLS). Vol. 1. eds. J. Kay and R. Luckin (London, UK: International Society of the Learning Sciences).

Google Scholar

Zhang, Y., Qin, F., Liu, B., Qi, X., Zhao, Y., and Zhang, D. (2018). Wearable neurophysiological recordings in middle-school classroom correlate with students’ academic performance. Front. Hum. Neurosci. 12:457. doi: 10.3389/fnhum.2018.00457

PubMed Abstract | CrossRef Full Text | Google Scholar