Introduction

A variety of high-level competencies is critical for students’ academic performance and life-long development (Becker & Luthar, 2002; Snell & Lefstein, 2018). A lack of such competencies, including inquiry, collaborative knowledge creation, and metacognition, may limit students’ opportunities to benefit from education (Snell & Lefstein, 2018; Yang et al., 2020). Knowledge Building has shown effectiveness in helping students, including academically low-achieving students, develop these high-level competencies (e.g., Chen, 2017; Chen & Hong, 2016; Yang et al., 2016, 2020; Zhang et al., 2018). Knowledge Building is an epistemological and pedagogical approach with associated technology support (i.e., Knowledge Forum) that emphasizes students collectively taking cognitive, metacognitive, social, and emotional responsibilities to advance the frontiers of their community knowledge (Scardamalia & Bereiter, 2014).

Various emotions may arise in the exciting but challenging Knowledge Building process because students continuously experience alternation between cognitive disequilibrium and equilibrium as they encounter ideas conflicting with their knowledge. This study focused on academic emotions during students’ Knowledge Building process (Pekrun et al., 2002; Zhu et al., 2019). This paper uses "emotion" to refer to "academic emotion" unless otherwise specified. Emotions are ubiquitous, fundamental, and influential in students’ learning and knowledge generation as they critically influence students’ regulative actions, motivation, and learning processes and outcomes (Di Leo et al., 2019; Harley et al., 2019; Muis et al., 2018; Zhu et al., 2019). The recent decades have witnessed an increasing awareness that emotions should be studied as a social phenomenon rather than only at the interpersonal level (van Kleef & Côté, 2022). Furthermore, an increasing number of researchers recognize learning and Knowledge Building as emotionally dynamic experiences (D’Mello & Graesser, 2012; Di Leo et al., 2019; Zhu et al., 2019, 2022). Emotions can facilitate or constrain students’ learning; generally, positive emotions, such as curiosity and joy, can facilitate students’ learning, while negative emotions, such as frustration and anxiety, can hinder students’ learning (Di Leo et al., 2019; Muis et al., 2018).

Low-achieving students may be more likely to experience negative emotions because of their limited competencies. These negative emotions may be relevant to their unproductive interaction, collaboration, and inquiry, resulting in shallow Knowledge Building discourse and limited development of competencies (Yang, 2019; Yang et al., 2016, 2020). Therefore, there is a need to investigate what types of emotions low-achieving students experience and how their emotions evolve during the Knowledge Building process in order to inform future development of scaffolding strategies for diverse students. CSCL research has made major progress in characterizing socio-cognitive and socio-cultural dynamics. Still, limited research has examined emotions in CSCL interaction and, in particular, what low-achieving students experience in Knowledge Building or collaborative inquiry learning in general. More recently, increasing interest has been given to supporting students’ emotion regulation in collaboration (Järvenoja et al., 2020). This study continues in this direction, examining this less explored but critical area of academic emotions in Knowledge Building.

Literature review

Knowledge building and knowledge forum

Innovation-driven societies demand schools to prepare students to learn new knowledge, collaborate to solve problems and innovate in areas that may not even currently exist (Bereiter & Scardamalia, 2003). A time-honored way for schools to meet these needs is by immersing students in a knowledge-creation environment (Bereiter & Scardamalia, 2003). Knowledge Building is a pedagogical approach that advocates transforming schools into knowledge-creation organizations and engaging students in the core work of a knowledge society (Scardamalia & Bereiter, 2014). Defined by twelve principles (e.g., improvable ideas, collective responsibility, embedded, concurrent & transformative assessment), Knowledge Building emphasizes students’ collective responsibility for continuous idea improvement within a learning community (Scardamalia, 2002). It transfers considerable cognitive, metacognitive, and emotional responsibilities to students (Scardamalia, 2002; Scardamalia & Bereiter, 2014) to work on the ideas and questions that they are passionate about and monitor inquiry, identify inquiry gaps, and adjust inquiry directions by making collective efforts. Just like a published biology experiment can be tested and built on by different biologists around the world, in Knowledge Building, ideas are considered to live in a public space open for evaluation, questioning, criticizing, putting up alternatives, and improvement (Popper, 1972; Scardamalia et al., 1994). Students improve their community ideas by engaging in progressive Knowledge Building discourse, such as asking authentic questions, theorizing ideas to respond to questions, introducing information to support or refute a theory, integrating various ideas to achieve syntheses, and improving theories (Scardamalia, 2004; Yang et al., 2022). Knowledge Building discourse usually occurs face-to-face, in online Knowledge Forum (Fig. 1), and during field trips. Owing to the affordances of Knowledge Forum to help students record and deepen the ensuing discourse, Knowledge Forum discourse has become the most studied discourse in the literature (e.g., Chen, 2017; Yang et al., 2016). In line with this past work, this study also focuses on Knowledge Forum discourse.

Fig. 1
figure 1

Knowledge Forum view and note

Full size image

As shown in Fig. 1, Knowledge Forum is an online platform developed to support Knowledge Building. A group of users with shared goals (e.g., a class of students) can form a Knowledge Forum community; and in that context, community members can create different views to organize their inquiry topics. In each view, students can post notes with the support of epistemic scaffolds such as “I need to understand,” “this theory cannot explain,” and “putting our knowledge together.” Students can read, build on, and reference each other’s notes; the building-on and referencing relationships are represented using blue and black lines to help students visually see the connections between notes. Furthermore, assessment tools embedded in the Knowledge Forum help students assess the status of their connections and Knowledge Building discourse. For instance, the Social Network Analysis (SNA) applet can generate sociograms that visualize reading and build-on activity and calculate the corresponding network densities. The Note Contribution applet can generate concise graphs to visualize students' participation (number of notes read or written, and scaffolds used). The public views within a Knowledge Forum community, epistemic scaffolds within a note, assessment tools, and underlying Knowledge Building principles together help students engage in sustainable discourse moves such as contributing authentic questions, co-constructing explanations, conducting sustained inquiry, and improving ideas in Knowledge Forum (Scardamalia, 2004).

Helping students of various achievement levels to benefit from Knowledge Building is an important research area. Yang and colleagues found that low-achieving students could engage in productive Knowledge Building and develop higher-level competencies (Yang, 2019; Yang et al., 2016, 2020). Furthermore, in Knowledge Building, students may experience various emotions. As students take high epistemic agency in determining what ideas to pursue and how to work on these ideas (Scardamalia, 2002), they may perceive the relatedness of learning and a great sense of control over learning, and therefore experience positive emotions (Pekrun, 2006; Zhu et al., 2022). However, creating new knowledge is complicated and involves challenging periods. For instance, students may experience confusion, surprise, and frustration when they cannot resolve the cognitive incongruity caused by the difference between new information and existing knowledge, experiences, or beliefs (D’Mello & Graesser, 2012; Pekrun & Stephens, 2012). Emotions can facilitate or hinder the Knowledge Building process. However, few studies have investigated students’ emotions in the Knowledge Building context, especially those of low-achieving students.

Low-achieving students and knowledge building

Low-achieving students, often from low socioeconomic and diverse ethnic backgrounds (Dietrichson et al., 2017; Slavin et al., 2011), enter schools with relatively weak collaboration, cognition, metacognition skills and interest in learning (Becker & Luthar, 2002). Previous studies suggest that Knowledge Building can not only help students improve their academic performance but also develop high-level competencies (Yang, 2019; Yang et al., 2020). Supporting low-achieving students’ Knowledge Building can greatly benefit them (So et al., 2010; Yang, 2019; Yang et al., 2020).

Advancing Knowledge Building inquiry and ideas involves students' social-cognitive, metacognitive, and emotional competencies. The social-cognitive aspects, such as posing questions, contributing diverse explanations, and judging promising inquiry directions (Chen, 2017; Zhang et al., 2007), are essential for productive Knowledge Building. However, low-achieving students often lack the competence to make such contributions. Moreover, academically low-achieving students generally lack metacognitive skills (e.g., collaborative planning, monitoring, and reflection) and require appropriate scaffolding to engage in metacognitive activities (Yang et al., 2016, 2020; Zohar & Dori, 2003). Low-achieving students have limited social-emotional skills, such as developing cohesion and trust and building motivation and confidence (Yang et al., 2016). The lack of these social-cognitive, metacognitive, and emotional competencies may lead low-achieving students to experience unfavorable emotions and benefit less from Knowledge Building. However, few studies have focused on the emotions that low-achieving students experience in Knowledge Building.

Academic emotions: achievement emotions and epistemic emotions

Emotions can be categorized in different ways: for example, based on valence and activation (e.g., positive activating emotions, such as curiosity, and negative deactivating emotions, such as boredom) or based on objective focus (e.g., academic emotions, such as confusion, or social emotions, such as empathy; Pekrun, 2006; Pekrun, & Linnenbrink-Garcia, 2012). This study focuses on academic emotions directly related to academic settings and outcomes (Pekrun et al., 2002). Under the umbrella of academic emotions, there are achievement emotions and epistemic emotions (Pekrun & Stephens, 2010). Achievement emotions are related to learning activities, such as problem solving and undertaking tasks (e.g., frustration and anxiety), or to learning outcomes, such as success and failure (e.g., pride and shame; Di Leo et al., 2019). Epistemic emotions (e.g., curiosity, surprise, and confusion) result from learners’ appraisal of the alignment or misalignment between new information and existing knowledge or beliefs during knowledge generation (Muis et al., 2018; Pekrun & Stephens, 2012).

Students’ emotions influence their learning. Positive emotions usually benefit learning. Putwain et al. (2018) found reciprocal relations between higher joy, lower boredom, and academic achievement among grade 5 and 6 students studying mathematics. Izard (2013) and Lazarus (1991) found that a sense of making progress or accomplishment contributed to joy. The literature has also established a link between achievement and self-confidence in creative learning (Davies, 2000; Kimbell et al., 1991). Compared to joy and self-confidence, the impact of negative emotions on learning is more complex. For example, anxiety may spark irrelevant thoughts, worries, or feelings of failure but may trigger extrinsic motivation to invest effort (Pekrun et al., 2017). Confusion accompanying a state of cognitive disequilibrium may promote learning in certain circumstances (D’Mello et al., 2014; Worsley & Blikstein, 2015). Being in or transitioning to an expression of confusion was related to positive learning outcomes, whereas being in or transitioning to surprise—a mediator of cognitive disequilibrium—was associated with less desirable learning performance (Worsley & Blikstein, 2015). Confusion may transition to frustration if students cannot resolve it, and frustration may eventually transition to boredom, resulting in disengagement from learning if students face persistent failure in overcoming learning challenges (Di Leo et al., 2019).

Emotions are dynamic and fluid. Therefore, researchers pay increasing attention to the temporality and sequence of emotions (D’Mello & Graesser, 2012). Several theories are relevant to the transitions of students’ emotions. Because providing a comprehensive review of relevant theories is beyond the scope of this study, here we only describe several from the perspectives of social psychology, cognitive psychology, and neuropsychology: namely, Emotional Contagion Theory (Hatfield et al., 2014), Cognitive Disequilibrium Theory (D’Mello & Graesser, 2010) and 3D (three dimensions) Mind Model (Thornton & Tamir's, 2020), respectively. Emotional Contagion Theory describes that individuals have "the tendency to automatically mimic and synchronize facial expressions, vocalizations, postures, and movements with those of another person and, consequently, to converge emotionally" (Hatfield et al., 2009, p19-20). Individuals tend to catch their peers’ emotions through mimicry, feedback, and contagion. More recently, Parkinson (2020) indicated limited support regarding mimicry-based emotional contagion. Rather, the following three processes account for interpersonal and intergroup emotional convergence: emotion contagion, social appraisal (i.e., emotions depend on interpretation and evaluation of things happening outside the body), and orientational calibration (i.e., reciprocal adjustments may result in convergent emotional orientations, Parkinson, 2020). Cognitive Disequilibrium Theory (D’Mello & Graesser, 2010) describes the transition between students’ cognitive-affective networks. It suggests that engagement (equilibrium) is a base state until learners encounter contradiction, anomaly, or system breakdown. Their equilibrium is then disrupted, and they may experience disequilibrium, usually accompanied by confusion. When the confusion is resolved, students can return to a state of engagement, but if the confusion cannot be resolved, learners may experience frustration and feel stuck, and their learning may be blocked. Furthermore, persistent frustration may result in boredom.

Thornton and Tamir (2020) framed humans’ experience of thoughts and feelings using a 3D Mind Model. They synthesized and verified that a mental state could be represented with three dimensions: rationality (i.e., degree of cognition), social impact (i.e., the impacts of a state on social relationships), and valence (i.e., positive or negative extents of a state). In a nutshell, a mental state is a package of how individuals experience in terms of rationality, negative or positive emotions, and the extent of social impacts. The 3D Mind model helps explain the transitions of mental dynamics—a mental state is more likely to transition to a near state in the 3D mind space. For instance, if rationality and social impact are controlled, the possibility of transitioning from boredom to frustration is likely to be higher than the transitioning from boredom to joy because both frustration and boredom are negative emotions, while enjoyment is a positive emotion. Therefore, we hypothesize that in the Knowledge Building context, emotions tend to transition to similar emotions (e.g., confusion to frustration, to boredom).

The present study

The above research suggests several research gaps. First, compared to the cognitive and metacognitive processes, the emotional process of learning did not draw enough attention from researchers until recent decades (Calvo & D’Mello, 2011; Muis et al., 2018). Second, we have a limited understanding of the emotions of low-achieving students. Third, previous studies have been mainly implemented in the individual learning context (e.g., Pardos et al., 2013) or in small groups (e.g., Isohätälä et al., 2020; Järvenoja et al., 2020). However, fewer studies were conducted in a community context. We aim to fill these research gaps by examining the types and evolution of emotions experienced by low-achieving secondary school students in Knowledge Building. We seek to answer the following research questions:

  • RQ1: What emotions can be identified among low-achieving students in Knowledge Building?

  • RQ2: What are the transition and sequential patterns of low-achieving students’ emotions manifested in the Knowledge Building process?

  • RQ3: How do the emotions of low-achieving students evolve in the Knowledge Building process?

Methods

Research context and participants

This study was conducted in a Band 3 secondary school that enrolled students at the 10th percentile on a pre-admission government examination at the end of elementary school (Grade 6) in Hong Kong. Secondary schools are rated from Band 1 (highest) to Band 3 (lowest) based on the examination performance of the majority of students enrolled. Most of the students in Band 3 are from low socioeconomic backgrounds, have low academic achievement, and are not adequately engaged in schoolwork (Yang et al., 2020). This study involved two Grade 9 classes and two Grade 11 classes (i.e., 120 students in total) taking a Visual Arts course taught by the same teacher. The main objective of the Visual Arts course was to help students develop inquiry and creative ideas concerning visual arts. The students inquired into different topics in the Visual Arts course in Chinese over five months and wrote different numbers of Knowledge Forum notes (see Table 1 for details). Generally, students had three lessons per week: one lesson for painting and two lessons for theory learning. The students in Classes A and B had no previous experience with Knowledge Building, while those in Classes C and D had some Knowledge Building experience. The teacher who facilitated the four classes had extensive experience teaching Visual Arts using the Knowledge Building pedagogy.

Table 1 Classes involved in this study
Full size table

Pedagogical design

The pedagogical design of the four classes aimed to help students develop critical and creative thinking skills and cater to learner diversity in academic abilities by engaging in Knowledge Building inquiry. The teachers adopted a three-phase pedagogical process to enable students to become accustomed to Knowledge Building. This pedagogical design was refined based on the previous four-phased pedagogical design (Chan, 2011). Table 2 summarizes the three-phased pedagogical design with principle-based activities to help low-achieving students engage in productive Knowledge Building inquiry. Next, we elaborate on the design of each phase.

Table 2 Pedagogical design for the four classes: phases, activities, and principles
Full size table

Phase 1: Creating a collaborative and reflective culture and helping students develop collaboration, inquiry, and reflection skills for productive Knowledge Building inquiry. In Phase 1, to support students’ motivation and agency and enhance their collaboration, inquiry and reflection skills, the teacher designed whole-class discussions, small-group collaboration, and individual and collaborative note-writing (offline) and reflection opportunities. For example, to motivate students’ Knowledge Building inquiry, they were first given opportunities to create any shapes using colorful wires to represent their ideas concerning inquiry topics (e.g., sustainable development), followed by constructing three-dimensional objects (Fig. 2a) using all shapes created by group members to represent the group’s ideas about their inquiry topic. Then each group shared their ideas and objects with the whole class. To help students develop collaboration and inquiry skills and basic disciplinary understanding, the students constructed collaborative concept maps (see Fig. 2b as an example) in small groups of 5 to 7 students. The concept maps created by each group were then made public to the whole class in each class. Each class also created a Knowledge Building Wall (Fig. 2c) by writing ideas and questions on index cards and attaching them to the classroom wall for their peers’ review, questioning and reflection (Chan, 2011). To help students develop reasoning discourse skills, they engaged in the activity of "Crossing Your Life Point" (Fig. 2d) with scaffolding. Each group has a game board divided into three regions: Yes, No, and Don’t Know. Each group member was able to put pictures related to their inquiry topic into one of the regions according to their interpretation. When they put photos in place, they needed to explain their reason to other members. Also, each group needed to present their findings to the whole class, and the other groups had the opportunity to challenge their ideas. To help students develop reflection skills and habits, they were encouraged to write reflections after each lesson using question prompts such as "What I have learned," "What I do not understand," and "What I want to know more about."

Fig. 2
figure 2

Examples of Knowledge Building principle-based activities

Full size image

Phase 2: Engaging in problem-centered collaborative and reflective inquiry on Knowledge Forum. In this phase, small groups of 5 to 7 students were first guided to formulate further inquiry questions based on Knowledge Wall discussions and present the questions that were generated (Fig. 2e). In this process, they selected the questions they were most interested in for further inquiry on the Knowledge Forum. In Knowledge Forum, the students extended and deepened their inquiry of the selected questions by writing notes, co-constructing explanations, generating idea-deepening problems, revising ideas, and synthesizing ideas. The teacher guided the students’ Knowledge Building through various strategies, such as frequently creating opportunities for collaborative reflection using the Knowledge Forum’s integrated assessment tools.

Phase 3: Deepening collective inquiry and advancing domain understanding and community knowledge through student-directed portfolio assessment. After the students contributed a reasonable number of notes and had some domain understanding, they were guided to conduct observations and field trips to the museums and other relevant places (Fig. 2f). The field trips allowed students to conduct investigations related to their inquiry topics, such as interviewing local people to understand how to use design to support environmental conservation. Doing so might support students in deepening their inquiry. After these activities, small groups were guided to deepen their inquiry and advance ideas by conducting a group portfolio assessment. When creating portfolio notes, small groups reflected on the advancement of community knowledge and explained their theories by selecting community notes as evidence (Lee et al., 2006; van Aalst & Chan, 2007). At the end of their course, the students were encouraged to conduct individual portfolio assessments. Each student reflected on their individual and community ideas in individual portfolio notes, synthesized and extended community ideas, and constructed new theories to explain their inquiry issues. Doing so might help facilitate students' continuous Knowledge Building. In addition, the students were involved in frequent reflection opportunities with the help of Knowledge Forum’s integrated assessment tools.

Data sources and analyses

The primary data of this study was 1,565 notes recorded in Knowledge Forum (42 unfinished or off-topic notes were excluded), written by the 120 participants from four classes. In addition to the textual ideas in each note, Knowledge Forum also recorded the author and creation time of each note and its relationship with other notes (e.g., build-on relations). The relationships between notes and their created time allowed us to analyze each note within their respective discourse contexts rather than as independent notes.

To respond to the three RQs, we conducted thematic analysis, content analysis, lag sequential analysis (LSA; Bakeman & Gottman, 1997), sequential pattern mining, and emotion evolution modeling. Next, we elaborate on how different analyses were employed to answer the RQs.

Identifying low-achieving students’ emotions reflected in Knowledge Building discourse (RQ1)

TO characterize low-achieving students’ emotions, we first preprocessed the Knowledge Forum notes, arranging them into different inquiry threads using inquiry thread analysis, followed by a content analysis involving labeling the emotions reflected in the notes in each inquiry thread.

Preprocessing the Knowledge Forum notes using inquiry thread analysis

The first author classified the Knowledge Forum notes of each class into inquiry threads. An inquiry thread is a sequence of notes that address the same problem or topic (Zhang et al., 2007). The inquiry thread analysis helped us understand the students’ inquiry and set the context for the subsequent content analysis. The identification of inquiry threads consisted of the following procedures: identifying the principal problems by reading all of the Knowledge Forum notes several times, synthesizing all of the inquiry, and dividing them into different sub- inquiry; and assigning the notes that addressed the same problem (e.g., sustainability and ecosystem or factors affecting sustainable development) to an inquiry thread. This process revealed 18 inquiry threads in Class A, 12 in Class B, 9 in Class C, and 9 in Class D. To verify the reliability of the inquiry thread analysis, another researcher with expertise in qualitative analysis of students’ postings independently analyzed the 788 notes from Class A (50.35% of all of the notes). This researcher identified the principal problems and clustered the notes into inquiry threads. The two researchers achieved inter-coder reliability of 0.83 (Cohen’s kappa) for identifying inquiry threads.

Content analysis

To qualitatively trace the emotions experienced by the participants, we used the coding scheme presented in Appendix Table 7 to code all of the notes within each inquiry thread into different academic emotions. The coding framework consisted of seven academic emotions: joy, surprise, curiosity, confusion, anxiety, frustration, and boredom. The code none was added to the coding framework to mark notes that did not contribute much to the development of the inquiry or did not add much value (e.g., "Yes." "OK." "This is my opinion"). Because these notes did not suggest students were working on the inquiry threads, we coded the emotion as "None." The development of the emotions coding framework drew on the model of impasse-driven emotions (D’Mello & Graesser, 2011, 2012; D’Mello et al., 2014) and the framework for measuring epistemic-related emotions (Pekrun et al., 2017). We also referred to Craig and colleagues’ (2008) definition of emotions. The data analysis was an iterative process of comparing the codes and the data. In each inquiry thread, we assigned one of the seven emotions or none to each note based on the linguistic clues and the coding scheme. Two raters independently coded 788 notes from Class A; the inter-rater reliability between the two raters was 0.86 (Cohen’s kappa). The rest of the notes were coded by one of the two raters.

Uncovering the transition and sequential patterns of low-achieving students’ emotions (RQ2)

Lag sequential analysis (LSA)

Based on the emotion labeling results, we conducted an LSA (Bakeman & Gottman, 1997) to identify the transition probabilities associated with various emotions per class of students. LSA was conducted using the Lag Sequential package in R (Draper and O’Connor, 2019). The sequential algorithm transforms a sequence of emotions into transition matrices with frequencies associated with transitions between emotions. We calculated the z-score for each transition (i.e., one emotion code transitions directly to another emotion code) to determine its statistical significance. A z-score value greater than 1.96 indicated a significant transition between two codes (i.e., 95% confidence interval, p < 0.05, Bakeman & Quera, 2011). This z-score criterion was also applied in other LSA studies (e.g., Dishion et al., 1996; Tlili et al., 2021). For each class of students, for the transition between any two emotions, only if the z-score is above 1.96 was the relationship visualized among the other significant sequential transitions.

Sequential pattern mining

To identify the frequent subsequences of emotions in all the inquiry threads, we conducted sequential pattern mining using the ArulesSequences package in R (Zaki, 2001). Sequential pattern mining identifies a set of frequent subsequences among the exhaustive set of sequences consisting of ordered elements/events. The length of a subsequence can range from two to the length of the shortest sequence. Support is computed as the fraction of inquiry threads where a subsequence appears in the database. We set 0.6 as the support value to identify subsequences that occurred in most inquiry threads.

Constructing an evolution model of low-achieving students’ emotions (RQ3)

Based on the significant sequential transitions between emotions noted in the four classes and the subsequences identified in the inquiry threads, we constructed a comprehensive model of students' emotion evolution in Knowledge Building communities following four steps. First, we placed the seven academic emotion labels in a figure. Second, we ensured that each significant transition generated in each class was noted with a directed arrow in the figure. Next, we compared the constructed model with the identified subsequences to examine whether the model was able to explain the subsequences and whether new lines needed to be drawn between emotions. The reason for doing so is that the emotion links that were not significant based on the transition analysis but occurred in most of the inquiry threads were also of interest to us. Once such links were found in the subsequences, they were added to the model. Finally, to void crossing arrows in the model, we dragged the emotion nodes and added repetitive representations of certain emotions. The rationale for employing these steps was to develop a comprehensive model that is able to explain all possible significant sequential transitions and subsequences of emotions reflected in the Knowledge Building discourse of low-achieving students.

Results

Categories of low-achieving students’ emotions in Knowledge Building

Table 3 presents the descriptive data of the emotions of the participants. The most frequently manifested emotion in the Knowledge Building inquiry threads across the four classes was joy (53.16%), followed by confusion (15.53%), anxiety (10.93%) and curiosity (9.07%). Surprise (2.17%) and boredom (1.41%) were less frequently expressed. These results suggest that these low-achieving students manifested a predominately higher percentage of positive emotions (e.g., joy and curiosity) than negative emotions (e.g., anxiety, frustration, and boredom) in the Knowledge Building process.

Table 3 The descriptive data of students’ emotions in the four classes
Full size table

Table 3 also shows that the participants in the four classes manifested varying percentages of different emotions. For example, the percentages of joy varied across the four classes from 50.63% to 56.33%. Class A demonstrated the greatest percentage of frustration (8.23%), and class D had the highest percentage of boredom (3.52%). Furthermore, the older low-achieving students (those in Classes C and D) manifested relatively higher percentages of confusion (22.36% and 23.94%, respectively) than the younger low-achieving students (those in Classes A and B, at 15.86% and 10.13%, respectively). Furthermore, the older classes had lower percentages of anxiety (6.83% and 1%) than the younger classes (12.18% and 13.29%). A similar pattern was found for frustration.

Sequential patterns of emotions manifested in Knowledge Building inquiry threads

We used the Viswork package in R to visualize the significant transitions of emotions reflected in the Knowledge Forum notes from each class. The transition matrix and adjusted residuals Table (z-scores) of students’ emotions in each class can be found in Appendix Table 8 and Appendix Table 9. Only emotion sequences with a z-score greater than 1.96 are visualized. The greater the z-score, the thicker the link between the two codes. The sizes of the emotion nodes denote the frequency of each emotion in each class, with larger nodes representing greater frequency.

As shown in Fig. 3(a), there were ten significant transitions of emotions in Class A: boredom → frustration (z-score = 5.1), frustration → frustration (3.74), joy → joy (3.6), confusion → none (2.88), confusion → confusion (2.71), none → curiosity (2.65), anxiety → anxiety (2.43), curiosity → boredom (2.29), frustration → boredom (2.18), and joy → curiosity (2.02). These results suggest a tendency for emotions to transition to the same or similar emotions, reflecting the spread of similar emotions in students’ inquiry threads.

Fig. 3
figure 3

The significant sequential transitions of low-achieving students’ emotions in the four classes. Note. JOY: Joy; CUR: Curiosity; SUR: Surprise; CON: Confusion; ANX: Anxiety; FRU: Frustration; BOR: Boredom; NON: None

Full size image

Figure 3(b) reveals four significant emotion sequences: anxiety → anxiety (3.43), surprise → frustration (2.76), joy → confusion (2.46), and frustration → boredom (2.25). Overall, the significant transitions suggest a greater possibility for transitions from anxiety to anxiety, surprise to frustration, and frustration to boredom, suggesting the spread of negative emotions if teachers or students take no specific actions to intervene in students' emotions.

Figure 3(c) presents Class C's significant emotional transition diagram. It shows that five emotional sequences reached a significant level in Class C: curiosity → frustration (2.08), frustration → surprise (5.73), surprise → surprise (2.2), surprise → anxiety (2.97), and confusion → boredom (2.64). As shown in Fig. 3(c), confusion could lead to boredom, and there was a path of emotions manifested by the low-achieving students: curiosity–frustration–surprise–anxiety. Consistent with Classes A and B results, these results suggest that emotion is likely to transition to a similar emotion. Different from Classes A and B, Class C featured the unique transition of frustration → surprise (5.73). These results might be due to the ability of the older low-achieving students (those in Grade 11) to use cognitive and metacognitive learning strategies to resolve the contradictions they faced.

Figure 3(d) shows the four significant transitions exhibited in Class D: joy → curiosity (2.09), frustration → confusion (3.60), boredom → boredom (2.03), and none → boredom (3.58). The spread of similar emotions found in the students’ Knowledge Forum discourse in Classes A, B, and C was also found in Class D.

To further analyze the subsequences of emotions reflected in students’ inquiry threads, we conducted a sequential pattern analysis of the low-achieving students’ emotions indicated in each idea thread. When the support value was set to 0.6 (meaning that a particular emotion subsequence appears in no less than 60% of the inquiry threads), 370 subsequences were generated. Table 4 shows the ten longest emotion subsequences reflected in students’ idea threads, each including nine units. These subsequences consisted of many manifestations of joy and a few of curiosity or confusion, suggesting the popularity and spread of positive emotions in students' Knowledge Building discourse.

Table 4 Longest emotion subsequences in students’ idea threads
Full size table

Of the 370 subsequences, 274 include confusion, suggesting the popularity and importance of confusion in the students’ Knowledge Building, and 112 subsequences include confusion, curiosity, and joy. Some representative subsequences are shown in Table 5 (SID 1 to 7). We intentionally selected subsequences starting and ending with different emotions to showcase the sequential patterns. Overall, in more than 60% of the inquiry threads of the four classes, curiosity was found to lead to confusion (e.g., SID 1, 6, and 7), and joy or confusion was found to trigger curiosity (e.g., SID 2, 3, 4, and 5). These representative subsequences confirm Path 1 (joy–curiosity–confusion–joy) and its bi-directional pattern.

Table 5 Representative subsequences that included confusion, curiosity, and joy or anxiety (support value is greater than 0.6)
Full size table

With the support value set to 0.6, 65 subsequences included anxiety. Similarly, we selected the subsequences that started and ended with different emotions among the 65 subsequences. As shown in Table 5 (SID 8 to 16), curiosity or confusion could lead to anxiety (e.g., SID 10, 11, 15, 16). Subsequences such as 10, 11, and 14 in Table 5 show that curiosity, confusion, and joy was found to lead to anxiety. These examples support Path 4 (joy–curiosity–confusion–boredom–frustration–surprise-anxiety-joy).

With the support value set to 0.6, no surprise or boredom was picked up in any subsequences because of their low frequency. To understand how surprise and boredom were manifested in students’ inquiry threads, we reduced the support value to 0.3, which generated 31,090 emotion subsequences. Of these, 439 subsequences included surprise, and 31 included boredom. As Table 6 shows (SID 1 to 10), anxiety might lead to surprise (e.g., SID 1), and surprise might trigger anxiety (e.g., SID 2, 6, and 10). To a certain extent, these subsequences confirm Path 3 (e.g., joy–curiosity–confusion–boredom–frustration–surprise-anxiety). However, given the low frequencies of surprise, frustration, and boredom, subsequences such as Path 3 may not be picked up by sequential pattern analysis when considered with other more frequent emotions (e.g., joy, curiosity, and confusion). As suggested by the LSA, if these negative emotions occur, they are more likely to follow the path. Concerning boredom, we selected five subsequences that started and ended with different emotions among the 32 subsequences, as shown in Table 6 (SID 11 to 15). These five subsequences suggested that confusion might lead to boredom and that boredom might be resolved and lead to joy. These subsequences confirm Path 2 (joy–curiosity–confusion–boredom) to a certain extent.

Table 6 Representative subsequences that include surprise or boredom
Full size table

Evolution of emotions of low-achieving students

Based on the significant sequential transitions and the subsequences identified in the inquiry threads, we constructed a model of the evolution of low-achieving students’ emotions, as shown in Fig. 4. Overall, students’ emotions follow five primary paths in Knowledge Building communities. Path 1 is joy–curiosity–confusion–joy. This path suggests that students may start an inquiry thread with joy or curiosity. They may then feel confused when they do not fully understand a phenomenon, theory, or piece of information or when there is a cognitive conflict between a new idea and their existing knowledge. When such confusion is resolved, they tend to feel joy again. However, it should be noted that this path is not linear or one-directional. Rather, each emotion on the path may go in two directions (e.g., curiosity → confusion, and confusion → curiosity), and it is also possible for a thread of notes to manifest one emotion only (e.g., joy → joy → joy). However, when such confusion cannot be resolved, it might lead to boredom and produce Path 2: joy–curiosity–confusion–boredom. Path 3 is joy–curiosity–confusion–boredom–frustration–surprise-anxiety, or joy–curiosity–confusion–boredom–frustration–surprise or joy–curiosity–confusion–boredom–frustration. Continuing the previous description, when confusion cannot be resolved, and community members lose interest in the inquiry threads, they may feel bored. They may then feel frustrated if they cannot progress their discourse (e.g., "What do you actually mean? I didn’t actively engage in the discussion."). When community members sense cognitive incongruity between different ideas, they might feel surprised. Being stuck in this state might make students anxious, indicating their uneasiness or uncertainty about the idea they had contributed. In Paths 2 and 3, frustration, surprise, and anxiety might lead to confusion if students continue exploring their inquiry threads without addressing their cognitive disequilibrium. Path 4 is joy–curiosity–confusion–boredom–frustration–surprise-anxiety-joy, suggesting students’ ability to move out of anxiety and achieve a joyful state. Path 5 is curiosity-boredom, which is likely to happen when a question raised by a community member does not arouse the curiosity of other community members.

Fig. 4
figure 4

The evolution model of low-achieving students’ emotions in Knowledge Building

Full size image

Discussion

Academic emotions are ubiquitous and critical in Knowledge Building. Understanding students’ academic emotions during Knowledge Building is of great significance and can inform how to design appropriate scaffolding strategies to support students, particularly low-achieving students who need additional support (Scardamalia & Bereiter, 2014; Zhu et al., 2019). However, little research has focused on what emotions low-achieving students express and how their emotions evolve during the Knowledge Building process. This study unpacked the types of academic emotions that low-achieving students expressed in Knowledge Building and the evolution of the expressed emotions. This emotion evolution model provides implications for designing scaffolding or interventions to facilitate low-achieving students’ learning and conducive emotion transitions. The following findings are worthy of further discussion.

Identifying academic emotions of low-achieving students

In this study, the most frequently expressed emotion by the low-achieving students in the Knowledge Building process was joy (53.16%), followed by confusion (15.53%), anxiety (10.93%), and curiosity (9.07%). Surprise (2.17%) and boredom (1.41%) were less frequently expressed. These results suggest that low-achieving students mainly exhibited positive emotions in Knowledge Building. The proportions of emotions should be influenced by the Knowledge Building design of this study and the teacher’s facilitating role. To make Knowledge Building more accessible to low-achieving students, we emphasized three design components: (1) establishing a collaborative and reflective culture for inquiry, collaboration, and idea improvement; (2) staging principle-based activities to help low-achieving students gradually develop skills of inquiry, collaboration, and metacognition; and (3) providing frequent reflection opportunities to help students engage in productive reflection. These designs provided explicit scaffolding for students to engage in inquiry, idea improvement, and reflective assessments. The teacher deliberately experimented with the designs to engage low-achieving students in collaborative Knowledge Building in his classes. He believed that the Knowledge Building approach was promising for helping low-achieving students develop higher-order thinking skills and motivate them. Low-achieving students were thus able to take collective responsibility for their knowledge advances in supportive Knowledge Building contexts.

In previous studies conducted with normal-performing students, Zhu and colleagues (2019, 2022) found that Grade 1 to 3 students experienced a higher percentage of enjoyment in Knowledge Building when compared with other emotions. However, some studies conducted in non-Knowledge Building contexts found that learners experienced frustration, confusion, and curiosity more often than other emotions (e.g., D’Mello & Graesser, 2012; Di Leo et al., 2019). To some extent, this difference in students’ emotions in Knowledge Building and other learning contexts might indicate the effectiveness of the Knowledge Building model in supporting students’ pleasant emotional experiences. However, emotions expressed by low-achieving and normal-performing students in Knowledge Building and other learning contexts need to be further examined in the future.

Low-achieving students from different classes expressed different proportions of academic emotions. For instance, Classes C and D had greater percentages of confusion than Classes A and B. Their emotional transition patterns also differed. The difference may be related to students’ cognitive and emotional development. Classes C and D had Knowledge Building experience and may have developed question-and-explanation skills to deepen their Knowledge Building inquiry progressively. For example, students in Classes C and D were more likely to help each other deepen the Knowledge Building inquiry by asking questions and requesting further explanations, coded as confusion in the study. Additionally, the older low-achieving students (i.e., grade 11 students) expressed a higher percentage of confusion but a lower percentage of anxiety and frustration than the younger students (i.e., grade 9 students). A possible reason for this difference is that the older low-achieving students were more capable of regulating and addressing negative emotions such as anxiety and frustration in Knowledge Building, which requires the effective use of appropriate cognitive, metacognitive, and emotional strategies. These results are consistent with prior research (e.g., D’Mello & Graesser, 2012) reporting that undergraduates are better than younger students at dealing with negative emotions, such as confusion.

As the first study investigating the academic emotions of low-achieving students in Knowledge Building, this study examined emotions at the community level rather than the individual level or group level. The analysis perspective aligns with the tradition of Knowledge Building research that generally emphasizes and focuses on community-level knowledge advancement and social interactions (e.g., Yang et al., 2016, 2022; Zhang et al., 2011). The community-level analysis can help us understand what emotions students experience as a community and how their emotions change and develop. However, group and individual analysis of low-achieving students’ emotions is also crucial to provide students with personalized support. This present study is the first round of a larger project that aims to help low-achieving students experience productive academic emotions and engage in productive emotion regulation. We are conducting systemic research such as testing the explanatory power of the evolution model, designing Knowledge Building environments to help low-achieving students engage in productive emotions, and examining individual and group characteristics that contribute to students’ productive emotions and emotion regulation.

Evolution of emotions among low-achieving students

The sequential pattern analysis suggests that emotions tend to transition to the same or similar emotions (e.g., joy to joy and boredom to frustration). This result is consistent with previous studies (e.g., Hatfield et al., 2014; Thornton & Tamir, 2020; Zhu et al., 2022) suggesting the contagion of emotions or mental states. Together, these findings indicate the importance of regulating students’ negative emotions, which tend to lead to other negative emotions without external support because students may not be competent to move out of these negative emotions by themselves.

Looking at Paths 1, 2, 3, and 4 (joy–curiosity–confusion–joy, joy–curiosity–confusion–boredom, joy–curiosity–confusion–boredom–frustration–surprise-anxiety, and joy–curiosity–confusion–boredom–frustration–surprise-anxiety–joy, respectively), we found that they all started with joy–curiosity–confusion and then developed in different directions: for example, immediately leading to joy or boredom or passing first through boredom–frustration–surprise and then ending with anxiety or joy. These results indicate that the students usually felt joyful and curious when they started an inquiry thread. Then they might feel confused due to the cognitive conflict or cognitive disequilibrium made salient by public community knowledge or as they further explored inquiry questions. Different emotional results might arise depending on whether this confusion can be resolved quickly. Previous studies (e.g., D’Mello et al., 2014; Worsley & Blikstein, 2015) have recognized the importance of confusion to students’ learning, finding that it can facilitate or hinder learning depending on how it is scaffolded. For instance, Worsley and Blikstein (2015) found that being in or transitioning to an expression of confusion was related to positive learning outcomes, whereas Pardos et al. (2013) found negative emotions, such as confusion and boredom, were negatively associated with learning gains. The paths identified in the current study help explain why confusion leads to different learning outcomes by triggering different subsequent emotions.

The path of curiosity–boredom suggests that some ideas or questions initially catch people’s interest, but this interest fades. Admittedly, there are diverse ideas contributed by various participants in Knowledge Building, but community members can not take up all ideas. Therefore, previous studies have developed tools (e.g., the Promising Tool, the Idea Thread Mapper, and Reflective Assessment, Chen et al., 2015; Yang et al., 2016; Zhang et al., 2018) to support students’ reflections on their community knowledge status and identify their knowledge gaps. For instance, the Promising Tool allows students to select the ideas that they think would lead to productive results and deserve more time and effort (Chen et al., 2015). However, as previous studies have focused on the socio-cognitive aspect, students’ emotions expressed in inquiry threads continuously improved or ignored by their Knowledge Building community are unclear. This study has addressed this research gap and suggests the importance of caring about students’ expressed emotions while advancing community knowledge. For instance, our interviews with primary school students suggest that the students felt bad if their ideas were not responded to by others (Zhu et al., 2020). More research is needed to examine how to develop a safe environment where students feel comfortable with some ideas not being taken up for community knowledge advancement.

Implications for educational practices

This study has several implications for facilitating low-achieving students’ productive academic emotions and Knowledge Building. First, appropriate scaffolding should be developed to help students deal with unproductive emotions. The participants in our study experienced a certain level of anxiety, frustration, and boredom. There is evidence that student-directed reflective assessment is an effective scaffolding strategy for helping students engage in productive Knowledge Building (Yang, 2019; Yang et al., 2016, 2020), and also has the potential to support students’ supportive emotions (Zhu et al., 2022) because it supports students’ control over learning and enhances the perceived value of their Knowledge Building inquiry (Pekrun, 2006). Therefore, future research may consider examining the impact of student-directed reflective assessment on students' emotions.

Second, the emotion evolution model has the potential to inform teachers about dealing with students' various emotions. For instance, it suggests that confusion may be followed by joy if it is resolved in a timely manner but might lead to boredom if it lasts too long. Therefore, on the one hand, teachers can help increase students’ tolerance of and ability to cope with confusion and negative emotions such as boredom and anxiety. For instance, teachers may convey a notion of the importance of different emotions, such as confusion, in students’ learning to help them be more comfortable when experiencing such emotions. On the other hand, teachers may need to help if students have been stuck with an idea and associated negative emotions (e.g., boredom, frustration, anxiety) for a while or have been transiting between these emotions, for example, by talking with students to understand their challenges. This is because young students tend to have difficulty managing their negative emotions (Boekaerts, 2011), which may lead them to experience the same or similar negative emotions. Without external support, students may be unable to move out of these negative emotions by themselves.

Third, it is fundamental to develop a collaborative and reflective culture among students; this culture is critical for helping students experience positive emotions and cope with negative emotions. Knowledge Building emphasizes students’ collective effort and treats each student as a legitimate contributor to advancing community knowledge (Scardamalia, 2002), not only in the cognitive aspect but also in the social and emotional aspects. Previous studies found that in 75% of face-to-face Knowledge Building discourse, at least half of grade 2 and grade 3 students contributed ideas vocally (Zhu et al., 2020, 2022); in their reflections, grade 2 students expressed empathy for those whose ideas were not responded to by the community. Therefore, we conjecture that in a supportive and reflective Knowledge Building environment, most students will actively participate in advancing community knowledge and help each other cope with negative emotions. The potential relationship between students’ differences and their roles in emotion regulation needs to be explored.

Finally, although the pedagogical design was developed to help low-achieving students collectively advance their knowledge-building discourse, we believe it may have important implications for the design of technology-rich environments in the collaborative inquiry context. We also hope this study will enhance educators’ understanding of low-achieving students' academic emotions in knowledge-building classrooms, along with constructive strategies for integrating knowledge-building models into traditional classroom curricula.

Limitations and future directions

There are several limitations to this study. First, we analyzed students’ emotions based on their written Knowledge Forum notes. There are gaps between students’ emotions expressed through text and their subjective feelings. Furthermore, students may vary in expressing emotions even if they have similar subjective feelings, and some may disguise their expressions of emotion (Ray & Chakrabarti, 2016). Future research should investigate how to accurately and reliably measure students’ emotions in collaborative inquiry. Moreover, manually coding students’ emotions is tedious work. Future research might explore how to build computing algorithms to analyze students’ emotions.

Second, we only examined the expressed emotions in the student’s written notes on Knowledge Forum. However, in Knowledge Building, it is critical to understand the emotions of all students over time, especially considering that students who feel bored, frustrated, or anxious at a certain time may not join in the Knowledge Building discourse. Graesser et al. (2014) suggested tracking moment-to-moment emotions using emotion-tracking technologies over longer intervals. It is important to detect prolonged periods of boredom, frustration, and confusion and to develop pedagogical interventions to address these issues to avoid student disengagement (D’Mello & Graesser, 2012). Future research may use wearable devices (e.g., wristbands that measure electrodermal activities; see Lee et al., 2019) to measure students’ physiological features to understand their emotions even when they do not speak or write.

Third, we focused on community-level emotions. However, it is also crucial to understand the individual and group level emotions that low-achieving students experience and their emotion evolution trajectory in order to provide adaptive and personalized support for effective emotion regulation. Future research might explore the emotions that low-achieving students express at the individual and collective levels and the factors contributing to these differences. Moreover, further research might explore how to scaffold low-achieving students' productive emotion regulation and how to help them develop regulation skills.

Finally, the evolution model of low-achieving students’ emotions in Knowledge Building was constructed based on the Knowledge Forum data from four classes of low-achieving students from Hong Kong. Our preliminary validation results of the model in two other low-achieving classes from Hong Kong suggest the explanatory power of this model. However, further research should systematically test this model with different students and more data. Furthermore, given that emotions are usually expressed differently in different cultural contexts (Elfenbein & Ambady, 2002), the extent to which the model may explain the emotional evolution of students other than in Hong Kong needs to be further researched.

Conclusions

Using the emotions extracted from the online Knowledge Forum notes of four low-achieving classes from Hong Kong, this study constructed the emotion evolution model of low-achieving students in Knowledge Buildings. This article is the first to examine low-achieving students' emotion categories, transitions, subsequences, and evolution model in Knowledge Building; this population has rarely been investigated. Such a model not only extends our understanding of low-achieving students’ typical emotion paths in Knowledge Building, but also contributes to the academic emotion literature. Despite being constructed using data from low-achieving students in a specific cultural and educational context, this model has the potential to explain other young learners’ trajectory of emotions as they engage in Knowledge Building and even other forms of collaborative inquiry. This study sheds light on the design of classroom practices for facilitating students’ engagement in productive emotions and collaborative inquiry.