University Students’ Knowledge Creation Practices in Face-to-Face and Hybrid Blended Learning: Development of Epistemic Views and Perceptions of the Community of Inquiry

Garrison et al., (1999, 2001) proposed the CoI framework, which comprises three types of presences deemed necessary for designing effective blended learning instruction to achieve productive outcomes. Among these, cognitive presence is considered the most challenging aspect in designing online learning environments (Celani & Collins, 2005; Garrison & Cleveland-Innes, 2005). It reflects the extent of learners’ agency in their learning process, such as constructing understanding through reflection. This process involves several steps: (1) triggering events, (2) exploration, (3) integration, and (4) resolution. In this process, learners initially identify problems for further inquiry, engage in individual and collaborative activities to explore these problems, integrate their ideas with scaffolding from teachers to elevate their thinking to a more sophisticated level, and apply their newly acquired knowledge to other contexts. This application helps them assess the appropriateness of their knowledge development.

Social presence, encompassing affective expression, open communication, and group cohesion, reflects how learners can perceive their social and emotional belonging within a community or group and be recognized as individuals worthy of communication. While numerous studies have highlighted a strong correlation between social presence, learning performance, and learner satisfaction, some research has delved deeper into the interrelationship between cognitive and social presence. For instance, Beuchot and Bullen (2005) conducted a longitudinal study analyzing messages posted in graduate students’ asynchronous forums, quantitatively and qualitatively. Their study focused on how interpersonal communication dynamics related to student interactivity. The results indicated that student participation increased, and discussions became more in-depth when students made concerted efforts to establish their interpersonal presence.

Teaching presence is the instructional component of the CoI framework. While cognitive and social presences are essential, they alone are insufficient to engage learners effectively in online learning (Garrison et al., 1999). Initially, the instructor must design the course structure. This includes outlining how learners initiate their learning journey, sustain their inquiry over time, interact with others for collaborative inquiries, and how their performance is evaluated (Swan, 2003). Secondly, learners are expected to engage in collaborative discourse to share and construct their ideas. To facilitate productive discourse, instructors must monitor the discussions and guide the direction of learners’ discourse (Xin & Feenberg, 2006). Lastly, as subject matter experts, instructors should provide feedback on learners’ participation in discussions and their understanding through direct instruction. This role does not imply exerting authority but instead supporting learners in reflecting on their discourse from an expert perspective.

Recent studies have provided robust evidence that teaching presence is a critical factor influencing learners’ perceptions of their cognitive and social presence (Kanuka et al., 2007; Swan & Shih, 2005; Varnhagen et al., 2005; Xin & Feenberg, 2006). Garrison et al. (2010) conducted an exploratory factor analysis and structural equation modeling based on responses from 287 learners to 34 items in a CoI survey. This research aimed to test hypotheses regarding the interrelationships among the three presences. The results statistically confirmed the three components of the presence and elucidated the hypothesized causal relationships among teaching, social, and cognitive presence. The findings highlight that teaching presence is crucial for facilitating social and cognitive presences and that social presence forms the foundation for developing cognitive presence.

Future research should focus on more rigorous empirical studies to design blended learning courses that enhance social presence, which could foster appropriate cognitive presence. Additionally, a systematic review of CoI studies by Stenbom (2018) suggests new methodological approaches beyond surveys to examine the interrelationships among the three presences, aiming to create more effective educational experiences within the CoI framework. In this study, we approach the design of CoI in blended learning by applying a methodology in the learning sciences to systematically examine how instructional design components, such as teaching presence, influence students’ perceptions of cognitive and social presence.

In the field of learning sciences, researchers have developed a range of theories and technologies to support learners in engaging productively with their inquiries within communities, such as classrooms (Cress et al., 2021; Fischer et al., 2018). Design-based research has emerged as the primary methodological approach in this domain, guiding the development and analysis of educational practices and interventions (Hoadley & Campos, 2022; McKenney & Reeves, 2013, 2018; Puntambekar, 2018). In this study, we propose design-based research as an approach to design and evaluate instructional designs of CoI in two formats of blended learning in higher education.

Sandoval (2004, 2014) described design-based research as a methodology for investigating hypotheses in the design of learning environments, such as blended learning. The methodology involves three main steps. Firstly, researchers should develop high-level conjectures, often referred to as design principles, regarding supporting learning in their specific context to design embodiment in a learning environment. The embodiment consists of tools and materials, task structures, participation structures, and discursive practices. Secondly, the components in the embodiment are expected to facilitate mediating processes in which learners should be involved. Finally, these mediating processes should lead to the learning outcomes that learners are expected to achieve. During the second step, researchers need to assess whether their embodiment could support the intended mediating processes. Otherwise, the final learning outcomes cannot be attributed to the embodiment. This is known as a design conjecture (Sandoval, 2004, 2014). Researchers iteratively refine their embodiment to effectively facilitate mediating processes. Once there is a meaningful link between the embodiment and mediating processes, researchers further evaluate whether the expected mediating processes lead to appropriate learning outcomes for the learners. This process is called a theoretical conjecture.

This study uses the design-based research methodology to design and examine blended learning. While the CoI framework has been helpful in evaluating blended learning, it has yet to provide guiding principles for embodiment. To overcome the problem, we use knowledge-building as a guiding principle. In design-based research, teaching presence is considered embodiment. This includes the structuring of tasks; the participatory structure of learning at individual, group, and classroom levels; and the use of technologies to facilitate learners’ inquiry (Sandoval, 2004, 2014). Within this framework, we explore how the embodiment contributes to establishing a social and cognitive presence in the CoI as a mediating process. We analyze learners’ activities during the course, such as discourse or transcripts, and artifacts created during activities to understand this conjecture. Furthermore, the connection between learning performance and educational experiences based on teaching, social, and cognitive presences is also a critical evaluation area for course design. Through the iterative process of practice-based studies, there is an opportunity for the interdependent refinement of theories and instructional designs. Because this study is the initial step of design-based research, we only focused on design conjecture.

The concept of learning as a process of knowledge creation has emerged as a novel approach in educational research. This idea was highlighted in the works of Paavola and Hakkarainen (2005) and Paavola et al. (2004), following Sfard’s (1998) discussion of two primary learning metaphors. This knowledge-creation metaphor is distinct from the other two metaphors—knowledge acquisition and participation. It focuses on developing the learner’s ability and confidence in generating new knowledge rather than just accumulating or integrating it into established cultural practices. The approach emphasizes inquiry as the central learning mechanism, differentiating it from the other two metaphors. Furthermore, this metaphor involves learners actively engaging in their inquiry within a community, highlighting the collaborative aspect of knowledge creation.

In this study, we adopt the perspective of learning as knowledge creation, aligning it with the cognitive presence in the CoI framework. This approach is informed by the concept of knowledge-building as proposed by Scardamalia and Bereiter (2021, 2022). Within the CoI framework, which includes triggering events, exploration, integration, and resolution, the knowledge-building perspective offers valuable conceptual tools. These tools facilitate an understanding of the teaching presence, emphasizing inquiry as a process of idea improvement and fostering learners’ epistemic agency to advance their inquiries.

Scardamalia and Bereiter (2021, 2022) have conceptualized learning as knowledge-building, drawing on their extensive research into human expertise (e.g., Bereiter & Scardamalia, 1993). They particularly emphasized two aspects of cognitive presence within the learners’ inquiry process, as framed by this knowledge-building approach. Firstly, in inquiry as knowledge-building, learners engage in refining their ideas. These ideas, defined as forms of thought developed through integrating various knowledge types during the inquiry process, aim to solve specific problems. Consequently, teaching presence should support learners in externalizing, sharing, and collaboratively examining their thoughts throughout this process. Secondly, learners’ agency in idea improvement through collaboration is crucial to their inquiry process (Scardamalia, 2002). This epistemic agency can be nurtured by designing task structures emphasizing inquiry as a process of idea improvement. Supporting this, recent studies by Chen et al. (2015) and Chen (2017) demonstrate that instructional support, specifically in helping elementary school students evaluate the potential of their ideas within collaborative discourse, effectively engages learners in their inquiry process.

As discussed, it is necessary to evaluate different aspects of learning when learners are expected to engage in knowledge-building practices: idea improvement and agency. While idea improvement in knowledge-building practices has been discussed in many studies (Oshima et al., 2012, 2019; Tong & Chan, 2023; Yang et al., 2022), it has not been sufficiently discussed how learners’ agency can be exerted in their practices.

Learners’ agency in knowledge-building practices can be examined from several different approaches. Firstly, the learners’ agency can be analyzed by interpreting their discourse during practice. Damşa (2014), for example, delved into the efficacy of interactions within small-group learning at the undergraduate level, focusing on the concept of shared epistemic agency—a continuation of the discourse from Damşa et al.’s (2010) work. This study leveraged qualitative methodologies alongside quantitative analysis to scrutinize how student collaborations form and evolve knowledge artifacts, like research reports, throughout a 20-week project. The findings indicated a spectrum of constructive interaction types and underlined that groups with a collective cognitive commitment were more adept at forging comprehensive and relevant knowledge artifacts. This research offers a nuanced perspective on collaborative learning mechanisms and the evolution of knowledge artifacts, providing pertinent implications for pedagogical strategies and future academic research.

An alternative method involves using rubrics to evaluate interview responses or discourse to explore learners’ evolving perceptions of collaborative learning (Lee et al., 2021). Drawing on Popper’s (1979) classification of knowledge across three worlds, Hong et al. (2016) devised a set of rubrics to assess learners’ epistemic views of knowledge. World 1 is the realm of physical entities and conditions, spanning the entire gamut of material existence from inanimate objects to celestial bodies, including humans. It is the world of observable and quantifiable phenomena. World 2 pertains to the internal experiences unique to each individual, encompassing thoughts, emotions, sensations, and awareness. These personal experiences, while subjective and not directly observable by others, are nonetheless significant to the individual. World 3 is the repository of abstract creations of the human intellect—once formulated, these entities, such as scientific theories, literary works, and social structures, persist beyond their originators’ cognition. Unlike the tangible nature of World 1 or the subjective nature of World 2, the constructs of World 3 are distinguished by their capacity for collective scrutiny, discussion, and refinement, constituting the shared heritage found in educational materials and cultural repositories. Knowledge-building is an intellectual practice for a community to improve objective knowledge, such as ideas in World 3, and learners who engage in the knowledge-building practice are expected to develop their views of knowledge in World 3.

Employing a mixed-methods strategy, Hong et al. (2016) involved 41 undergraduates from a university in Taiwan enrolled in a course over a semester to enhance collaborative knowledge-building. This approach integrated surveys to evaluate the students’ epistemic perspectives by examining online interactions on the Knowledge Forum platform, thereby classifying participants labeled as “more informed” and “less informed” based on their preliminary grasp of knowledge creation. The findings indicated notable advancements in the students’ perceptions aligned with World 3, signifying a more profound comprehension of knowledge as a collective endeavor. Despite high levels of participation, the volume of input was not necessarily indicative of epistemic development; instead, the emphasis on the quality of ideas’ integration and examination was pivotal. Groups identified as more informed exhibited superior consolidation of concepts, highlighting that successful knowledge construction demands more than mere participation—it necessitates careful amalgamation and application of concepts to cultivate an advanced understanding of knowledge dynamics. Thus, the rubrics developed by Hong et al. (2016) successfully captured the shift in the development of students’ epistemic views of knowledge in their learning from World 2 to World 3.

In studies on the CoI, evaluating the process of learners’ inquiry over time has been identified as a challenge. In addition to assessing students’ perceptions of presences using the CoI survey, recent studies have proposed transcript analysis—specifically, the analytics of students’ discourse in their inquiry over time (Kovanović et al., 2019; Rolim et al., 2019; Stenbom, 2018). For instance, Rolim et al. (2019) examined how learners engaged in their inquiry process from the perspectives of cognitive and social presence by using a recently developed network analysis method, epistemic network analysis (ENA) by Shaffer and colleagues (e.g., Shaffer, 2017; Shaffer et al., 2016).

ENA was initially developed to visualize and examine epistemic frames seen in specific cultural communities and cultural grammar. According to Shaffer and colleagues (Shaffer, 2017), communities of practice exist as cultures with their unique grammar, which is comprised of skills, knowledge, identity, values, and epistemology. With discourse as data, the meaningful relationship between codes representing components of cultural grammar is calculated by using the co-occurrence of the codes within a unit of analysis (Shaffer et al., 2016). Firstly, based on an original dataset of codes, adjacency matrices for meaningful sequences of discourse exchanges called stanzas are created. Then, a cumulative adjacency matrix \({C}^{\mu }\) is calculated by accumulating all the adjacency matrices for each group. In, each cell (\({C}_{i j}^{\mu }\)) is the number of units in which a pair of codes (i and j) co-occurred. The cumulative adjacency is converted into vectors that are further normalized through singular value decomposition, a technique similar to that used in principal component analysis. Finally, another optimization routine is performed for positioning network nodes (codes) (Shaffer et al., 2016). The advantage of ENA in evaluating learners’ perceptions of presence in their CoI is to identify their perceptions as integrative epistemic frames. With the calculated epistemic frames, frames of different groups or conditions can be statistically compared.

The ENA approach taken by Rolim et al. (2019) was informative in evaluating the designed online learning environment by identifying epistemic frames of learners’ inquiry practices over time through coding their transcripts from both perspectives of cognitive and social presence in CoI. Results revealed that the interactive category of indicator of social presence was more strongly connected with the early phases of cognitive presence, such as triggering events and exploration, while the affective category of the social presence related to later phases of cognitive presence (integration and resolution).

In this study, we aimed to consider an instructional design for blended learning in the post-COVID era by leveraging the infrastructure developed during the years of the COVID pandemic through design-based research. To do so, we conducted design conjectures for two blended learning instruction formats using CSCL (Computer-Supported Collaborative Learning) technology, guided by knowledge-building principles (Fig. 1). The first format was traditional blended learning, wherein students interacted face-to-face in class and engaged in asynchronous communication within the CSCL environment. We have designed the traditional blended learning format for years and found it worked to facilitate students’ knowledge-building practices (e.g., Splichal et al., 2018).

Full size image

The second format was a hybrid blended learning model conducted in another year, allowing students to choose their participation mode, involving in-person and online attendance, complemented by asynchronous communication in the CSCL environment. It is a new format of blended learning, and many studies have not yet discussed hybrid blended learning. We had an opportunity to design a hybrid blended learning format in the late phase of the pandemic era. Students with masks as a requirement could come and communicate face-to-face. However, they could participate in the course remotely through the Zoom online conference system because they had to stay home to recover from their symptoms or for family issues. Despite the different communication formats over 2 years, we maintained consistent task structures, participation structures, and the CSCL system. Under the conditions we described, we compared the two blended learning formats by conducting our design conjectures to examine the influences of two different communication modes (traditional and hybrid) as participatory structures on two mediating processes, the development of students’ epistemic views and perceptions of CoI (cognitive and social presence) addressing the following research questions within and between the formats of blended learning:

Based on the findings, we discuss how to improve both blended learning formats to facilitate students’ knowledge-building practices.

In this study, students’ individual reflection notes every week were used as data. Their written discourse was coded with two rubrics to answer our research questions.

Regarding students’ epistemic views of knowledge, we developed rubrics to evaluate students’ reflections by adapting the original coding scheme by Hong et al. (2016) to suit our course activities. The coding scheme from Hong et al. was structured around six criteria, with three reflecting the World 2 view and three reflecting the World 3 view of knowledge. In this study, we focused on the developmental trajectory of the World 3 view of knowledge. While Hong et al. (2016) examined if their engagement in knowledge-creation practices would facilitate the development of students’ epistemic views as learning outcomes before and after their engagement in knowledge-creation practices by evaluating both World 2 and 3 views, we were more concerned with the development of students’ World 3 view of knowledge as a mediating process in our design conjecture, i.e., how our embodiments in our instructional design would facilitate our expected mediating process. The World 3 view encompassed criteria such as (1) knowledge as concrete products of students’ thought processes, (2) ideas formed through students’ interactions with their environment, and (3) the development of ideas through collaborative scrutiny (see Table 1). Reflection notes were coded 1 when each of the three perspectives was identified. Two independent raters coded students’ reflections, achieving Cohen’s kappa (κ) of 0.92 and 0.76 for the traditional blended learning year and the hybrid blended learning year, respectively. Discrepancies were resolved through raters’ discussions.

Students’ perceptions of cognitive and social presence within a CoI framework were evaluated using a modified version of the coding scheme developed by Rolim et al. (2019) comprised of the foundational categories of cognitive presence (Garrison, 2007) and social presence (Rourke et al., 1999). We refined Rolim et al.’s scheme by distinguishing between social presence in thought processes (“cognition”) and behaviors (“action”). The original categories, such as “the use of humor,” “continuing thread,” and “contribution,” were bifurcated into aspects of (1) cognition, where codes were intended as a strategy, and (2) action, where they were actively employed in collaborative efforts (see Table 2). Each reflection note was assessed by two independent raters, who coded the presence of these adjusted categories as “1” for observed and “0,” achieving Cohen’s kappa (κ) of 0.64 and 0.79 for the year of traditional blended learning and the year of hybrid blended learning, respectively. Discrepancies were resolved through raters’ discussions.

The first research question focused on our design conjecture regarding the development of students’ views on knowledge as a mediating process. We anticipated that students participating in the designed blended learning environment would continuously develop their understanding of knowledge from a World 3 perspective. To examine the developmental trajectories of students’ knowledge views, we initially conducted a clustering analysis of their knowledge view scores to identify the patterns of developmental trajectories they followed. Following this, we performed ANOVAs on the mean knowledge view scores within the identified clusters to determine significant differences in these trajectories. These analyses enabled us to explore how many students adhered to our expected mediating processes and what alternative mediating processes we needed to consider for further refining our blended learning designs. Finally, we discussed our findings within each blended learning format as well as between the two formats.

The second research question focused on our design conjecture regarding students’ perceptions of CoI, specifically cognitive and social presences. If our teaching presence within the CoI framework was effective, students would successfully perceive both cognitive and social presence. To explore how students experienced CoI during their learning, we conducted ENA to identify connections between cognitive and social presence codes. The ENA results also revealed differences in how students perceived CoI across two blended learning formats. We examined students’ perceptions of CoI in each blended learning format and the variations in their experiences to identify further improvements for the designs.

To answer our first research question, we coded students’ reflection notes according to the rubrics of epistemic views of knowledge. Each student’s epistemic view score was calculated as the averages of the first, second, and third modules each. Thus, each student’s epistemic view of knowledge was represented by a string of three scores ranging from 0 to 3. Based on the epistemic view scores, we conducted a hierarchical clustering analysis using the Ward method to identify groups of students who manifested different trajectory patterns. Four clusters were identified each year based on the dendrograms (Figs. 4 and 5). Students were grouped into Cluster #1 (n = 10), Cluster #2 (n = 32), Cluster #3 (n = 11), and Cluster #4 (n = 21) in the year of traditional blended learning and into Cluster #1 (n = 13), Cluster #2 (n = 34), Cluster #3 (n = 5), and Cluster #4 (n = 22) in the year of hybrid blended learning.

Full size image

Full size image

We conducted ANOVAs to examine how different the developmental trajectories of students’ epistemic views over three modules were among the identified four clusters each year. A 4 (Clusters) X 3 (Modules) ANOVA on the scores in the year of traditional blended learning revealed two significant main effects, F (3, 70) = 24.30, p < 0.01, partial η²  = 0.51 for the cluster effect, F (2, 140) = 12.66, p < 0.01, partial η² = 0.43 for the module effect, and the interaction effect, F (6, 140) = 7.73, p < 0.01, partial η² = 0.58 (see Fig. 6). Further, post-hoc comparisons using Tukey’s LSD test were conducted to examine the interaction effect. The results were summarized as follows. Students in Cluster #1 and #2 were identified to develop their epistemic views over the modules based on the results showing that the mean score in Module 3 was significantly higher than that in Module 1 in Cluster #1 and that scores increased significantly over the three modules in Cluster #2. Students in Cluster #3 and #4 were identified to be partially successful. Students in Cluster #3 could not develop their epistemic view from modules 2 to 3. Students in Cluster #4 could not sustain their epistemic view scores in Module #3. Their mean score in Module 3 was significantly lower than in other modules.

Full size image

A 4 (Clusters) X 3 (Modules) ANOVA on the scores in the year of hybrid blended learning revealed two significant main effects, F (3, 70) = 46.72, p < 0.01, partial η² = 0.67 for the cluster effect, F (2, 140) = 3.69, p < 0.05, partial η² = 0.05 for the module effect, and the interaction effect, F (6, 140) = 2.87, p < 0.05, partial η² = 0.11 (see Fig. 7). Further, post-hoc comparisons using Tukey’s LSD test were conducted to examine the interaction effect. The results were summarized as follows. Students in Clusters #1 and #2 were found to fail to develop their epistemic views at all through three modules. There were no significant differences in the scores over the three modules. Those in Clusters #3 and #4 were identified as partially successful. They could develop their epistemic views from Modules 1 to 2 but could not continue the trend from Modules 2 to 3.

Full size image

We identified distinct patterns in the developmental trajectories of students’ epistemic views of knowledge over 2 years. During the year of traditional blended learning, two patterns emerged: (1) a consistent increase in scores and (2) an initial increase or sustainment followed by a decrease in scores. Conversely, in the year of hybrid blended learning, we observed two distinct trajectories: (1) no change in scores and (2) an increase in scores then plateaued.

To address our second question, we conducted an epistemic network analysis (Shaffer et al., 2016) to examine students’ perceptions of the CoI within each blended learning format and the differences between the two formats. To examine differences in students’ perceptions of CoI between the two blended learning formats, we first set our ENA model with students in different years (formats) as units to see how students in the two blended learning formats perceived CoI. We compared the means of vectors of students between the two blended learning formats in two dimensions of the epistemic network space. Second, we set phases (modules) as conversations in each of which cooccurrences of codes were calculated. In our designs, the composition of students in groups was changed across the three modules (phases) from the original jigsaw groups through the expert group back to the jigsaw group. Therefore, the module (phase) was the appropriate conversation. Finally, we applied the sliding stanza window (its size was 4) to detect the epistemic frame from students’ dialogue in the CSCL system. They engaged in asynchronous dialogue by reading and commenting on each other in the CSCL system. Their perceptions of CoI were assumed to be constructed through such a CSCL discourse. We set the bandwidth of the dialogue, named Stanza in ENA, as four inputs because of the size of the groups (three or four group members).

Figures 8 and 9 show epistemic networks of students’ perceptions of CoI in years of the traditional and the hybrid format of blended learning, respectively. The first dimension (MR1) explained 23.9% of the variance, and the second dimension (SVD2) explained 8.3%. The goodness of fit for both dimensions was high (Spearman correlation coefficients were 0.99 for both). Figure 10 illustrates the comparison network between the years. On the X-axis (MR1), a Mann–Whitney U test indicated a significant difference between the year of traditional blended learning (Mdn = 0.30) and the year of hybrid blended learning (Mdn =  − 0.31), U = 6.00, p = 0.00, r = 1.00. No significant differences were observed on the Y-axis (SVD2). In both years, students’ perceptions of the cognitive presence codes were interconnected and connected to social presence codes, and the findings revealed that students understood what they were supposed to do to attain their cognitive goals in their activities. However, there were critical differences in their perceptions between the years. Students in the year of traditional blended learning perceived cognitive presence more strongly than those in the year of hybrid blended learning. Further, students in the year of hybrid blended learning perceived stronger and more connections between cognitive and social presence. Figure 10 shows that students in the year of hybrid blended learning perceived connections between their cognitive presence and codes such as “Vocatives,” “Quoting message,” and “Contribution_action” that had not been seen in the year of traditional blended learning. Thus, during their cognitive engagement in the courses, students in the year of traditional blended learning were more focused on cognitive presence in their activities, while those in the year of hybrid blended learning paid more attention to their social presence.

Full size image

Full size image

Full size image

To examine our first research question, we identified clusters of developmental trajectories of students’ epistemic view scores through three modules. In the year of traditional blended learning, we found a group of students (n = 42) who consistently developed their World 3 views (Clusters #1 and #2) and another group (n = 32) who failed to improve their views in the third module (Clusters #3 and #4). In the year of hybrid blended learning, different patterns emerged. One group of students (n = 27) improved their views from the first to the second module but not to the third module (Clusters #3 and #4). Then, another group (n = 47) did not manifest any improvement at all over the three modules (Clusters #1 and #2). The results show that while the instructional design of CoI in traditional blended learning succeeded in facilitating our expected development of students’ epistemic views over the course (57% successful and 43% partially successful), the same instructional design in hybrid blended learning did not (36% partially successful but 64% unsuccessful). Those findings suggest we must consider a different approach to the embodiment rather than replacing face-to-face communication with its complementary mode, such as hybrid. Here, we discuss how our embodiments in the instructions facilitated the development of students’ epistemic views in traditional blended learning and how such a mediating process was not triggered in hybrid blended learning.

A partially successful group of students in the traditional and hybrid blended learning failed to sustain their development of epistemic views in the third module. The third module required students to discuss their ideas of new happiness indices from different perspectives learned in the second module. They were required to make their own knowledge objects in the third module for the first time. In the previous modules, students were encouraged to scrutinize others’ knowledge objects provided by the instructional team. Since they were given concrete products to share and discuss in their groups, they might be more easily engaged in the knowledge-building practices by exerting their World 3 view of knowledge. We expected that their experiences in the first two modules would support students’ World 3 knowledge view development in the third module. More than half of them were found to develop their World 3 views further in traditional blended learning, but not all. This suggests that we need to consider further embodiment to support students in creating their new knowledge objects to share and improve in their group activities. Further detailed comparative analysis of discourse in the third module between the successful and the partially successful groups should be conducted to develop new embodiment. Our current hypothesis of the difference between the two groups is whether they meaningfully used their learned knowledge in the first two modules to create their new ideas of happiness indices by identifying remaining problems in the current indices and discussing how to overcome them.

In the hybrid blended learning format, more than half of the students failed to develop their World 3 views of knowledge. The results suggest our embodiment did not function at all for the group of students. However, it must be worth discussing why we saw this group of students only in the hybrid blended learning format. Based on the differences in embodiment between the two formats, it may be reasonable that the hybridity in the embodiment influenced the development of students’ epistemic views. Hybrid communication might give students a new challenge because they have to manage different communication channels simultaneously (e.g., Zenouzagh Mohammadi et al., 2024). For remote students, catching up with the discussion unfolding on the face-to-face site had to be challenging. When they had difficulties listening to and commenting on ideas on-site, they had to emphasize their conversational turns more than when doing so on-site. It might also be another challenge for on-site students to care for their remote peers so that they can catch up with on-site discussions. Monitoring and adjusting progress in their discourse in such a hybrid communication might take more cognitive load than usual (Zenouzagh Mohammadi et al., 2024) and consequently reduce their engagement in idea improvement. To solve the problem, we must consider effectively using multimodal communication channels to empower students’ discourse in knowledge-building practices. Text-based asynchronous communication could be valuable for on-site students to reflect on their discussion and for remote students to catch up with the discussion. The instruction to ask students to report their on-site discussion in detail in the Knowledge Forum may facilitate their in-depth discussion across different communication channels and develop their epistemic views by reducing their cognitive load.

This study provided a more detailed analysis of the epistemic view development during learning beyond Hong et al. (2016). The results of the partially successful student groups suggested a pivotal phase to divide students into successful or less successful groups. In the knowledge-building practice, learners shift from a phase of learning what others found to another phase of creating their new knowledge objects. While understanding knowledge objects developed by others, students can engage in their activities as if they have the World 3 view of knowledge because they are given knowledge objects by others. The critical timing, however, should be in the phase where students must develop their own knowledge objects based on what they learned.

Our findings further suggest the critical influences of communication modalities on students’ knowledge-building practices. Multimodality in their group work is considered to increase the cognitive load for students to engage in their interaction (Zenouzagh Mohammadi et al., 2024). The more load they allocate for their interaction, the less epistemic engagement is expected. In the context of multimodal communication, such as hybrid blended learning, it was found more support should be invented to effectively use the modalities to conduct their knowledge-building practices.

To answer our second research question, we examined students’ perceptions of CoI (cognitive and social presences) based on our design conjecture. ENA revealed that students in both years perceived all phases of cognitive presence in their group activities, suggesting that students attempted to engage in our designed knowledge-building practices. To engage in their knowledge-building practices, students were found to have to pay attention to different social presences between traditional and hybrid blended learning. As the comparison network showed in Fig. 10, students in the year of traditional blended learning were more focused on cognitive presence, such as “Triggering_events” and “Exploration,” and a social presence, “Contribution_cognition.” Their stronger perception of the two phases of cognitive presence suggests that students in the year of traditional blended learning were actively engaged in the initial phase of their inquiry process. These phases should be critical in creating happiness indices as new knowledge objects in the third module. Furthermore, those cognitive presence codes were connected to a social presence “Contribution_cognition.” This social presence code represents students’ descriptions of their intentions to contribute to their group in reflection notes, and the result reveals that students in the year of traditional blended learning more deeply considered how they engaged in their group activities, particularly in the initial phases of their inquiry compared with those in the year of hybrid blended learning.

Compared with those in traditional blended learning, students in the year of hybrid blended learning were found to perceive more aspects of social presence strongly. The tendency of such a strong perception revealed that they were more concerned with the social aspects of their group activities rather than cognitive ones. Some connections of social presence codes were found in students in traditional blended learning, but the strength was stronger in hybrid blended learning, suggesting that students in hybrid blended learning had to be more conscious about the social aspects than cognitive presence. The unique connection of social presence codes only seen in hybrid blended learning, such as “Quoting_Message,” “Vocatives,” and “Contribution_action,” revealed students’ intentional efforts to make their collaboration visible.

The results of students’ perceptions of CoI, together with those of the development of students’ epistemic views, suggest that the differences in perceptions of social presence may underscore the importance of establishing a strong group identity to facilitate more productive cognitive engagement by students (Miyake & Kirschner, 2014; Rolim et al., 2019). While many questionnaire studies have discussed the relationship between cognitive and social presence, more instructional design studies are needed to identify what causes missing links between the presence (Stenbom, 2018). This study applied the design-based research methodology to examine how teaching presence as embodiments triggered students’ knowledge-building practices and how teaching presence was related to cognitive and social presence. ENA results in this study could provide additional findings beyond Rolim et al. (2019) by designing the instructions for knowledge-building practices. As Rolim et al. (2019) found, we also saw the connection between codes in “Interactive” and “Cohesive” codes of social presence and early phases of cognitive presence codes like “Triggering Events” and “Exploration.” However, we did not find the connections between “Affective” codes and later phases of cognitive presence codes like “Integration” and “Resolution.” Rather, in the year of hybrid blended learning, the connections were found to sustain over all the cognitive presence codes, and no connections were found in later phases in the traditional blended learning. Thus, in the knowledge-building approach to traditional blended learning, students could focus more on their cognitive presence and develop their epistemic views in their inquiry process. In the hybrid blended learning format, in contrast, students had to pay more attention to social presence in their inquiry and failed to focus on their cognitive presence to develop their epistemic views.