Introduction

In recent years, Professional Learning Communities (PLCs) have served as central frameworks for Professional Development (PD) for teachers, and have been implemented alongside more traditional structures (Vangrieken et al., 2017). PLCs offer a flexible social-professional environment for collaboration, geared to its participants' needs. PLC providers should constantly evaluate the participants' attitudes, interests, and involvements so they will be able to effectively respond (in real-time) to unexpected changes, stay attuned and relevant and ensure fulfillment of the PLC's objectives. Yet, monitoring PLCs in the era of digital media and emerging online environments for professional development calls for new approaches. The replacement of face-to-face communication with convenient digital interactions creates huge and constant volumes of various data, requiring technological support to analyze and understand the meaning of the digital traces, which in turn could be translated into actionable insights for better decision-making.

The PLC in this study was active before and during the COVID-19 pandemic. Participants utilized the WhatsApp platform as their most prominent channel for spontaneous-unguided online communication, which gave us the opportunity to monitor the participants' performance and needs throughout the period, and to investigate the impact of COVID-19 on activity and performance.

In the following sections, we present a general background about teachers' PLCs, including some theoretical frameworks and models. Then, we elaborate on the WhatsApp environment as an online platform for teachers' PLCs, as well as a research data source. Finally, we address the context in which this study took place.

Background: Professional Learning Community (PLC)

Several theoretical and experimental models in the area of teacher professional development (Bolam et al., 2005; Borko et al., 2010; DuFour, 2004; Grossman et al., 2001; Little, 2012; Vescio et al., 2008), as well as our own research work in the field (Eylon et al., 2020), laid the ground for the following PLC definition, framework, and models.

PLC Definition

DuFour (2004) was among the first to introduce PLC as a framework for teacher professional development and identified three big ideas for PLC: (1) to ensure teacher learning, (2) to create a culture of collaboration, and (3) to focus on outcomes (such as the participents' achievements and attitudes). Shulman and Sherin (2004) claimed that active membership in a PLC can boost teachers' professional development, and will facilitate their goal of becoming accomplished teachers. There is no consensus in the literature regarding the definition of PLC. Influenced by the above, we define a PLC as: A framework for a group of educators to meet regularly and develop norms of trust and sharing. The educators actively investigate their teaching, collect evidence from their students’ learning, reflect collaboratively on their practice, and learn from one another.

Typical key characteristics of teachers' PLCs and their descriptions are presented in Table 1. The PLCs in our project followed these characteristics.

Table 1 Characteristics of teachers' PLC*
Full size table

PLC performance can be evaluated in the light of different theoretical frameworks such as the model of teachers knowledge (Shulman, 2015) or linkage between teacher knowledge progression and student learning (Fishman et al., 2003). Vangrieken et al. (2017) mentioned additional parameters, such as development of a discourse, tracing personal growth, meeting stakeholders interests, and following PLC characteristics (e.g., group dynamics, regularities). This study monitored teacher learning with regard to several aspects of teacher knowledge. The first three aspects correlate with Shulman's (1986, 2015) components of teacher knowledge: (a) disciplinary Content Knowledge (CK), which relates to the acquisition and expansion of content areas in the discipline, (b) Pedagogical Knowledge (PK), which relates to the art of teaching and instruction, learning and learners, class management, and assessment methods, and (c) Pedagogical Content Knowledge (PCK), which combines principles of meaningful teaching of disciplinary contents with concrete pedagogical aspects such as student's needs. Later, when technology became an inseparable component of education, Koehler and Mishra (2009) added (d) Technological-Pedagogical-Content-Knowledge (TPCK) component, which is the knowledge required for the successful integration of technology into teaching to ensure meaningful learning and optimal use of a specific technology. In the following years, TPCK grew to a major yet multifaced framework (Brantley-Dias & Ertmer, 2013). Angeli and Valanides (2009) interpreted TPCK as Information Communication Technology (ICT)-related PCK.

PLC Models

Two models emerged from our prior studies on teachers' PLC development and implementation (Eylon et al., 2020; Scherz et al., 2021): the Collaboration Model and the Network Model, both of which influenced subsequent PLC planning and implementation.

The Collaboration Model

Collaboration is defined as a “coordinated, synchronous activity that is the result of a continued attempt to construct and maintain a shared conception of a problem” (Roschelle & Teasley, 1995: 70). In former studies (Eylon et al., 2020; Scherz et al., 2021), we defined a Collaboration Model that specifies four levels of group effort performance, which differ in the intensity of required commitment and complexity (see Table 2).

Table 2 The four-level collaboration model
Full size table

The Network Model

The Network Model demonstrates relationships between individual participants within a specific teachers' community (PLC), as well as direct and indirect relationships between associated communities. Accordingly, the Network model represents various graphic paths of knowledge transmission as shown in Fig. 1 (Eylon et al., 2020; Scherz et al., 2021). This model is research-based and has emerged from many regional PLCs that were operated under the same project. An initial rigid top-down model gradually evolved into a Network Model that emerged alongside implementation, via analysis of documented evidence-based PLC stories. These PLC stories triggered various knowledge transmissions, both within a single PLC and between PLCs.

Fig. 1
figure 1

PLC Network Model: paths of knowledge transmission among teachers' PLCs ; (Source: Eylon et al., 2020 Scherz et al., 2021)

Full size image

WhatsApp as an Online Platform for Teachers' PLCs

The ever-growing popularity of social media — such as Facebook, Twitter, WhatsApp, and other platforms — has enabled researchers to scale-up and leverage data-mining techniques in order to collect and analyze the digital footprints left by users. Their importance stems from the fact they can serve as rich resources to characterize and detect various patterns regarding behavior, relationships, and social dynamics (Bakhshinategh et al., 2018; Nettleton, 2013; Romero & Ventura, 2010). Such patterns may be transformed into Learning Analytics, which can denote productive learning or its absence (Clow, 2013; Ferguson, 2012).

WhatsApp has been found to have many potential benefits for teacher development, especially within in-service programs and challenging periods, including the changes brought on by the global COVID-19 pandemic. Studies that focused on WhatsApp communication behavior reported an increased activity following the onset of the pandemic (Seufert et al., 2022; Tan et al., 2021). Regarding teachers' PLCs, English language teachers used their PLC WhatsApp group during remote teaching to share resources, raise questions, and propose solutions and teaching strategies (Defianty & Wilson, 2021). WhatsApp was also used to facilitate communication and coordination between groups of teachers in low-income settings (Varanasi et al., 2021). Discourse in WhatsApp groups of science teachers often relates to specific types of know-how, e.g., field knowledge, pedagogical content strategies, and in-school teaching practices (Cansoy, 2017; Waldman, 2020). In line with these reports, monitoring a WhatsApp group aimed at supporting language teacher development, identified three related themes of usage: interpersonal interactions, professional development, and organizational purposes (Motteram et al., 2020).

Besides its potential contributions to professional development, there are also important gains relating to socioemotional aspects and user well-being. For example, a PLC WhatsApp group reportedly helped teachers to develop a sense of community (Waldman, 2020). Similarly, using WhatsApp supported a social atmosphere that promoted collaboration among PLC members (Bouhnik et al., 2014; Naicker & Ebrahim, 2018; Zahedi et al., 2021). Yet, the effectiveness of using WhatsApp in teachers' PLCs depends on participants' awareness of the context within which the community exists, as well as their willingness to accept differing views and opinions (Moodley, 2019).

Context: Leaders PLC (L-PLC)

A PLCs project was established to create both nation-wide and region-oriented PLCs for STEM teachers, with the ultimate goal of improving student achievements and motivation to study science. The project was launched in 2016 with one Leaders PLC (L-PLC), whose members already served as mentors of regional PLCs. The L-PLC members were carefully selected from different regions of the country and were continuously tutored by the project supervisors, while in parallel, 20 regional PLCs were gradually established and co-mentored by these L-PLC members. Each PLC had 40 h of face-to-face meetings, 8 h of a full-day conference, and 12 h of independent learning (a total of 60 h per year).

Most meetings had a recurrent structure of: (1) a STEM and STEM-education-oriented session, (2) a PLC developing session that was dedicated to leadership, mentoring, and co-mentoring capabilities, and (3) a sharing session where participants shared practices and resources. After each meeting, the participants answered a feedback questionnaire that served for evaluation and to refine meeting contents and structure. Thus, the L-PLC meetings were also intentionally planned, implemented, and explicated as part of our modeling strategy, which was geared to develop the participants as accomplished teachers and PLC leaders.

WhatsApp groups were established in all the PLCs of the project. These groups were initially and mainly meant to be used for administrative and spontaneous messaging, with minimal intervention by the project supervisors. For professional correspondence, email, PLC moodle, and other forms of communication were used as well. The current research follows only the WhatsApp group of the L-PLC.

During the COVID-19 pandemic (which started in March 2020), PLC activities continued, and face-to-face meetings were replaced by online meetings using the ZOOM platform. Following extensive discussions, we — as the supervisors of the PLCs project — decided to stick to its original plan as much as possible by introducing or modifying suitable technological and techno-pedagogical applications.

Research Design

Rationale

The onset of the COVID-19 pandemic accelerated changes in the focus of many world interactions from face-to-face to online communication. Therefore, we sought for new ways and tools to follow the professional development and the dynamics in our L-PLC. Throughout the pandemic, we noticed an unusual volume of activity — “traffic” — in the L-PLC WhatsApp group. Even though such change is expected, as the educational system was forced to move into remote teaching and learning, we noted professional discourse initiated by the teachers themselves, and not by the supervisors. This new level of interaction was more prevalent than before and seemed to present new opportunities for professional sharing and learning. The spontaneous-unguided online communication directed us to investigate participants' interactions, knowledge development, and needs.

Research Questions

Guided by the rationale above, this study sought to answer the following research questions (RQs):

  1. 1.

    What can be learned about community processes and professional knowledge development of L-PLC by analyzing their WhatsApp discourse?

  2. 2.

    How were L-PLC online dynamics, as reflected in their WhatsApp discourse, influenced by the COVID-19 pandemic?

Methods

The longitudinal development of the L-PLC, and the impact of COVID-19 on it, were analyzed by characterizing the structure, the content, and the participants of the WhatsApp discourse across four consecutive semesters (cohorts). Aside from the intense usage of WhatApp during COVID-19 as a medium for ongoing communication between the L-PLC participants, the decision to trace this discourse as the main data source was also derived from its neutrality, with minimal presence and no interference by the supervisors. Consequently, it presented as a noncompulsive observation tool that reflects the L-PLC participants' needs, activities, state of mind, behaviors, and professional development, which in turn enabled tracing of processes and outcomes. The following paragraphs describe the analysis methods employed and how the results were visualized.

Data Collection

Data for this study were extracted from the log of the L-PLC WhatsApp group between March 2019 and March 2021. The following parameters were collected: the number of L-PLC participants, the number of messages sent by each participant, the message content, and the message timestamp. The dataset contained a total of 6599 messages. In addition, L-PLC participants' names, gender, age, higher education degree, and years of participation in the L-PLC were collected from background questionnaires.

Study Duration and Cohorts

The study covered four academic semesters (2 years), between March 2019 and March 2021, which included two semesters prior to the COVID-19 outbreak, and two semesters during the pandemic. Consequently, the data were divided into the following four cohorts (6 months each).

  • Cohort I: 3–8/2019 (the first half-year before COVID-19 outbreak)

  • Cohort II: 9/2019–2/2020 (the second half-year before COVID-19 outbreak)

  • Cohort III: 3–8/2020 (the first half-year during COVID-19)

  • Cohort IV: 9/2020–2/2021 (the second half-year during COVID-19)

As a result, we were able to compare relevant variables in each periodic cohort in order to identify and characterize various patterns, specifically, those emerging before and during the COVID-19 outbreak in our country. The cohorts were incorporated in all the analyses and visualizations.

Participants

The L-PLC was comprised of 64 college-educated participants, including 7 supervisors, 2 project secretaries, and 55 leading STEM teachers (2 males and 53 females). The actual number of participating teachers per semester varied between 43 and 48, and their ages ranged between 35 and 62 years. The majority of teachers (n = 31) participated in the L-PLC throughout the entire study period, while some left (n = 14) and others joined (n = 19) in the middle. The teachers came from a variety of regional schools, thereby representing different cultures, religions, socioeconomic and personal backgrounds.

Content Analysis

Content analysis was applied by classifying the WhatsApp messages into three main categories defined in a coding scheme that we synthesized for assessing PLC discourse (see Table 3). The coding scheme was designed based on professional knowledge frameworks (Angeli & Valanides, 2009; Koehler & Mishra, 2009; Shulman, 1986, 2015) and characteristics (Benaya et al., 2013; DuFour, 2004), and aimed to capture expressions that indicated: (1) Community Development (e.g., supportive expressions as evidence of a positive social climate, sharing knowledge and practice, norms, level of collaboration, cooperation, and partnership), (2) Knowledge Development, including CK, PCK, and TK. We decided to use these three categories after shared coding of part of the data and finding that PK without content knowledge simply does not exist and by acknowledging that TPCK is part of PCK, and (3) Administrative Issues. The scheme enabled us to identify and annotate relevant and important utterances that function as indications of teachers' learning or PLC development.

Table 3 Coding scheme for assessing PLC discourse
Full size table

Table 3 provides a short description and an example for each annotation category and sub-category. The Media category differs between Professional Media which can be inspected and coded, and Other Media which refers to media that were only mentioned in the log and were unaccessible (omitted).

Annotation Procedure

The dataset was annotated by two experts in the fields of STEM and teacher professional development. The unit of analysis was a single WhatsApp message, to which several tags could be assigned (multi-labeled annotation). Also, discursive sequences in the dataset were manually delimited according to the beginning and the end of consecutive related messages. As a result, the flow of messages was transformed into different threads grouped under the same discussions.

Annotations were considered valid only if both annotators agreed on a tag, following a two-fold process. First, after individual training to practice message content classification in accordance with the coding scheme, the annotators jointly annotated 7% of the dataset (456 messages); working together on the same threads, while consolidating and reaching full agreement. Then, both annotators independently tagged the rest of the dataset, while conducting consolidation sessions after annotating a certain amount of the same threads. These sessions aimed to minimize the variance in their annotation, thereby increasing its quality by reaching a higher level of an overall agreement. The final Inter-Rater Reliability (IRR) reached an average agreement of > 95%. Table 4 details the IRR for each annotated message tag (label). In this regard, it is important to note that Media tags were annotated automatically by using the existing remarks in the WhatsApp log — such as “image omitted”, “https://…” etc. — and thus were excluded from Table 4.

Table 4 Inter-rater reliability per message tag
Full size table

Heat Map and Bar Graph Visualizations

The annotations were visualized by tailored heat maps and bar graphs. Both forms of visualization enable quick and effective comparison of content frequencies, which in turn, can be translated into discourse trends and patterns. The classic heat map was customized by adding a layer of a vertical bar graph, to illustrate the number of messages sent by each participant, irrespective of the discursive function.

In addition, a Chi-square test was applied to examine differences in the annotated sub-categories between the cohorts. This analysis was applied in two steps: (1) an overall Chi-square test comparing the four cohorts to obtain an overall perspective of development over time; and (2) specific comparisons between cohort clusterings. A cohort was either compared to a sum of a cluster of other cohorts or to a specific cohort, to assess the significance of the change. The alpha (type 1 error) for each comparison was set to 0.01—0.05 divided by the number of comparisons (i.e., 0.05/5 = 0.01).

Social Network Analysis (SNA)

SNA is one of the core methods harnessed for identifying role and power structures within a given network. This method measures and visualizes the relationships between different participants, such as interaction type, frequency, various centrality degrees, and unique meaningful structures (e.g., star graphs, cliques) (Liu et al., 2022; Tabassum et al., 2018). Hence, when integrated with corresponding content analysis, SNA can shed light on patterns relating to PLC productivity and development (Baker-Doyle & Yoon, 2020; Matranga & Silverman, 2020; Polizzi et al., 2019). For these reasons, SNA and content analysis — along with appropriate visualizations — were jointly adopted to investigate the L-PLC discourse.

The structure of the L-PLC WhatsApp discourse was modeled, analyzed, and visualized through SNA, using the open-source software Gephi (Gephi Consortium, 2022). Since the WhatsApp messages appear one after the other — usually directed to all the participants — with no graphical hierarchy or clear indication regarding the direction of the communication, an undirected network graph is utilized to represent the discourse structure.

Figure 2 demonstrates a typical SNA graph of the L-PLC WhatsApp discourse. It is important to emphasize that the edges do not represent sent messages, but rather, interactions. Namely, when a participant took part in a discursive sequence, only one interaction (edge) was counted between him/her and all other participants who also took part in the same discursive sequence. The edge thickness reflects the exchange of interactions between two participants – the thicker the edge, the greater the interaction, also denoted by a corresponding color scale – from black (low) to red (high). Therefore, the more sequences both participants took part in, the thicker and redder the edge drawn between them. Discursive sequences initiated by a participant that did not receive responses are presented as self-loops.

Fig. 2
figure 2

Demonstration of a typical SNA graph of the WhatsApp discourse

Full size image

Network visualizations were further enhanced by adding two unique graphical layers of information to the graphs: (1) the number of messages sent by a participant, which is marked above each node (representing a unique participant), and (2) the years of participation in the L-PLC, which is denoted by a small colored square. A unique symbol instead of the colored square marks supervisors or secretaries (star or triangle, respectively).

Network Metrics

By employing and integrating various network metrics (as detailed below) with the content analysis annotations, we were able to quantify and measure the overall involvement of each participant in the L-PLC discourse, as well as structural changes, coherence degree, and formation of subgroups (i.e., network clusters).

Weighted Degree Centrality denotes the centrality of a participant by considering both his/her connection to other participants and the frequency of these interactions. Namely, this measure is based on the number of edges for a node but also considers the weight of each edge. In the visualization, the size of the node denotes the centrality degree of a participant. The Average Degree Centrality and the Average Weighted Degree Centrality of the entire network (i.e. taking into account all the participants) were computed as well (Ayyappan et al., 2016).

Modularity measures the strength of division of a network into groups, and thus is utilized as a community detection algorithm that uncovers the number of groups or clusters within a network. A relatively high modularity (coefficient) means that there are dense connections between the nodes within groups, but sparse connections between nodes in different groups. Nodes with identical colors denote participants that belong to the same group (Oliveira & Gama, 2012).

Graph Density quantifies how many edges exist between nodes in the network, compared to the number of theoretically possible edges between nodes; namely, it measures how interconnected the participants are. As a supplementary metric, the Clustering Coefficient measures the average probability that two neighbors of a node are themselves neighbors; in other words, the degree to which participants tend to cluster together, to form a clique. Thus, the more participants "know" each other, the higher the average clustering coefficient will be (Feng & Law, 2021; Hansen et al., 2010).

Results and Discussion

The results and discussion section addresses the two research questions together, referring to both dimensions: community processes and professional knowledge development (RQ1), before and during COVID-19 pandemic (RQ2).

The content analysis results of the WhatsApp messages are detailed in Table 5. The number of annotated tags in almost all the categories increased during the pandemic period (cohorts III and IV), with a notable peak in cohort III, corresponding to the first half-year after the COVID-19 pandemic broke out. While the number of annotated tags dropped in cohort IV, it was still higher as compared to cohort I, and in some categories also compared to cohort II. These findings imply that WhatsApp communication played an important role for teachers during the lockdown, and addressed their need for help when shifting to remote teaching. Interestingly, the overall number of annotated Media tags dropped in cohort III and cohort IV, as these messages usually included pictures generated during face-to-face meetings, which were not held during the pandemic. A moderate decline was observed in several other annotated tags (e.g., PF, PCK) between cohort III and cohort IV, which may reflect adjustment of the L-PLC’s online dynamics to the new reality resulting from the pandemic.

Table 5 Content analysis of L-PLC WhatsApp discourse
Full size table

The original goal of the WhatsApp group — i.e., to support communication regarding administrative issues — was still evident across all four cohorts. In parallel, the L-PLC participants utilized WhatsApp as a tool for professional development – both as part of building and enhancing their community and as a channel for obtaining and sharing professional knowledge, as evidenced by the number of tags relating to professional knowledge, and the numbers of documents and links attached, which were mostly professional in nature, and increased as the pandemic progressed.

The following sub-sections detail results regarding two main areas of professional development: (1) Community Development and (2) Knowledge Development. In this study, Community Development was inferred from integration between the SNA results, presented in Figure 3 and Table 6, and the content analysis of the WhatsApp messages, presented in Figs. 4, 5, and Table 5. Specifically, Knowledge Development outcomes are presented and visualized in Figs. 5 and 6.

Fig. 3
figure 3

SNA graphs of the L-PLC WhatsApp discourse in four consecutive cohorts

Full size image
Table 6  SNA Metrics of the L-PLC
Full size table
Fig. 4
figure 4

Bar graph of the frequency of Community Development sub-categories in WhatsApp messages, and their statistical significance across cohort clusters

Full size image
Fig. 5
figure 5

Bar graph of the frequency of Knowledge Development sub-categories in WhatsApp messages, and their statistical significance across cohort clusters

Full size image
Fig. 6
figure 6

Tailored heat maps that indicate the frequency of annotated sub-categories of WhatApp messages of individual L-PLC participants in each cohort

Full size image

The findings were also interpreted using three levels of granularity: macro, meso, and micro. In our case, a macro-level analysis can identify behaviors and patterns of the whole L-PLC, and how they correlate with the PLC characteristics. The meso-level refers to subgroups within the entire community, that share attributes or demonstrate a unique engagement. The micro-level refers to behaviors and professional development of individuals.

Community Development

In this section, the results refer to SNA metrics and visualization that were explicated in the Methods section.

Macro-Level (The Whole Community)

Figure 3 shows the SNA graphs of the L-PLC discourse across four consecutive cohorts; their corresponding network metrics are presented in Table 6. Of note, each network graph should be interpreted as a stand-alone unit, as it was adapted to visualize the information of a specific cohort. Therefore, the sizes of nodes are valid only for a specific cohort and cannot be compared across cohorts. Further support for this point is provided in the supplementary material.

The SNA graphs in Figure 3 show a greater level of interconnectivity regarding the L-PLC networks of cohorts III and IV compared to those of cohorts I and II, as indicated by both the greater number of edges (interactions between participants) and their thickness (frequency). This difference is supported by an increase in both the computed average weighted degree centrality and graph density (see Table 6). Specifically, cohorts I and II — which were active in the period before the COVID-19 outbreak — generally reflected an experienced L-PLC, that already developed a certain level of community characteristics. Cohort III showed notable differences in the network metrics, as well as in the number and frequency of the interactions, with more connections between different participants, denoting an increased level of communication. The pandemic period — cohorts III and IV — showed reinforced L-PLC cohesiveness (graph density of 0.859 and 0.729, respectively), compared to the pre-pandemic cohorts I and II (from a graph density of 0.620 in cohort I to 0.859 in cohort III, and from a graph density of 0.635 in cohort II to 0.729 in cohort IV). In cohort IV, the cohesiveness level was still relatively high, but less compared to cohort III, suggesting the beginning of an adjustment of the L-PLC online dynamics to the new reality resulting from the pandemic. The average weighted degree centrality were also higher in cohorts III and IV (170.7 and 100.6 respectively) than in cohorts I and II (63.5 and 62.1 respectively). This finding indicates a growing number of interactions between the participants in cohorts III and IV, and demonstrates that the L-PLC network became more interconnected.

A significantly higher number of Sharing (SHAR) tags — which refers to shares in teaching and learning materials, instructional apps, and information of scientific contents — was noted in cohorts III and IV, as compared to cohorts I and II. This finding can be attributed to the increased need for remote teaching materials during the COVID-19 pandemic. After the outbreak, the high volume of Sharing was relatively steady, with no significant change between cohort III and cohort IV; indicating that Sharing became a norm in the L-PLC.

A relatively low number of Cooperation (COOP) tags was obtained. This can be explained by the fact that complex and full cooperation mostly occurs in small groups or between two individuals, which requires other means of communication (phone calls, synchronous and face-to-face meetings, or shared documents) and not via WhatsApp only. In addition, the results suggest that group facilitators should elaborate more on the cooperation competency.

We perceive Positive Feedback (PF) as an important norm indicative of a safe environment with mutual respect among PLC participants. In the current analysis, the annotated PF tags were similar in cohorts I, II, and IV, with a statistically significant increase in cohort III. This peak may reflect reinforcement in the forms of sympathy and socioemotional support during a time of crisis.

Building Community (BC) values were significantly higher in cohort II. This may be a result of special time-consuming activities that were conducted during that period, explicitly to enhance collaborative community projects in the L-PLC program.

Meso-Level (Particular Subgroups)

Analysis of each of the four SNA graphs revealed a central cluster of senior participants (defined by the number of years they participated in the L-PLC), who frequently interacted with each other. For example, participants B, Q, R, and K who had been in the L-PLC for more than three years, formed a sort of a core of leaders within the L-PLC. The members of this senior group usually had higher values of weighted degree centrality compared to the participants of the other subgroups. They contributed more to the L-PLC discourse by asking for professional knowledge-related help, providing knowledge to the community, and sharing professional materials. On the other hand, the discourse of the newcomers' subgroup was comprised mainly of messages of positive feedback, indicating that the newcomers were attuned to the L-PLC discourse, and appreciated its value from their comfort zone as reflective observers. The SNA graphs did not identify any disconnected subgroups. In conclusion, the L-PLC proved to be a solid and cohesive community, which became even more unified in response to crisis.

Knowledge Development

The existing repertoire of teacher knowledge assessment tools includes design tasks, observations, teachers' pedagogical discussions, and self-reported tools such as surveys and questionnaires (Brantley-Dias & Ertmer, 2013; Tucker & Quintero-Ares, 2021). Here we suggest another methodology to this set, which is based on teacher communication with colleagues as reflected in their WhatsApp discourse analysis. Professional knowledge was reflected by utterances relating to CK, PCK, and TK, and in questions indicated by ASK-CK/PCK/TK. Figure 5 presents the frequency and significance level of the annotated Knowledge Development sub-categories. In addition, the individual contributions of the L-PLC participants in each cohort to the online WhatsApp discourse were visualized through tailored heat maps (see Fig. 6).

The CK and PCK tags generally increased between the cohorts, indicating positive and enhanced knowledge development processes over time within the L-PLC. Of note, the number of CK and PCK tags was significantly higher in cohorts III and IV compared to pre-COVID cohorts, suggesting that knowledge exchange was accelerated by the COVID-19 outbreak.

Furthermore, there was a significant increase in the number of both TK (from 9 tags in cohort I to 206 tags in cohort III), and ASK-TK (from 6 tags in cohort I to 82 tags in cohort III) tags across cohorts. In particular, the L-PLC participants actively requested and shared technological tools and methods in cohort III. These findings reflect the critical need for technological applications and solutions when shifting to distance teaching and learning. The findings are in line with recent studies which found that extensive pedagogical support is needed when designing digital teaching during the COVID-19 pandemic (Sarı & Keser, 2021). Rap et al. (2020) found that chemistry teachers participating in PLCs' WhatsApp groups during the COVID-19 pandemic mainly requested operational information related to TK and TPCK.

Micro-Level (Individuals)

The personal contributions of each L-PLC participant, in each cohort, are reflected by the content analysis visualized in the heat maps (see Fig. 6). The heat maps indicate high frequency of annotated Knowledge Development in cohort III, as visualized by the growing number and deeper blue squares.

A gradual personal development process of several L-PLC participants (e.g., I, K, and Y) was observed, as indicated by a significant increase in their Knowledge Development tags' values. In addition, the visualization enabled identification of roles taken by different participants. Some participants (e.g., D, I, K, and L) were found to continuously contribute to the knowledge development of the L-PLC, by frequently asking for different kinds of professional knowledge, and by sharing such knowledge with others. Such participants can be considered as knowledge promoters.

We suggest that the method described in this section is a valid and robust means of assessing PLC teachers' knowledge.

Conclusion

Development of Community and Professional Knowledge

In all the cohorts, the structure of the L-PLC demonstrated four network assemblies of connected interactions, with increasing cohesiveness over time. While the L-PLC WhatsApp was initially established as a platform for delivering administrative and formal messages, the participants gradually began to use it for professional purposes as well. The content analysis of the WhatsApp messages indicated that all types of professional knowledge, i.e., CK, PCK, and TK, increased over time, aligning with previously reported qualitative analyses regarding the same L-PLC (Scherz et al., 2020). Interpretation of the data using the Collaboration Model (Eylon et al., 2020), showed that over the two years of the study, the L-PLC developed higher levels of collaboration, which shifted from sharing towards partnership.

These findings echo the concept of “commognition” — a combination of both cognitive and interpersonal communication, coined by Sfard (2020), which suggests that knowledge is being developed through commognitive processes. While Sfard usually refers to face-to-face discourse as a mean of knowledge development, it can also be applied to online discourse, such as in WhatsApp groups. The current research suggests a strong link between professional knowledge development, and the level of community development. These two dimensions are intertwined, supporting, nourishing, and complementing each other.

L-PLC Development During COVID-19 Pandemic

Professional Interactions

During the pandemic period, the WhatsApp group instantaneously became a readily available resource for professional knowledge, in addition to the regular zoom meetings. The outbreak of the pandemic enhanced professional interactions, as evidenced by the increased volume of science and science teaching consulting issues, responses to queries, and sharing of teaching and learning materials, professional-technological applications, and online experiments.

Socioemotional Support

The WhatsApp platform provided an opportunity for community bonding, and served as a source of socioemotional resilience for the L-PLC participants struggling with a complex reality. Thus, without deliberate planning, the scope of support, sharing, and collaboration within the community expanded. As the L-PLC supervisors, we were concerned that the L-PLC may disperse due to COVID-19 lockdowns and the shift to distance teaching and learning. Yet, it appears that the decision we made at the very beginning of the pandemic — namely, to continue with the principal outline and pedagogical approach of L-PLC meetings through technological interfaces — encouraged online interconnections in the community.

Integrating Structure and Content

The findings of this study underscored the importance of integrating both structure and content analyses of the L-PLC's WhatsApp discourse, as also demonstrated in another study with similar goals (Alwafi, 2021). This methodology provided insights into processes and profiles of professional development at the macro, meso, and micro levels. The SNA measured degrees of coherence and types of interactions within the L-PLC, and subsequently illuminated the intensity of various community characteristics. The content analysis exposed the professional knowledge and know-how associated with the L-PLC dynamics. This approach enabled us to identify and monitor changes in key PLC characteristics, and to evaluate their fluctuations under the influence of chaotic changes, like the COVID-19 pandemic.

Implications and Limitations

This study was conducted before and during the first year of the COVID-19 pandemic. While future research will be needed to assess the longevity of shifts and developments detailed in this study, PLCs facing times of lockdown or other challenging situations may benefit from the insights of this study.

This study demonstrated the value of using WhatsApp, by researchers and community leaders, as a source of data for a noninvasive continuous diagnosis of group behavior, relationships, and online dynamics. Thus, it will enable responsive feedback to the community as a whole and to individual participants, as well as fine-tuning of PLC meeting programs.

The study was carried out within an established community, with a previous acquaintance and common professional knowledge, and thus with an appreciation of the value of a PLC. Studies investigating PLC dynamics in communities that were newly established during the pandemic, may yield different results. Therefore, research comparing PLCs with community tradition versus PLCs without is required. In this notion, relevant models of online professional development should be employed (e.g., Duncan‐Howell, 2010; Salmon, 2013).

Our decision to set the unit of analysis as a single WhatsApp message likely affected the results. There is no doubt that in messaging applications, such as WhatsApp, individuals may use multiple messages to talk around the same point. Therefore, while two or multiple messages are utilized by the same user to address the same issue, tagging their content and counting them as stand-alone messages may have biased the analysis. Hence, future studies may choose to set different units of analysis to overcome this potential limitation. However, it is important to note that a message-level analysis is common and acceptable when annotating the content of forums and social media sites (Alwafi, 2021; Cansoy, 2017).

Crossing Boundaries: Computer Science and Professional Development

The growing usage of social media platforms — such as WhatsApp — in online learning and teaching environments, calls for collaborative work between science teaching researchers and data and computer scientists. This will advance the development of appropriate and supportive artificial intelligence-based tools, particularly, by harnessing state-of-the-art Machine Learning and Natural Language Processing techniques. In the absence of such dedicated tools to apply on platforms where rich and big data are generated, it will be difficult for researchers and supervisors to track the online dynamics of a PLC to evaluate their productivity or relevance to the community objectives. Similar tools are used in learning analytics, to understand and optimize learning (Clow, 2013; Ferguson, 2012). Taken together Professional Development Analytics may serve as an authentic means for monitoring community development, and gaining insights into its social and professional knowledge expressions. Indeed, the analytical approach used in this study to identify, understand, and formalize both community and professional development processes within the L-PLC, can be applied to other learning communities.