Research has shown that student participation in classroom dialogues is associated 
with learning gains, and initiatives to encourage more dialogic forms of learning and teaching 
are abundant. Yet, less is known about how different students may experience, participate in, and what they may gain from dialogic classroom activities. In the current work, we explore potential differences in participation of high and low-achieving students (from 6 different 
classrooms) in upper elementary Hebrew lessons of teachers who participated in a professional development program on academically productive dialogue. We used Epistemic Network Analysis to identify differences across twelve lessons. Findings reveal that the network model of low-achieving students is characterized by simpler talk moves, reduced connectivity, and 
repetitive loops. In contrast, high achieving students’ network model is more interconnected, and the strongest connections formed among codes there are indicative of a reasoned argumentation and critical stance. Analyses of selected excerpts further explored the dynamics that may have led to these different patterns.  

There is increasing scholarly agreement about the key features of academically productive classroom dialogue, as well as about its role in student learning and growth. Yet, despite this emerging conceptual consensus, the ways in which it is measured and coded in quantitative research efforts vary significantly across settings, research teams, and studies. In order to communicate, compare and integrate findings from this rich body of empirical research and to further theory refinement, a more uniform approach to measuring classroom dialogue is needed. In this paper, we argue that for such an approach to succeed it should meet the following requirements: It should enable stable coding across settings and raters (reliability), be capable of capturing a wide variety of dialogue features (comprehensiveness) and enable different types of research questions and analyses (flexibility). Given the current conceptual maturity, as well as the vast number of quantitative studies that have accumulated in the last two decades, we believe that the field has sufficiently matured to achieve this goal. We selected seven well-known and validated coding frameworks for academically productive classroom dialogue. Through an iterative process of comparison, deconstruction, and application to classroom dialogue transcripts, we identified a set of nine elementary particles, so-called dialogue elements (DEs), that are common across the different coding categories and can be reliably coded at the conversational turn level. We then demonstrate how a much larger set of "compound" dialogue constructs can be identified post-coding by flagging co-occurrences of different DEs that recreate higher-order dialogue constructs. With the help of this Dialogue Elements to Compound Constructs Approach (DECCA), the majority of coding categories from each of the seven selected coding frameworks could be recreated. DECCA thus enables interrater reliability, while simultaneously maintaining the flexibility and comprehensiveness needed to enable research on a large variety of questions with a single methodological approach. The implications, limitations and for future research and theory are discussed.

Within the scholarly field of academically productive classroom dialogue, several elusive, yet rich and potent discourse moves have received special attention. However, their rarity and complexity also poses significant challenges to meeting interrater reliability thresholds, and they are often omitted from quantitative research efforts. We propose a different approach for coding that circumvents these issues, called DECCA. In this presentation, we showcase this methodology by focusing on teacher revoice. We demonstrate how it is possible to deconstruct this complex phenomenon into smaller and simpler elements. We then code each turn in the corpus for the existence or absence of each element (which we term Dialogue Elements, DEs). Adding a post-coding, pre-analysis stage allows us to extract turns which contain the specific combination of DEs relevant for revoice and distinguish it from similar teacher dialogue facilitation moves. Theoretical and practical implications are discussed.

During the height of the COVID-19 pandemic, teachers around the globe had been
forced to move their teaching to full-time online, remote teaching. In this study, we aimed
at understanding teacher burnout during COVID-19. We conducted a survey among
399 teachers at the peak of a prolonged physical school closure. Teachers reported
experiencing more burnout during (vs. before) the COVID-19 pandemic. Contributing
factors to this burnout were high family work conflict and low online teaching proficiency.
Burnout was associated with lower work-related wellbeing: Lower work commitment,
and higher turnover intentions. It was also associated with lower psychological
wellbeing: More depressive and anxiety symptoms, and lower subjective wellbeing.
Approach (but not avoid) coping strategies served as a protective factor against the
burnout-turnover intentions association. We conclude with recommendations on how
to mitigate teacher burnout, thereby contributing to teacher wellbeing.

Academically productive talk (APT) in classrooms has long been associated with significant gains in student learning and development. Yet, due to COVID-19 related restrictions, teachers around the world were forced to adapt their teaching to online, remote settings during the pandemic. In this investigation, we studied APT in junior high school during extended online, remote teaching spells. Specifically, we focused on the extent APT was a part of online teaching practices, what characterized teachers who tended to promote APT more in online, remote teaching, and associations between APT and teacher well-being, as well as student motivation and engagement. Findings from two survey studies (Study 1: 99 teachers, and 83 students; Study 2: 399 teachers) revealed the following patterns: Students and teachers agreed that APT was used to a lesser extent in remote, online classes, and associated with more interactive instructional formats (whole classroom discussion, peer group work, and questioning), but not with frontal teaching and individual task completion. Teachers with a higher sense of teaching self-efficacy, autonomous orientations, and higher empathy tended to promote APT in online, remote teaching more. More APT was associated with greater teachers’ work-related (i.e., lower burnout, more commitment to teaching, and lower turnover intentions) and psychological well-being (i.e., less depressive and anxiety symptoms, and higher subjective well-being). Finally, student experiences with APT in online, remote learning was positively associated with learning motivation and engagement. Theoretical and practical implications are discussed.

We explore how problem framing shapes teacher dialogue in teacher-led, school-based peer consultations. Twenty audio-recorded workgroup conversations were analyzed using a mixed-methods approach. Three different frames for presenting problems of practice were identified: teaching-, student- and classroom composition-oriented. Quantitative analyses showed associations between problem frames and the ensuing positioning of teachers as main agentive actors. In-depth qualitative analysis of two focal cases of low-teacher-agency problem frames (student- and classroom composition-oriented) revealed that psychologized discourses and attribution of responsibility to parents contributed to reduction of teacher responsibility and concomitant limited agency, and that initial problem frames were resistant to reframing.

Despite visions of social network technology (SNT) for collaborative knowledge construction, recent research in secondary schools suggest that students use these tools mainly for knowledge sharing of study-related artifacts. We extend these findings to higher education settings and report on two survey studies that map characteristics of students’ self-directed use of SNTs for study purposes, in undergraduate university programs (N = 264) and teacher training colleges (N = 449). The combined findings confirm that students use SNTs extensively for uploading, linking and downloading study-related artifacts in peer-directed SNT groups. They regard these practices positively and believe they improve academic achievements. Sharing was predicted by positive attitudes toward sharing and collectivist value orientations, motivated overall by prosocial reasons and less frequent in competitive study programs. Use of shared materials was associated with performance-avoidance achievement goals and lower GPA. Findings, directions for future research and implications are discussed in the context of learning theories, as well the knowledge sharing literature.

Scholarly efforts to identify core design features for effective professional development (PD) efforts have rapidly grown in the last two decades. Based on extensive literature reviews, meta analyses and large-scale quantitative studies, scholars have arrived at short lists of core design principles for effective PD programs. These design principles are presented as based on strong evidence from large-scale, replicated and rigorous research studies, and as at the heart of consensus among PD scholars. In the present essay, we appraise the quality of the evidence on which this claim is based. We identify several major flaws in the research base on which such claims are based and conclude that, overall, the evidence is weak and claims about strong evidence-based findings are misleading. Additional reservations about this research program are discussed.

This study contributes to growing scholarly interest in teacher-led, on-the-job learning communities and how collaborative inquiry into practice can be supported in such 
contexts. We particularly focus on the relative advantages and limitations of three teacher team activity types: video analysis, peer consultation and pedagogical planning. Fifty-four 
transcribed teacher meeting excerpts were analyzed using the CLIP coding scheme for teacher collaborative inquiry into practice, assessing aspects of inquiry-based reasoning, collaboration and focus on pedagogy (teacher actions, student thinking and disciplinary content). Quantitative comparisons showed that, overall, collaborative inquiry into practice was lowest during peer consultations, in part because teachers were often not positioned as agents of change in such conversations. Teachers tended to inquire into each other’s ideas more often during pedagogical planning. Surprisingly, teacher team video analysis activities were not characterized by higher measures of attention to student thinking, nor inquiry orientation. Practical and theoretical implications are discussed. 

Detection of suicide risk is a highly prioritized, yet complicated task. Five decades of research have produced predictions slightly better than chance (AUCs = 0.56 – 0.58). In this study, Artificial Neural Network (ANN) models were constructed to predict suicide risk from everyday language of social media users. The dataset included 83,292 postings authored by 1,002 authenticated Facebook users, alongside valid psychosocial information about the users. Using Deep Contextualized Word Embeddings for text representation, two models were constructed: A Single Task Model (STM), to predict suicide risk from Facebook postings directly (Facebook texts → suicide) and a Multi-Task Model (MTM), which included hierarchical, multilayered sets of theory-driven risk factors (Facebook texts → personality traits → psychosocial risks → psychiatric disorders → suicide). Compared with the STM predictions (.621 ≤ AUC ≤ .629), the MTM produced significantly improved prediction accuracy (.697 ≤ AUC ≤ .746), with substantially larger effect sizes (.729 ≤ d ≤ .936). Subsequent content analyses suggested that predictions did not rely on explicit suicide-related themes, but on a range of text features. The findings suggest that machine learning based analyses of everyday social media activity can improve suicide risk predictions and contribute to the development of practical detection tools.

Detection of suicide risk is a highly prioritized, yet complicated task. Five decades of research have produced predictions slightly better than chance (AUCs = 0.56 – 0.58). In this study, Artificial Neural Network (ANN) models were constructed to predict suicide risk from everyday language of social media users. The dataset included 83,292 postings authored by 1,002 authenticated Facebook users, alongside clinically valid psychosocial information about the users. Using Deep Contextualized Word Embeddings for text representation, two models were constructed: A Single Task Model (STM), to predict suicide risk from Facebook postings directly (Facebook texts → suicide) and a Multi-Task Model (MTM), which included hierarchical, multilayered sets of theory-driven risk factors (Facebook texts → personality traits → psychosocial risks → psychiatric disorders → suicide). Compared with the STM predictions (.606 621 ≤ AUC ≤ .608629), the MTM produced significantly improved prediction accuracy (.690 697 ≤ AUC ≤ .759746), with substantially larger effect sizes (.701 729 ≤ d ≤ .994936). Subsequent content analyses suggested that predictions did not rely on explicit suicide-related themes, but on a range of text features. The findings suggest that machine learning based analyses of everyday social media activity can improve suicide risk predictions and contribute to the development of practical detection tools.

Effective instruction for conceptual change should aim to reduce the interference of irrelevant knowledge structures, as well as to improve sense-making of counterintuitive scientific notions. Refutation texts are designed to support such processes, yet evidence for its effect on individual conceptual change of robust, complex misconceptions has not been equivocal. In the present work, we examine whether effects of refutation text reading on conceptual change in biological evolution can be augmented with subsequent peer argumentation activities. Hundred undergraduates read a refutation text followed by either peer argumentation on erroneous worked-out solutions or by standard, individual problem solving. Control group subjects read an expository text followed by individual problem solving. Results showed strong effects for the refutation text. Surprisingly, subsequent peer argumentation did not further improve learning gains after refutation text reading. Dialogue protocols analyses showed that gaining dyads were more likely to be symmetrical and to discuss core conceptual principles.

Data collection from online platforms, such as Mechanical Turk (MTurk), has become popular in clinical research. However, there are also concerns about the representativeness and the quality of this data for clinical studies. The present work explores these issues in the specific case of major depression. Analyses of two large data sets gathered from MTurk (N₁ = 2,692 and N₂ = 2,354) revealed two major findings: First, failing to screen for inattentive and fake respondents inflates the rates of major depression artificially and significantly (to 18.5% to 27.5%). Second, after cleaning the data sets, depression in MTurk is still 1.6 to 3.6 times higher than general population estimates. Approximately half of this difference can be attributed to differences in the composition of MTurk samples and the general population (i.e., socio-demographics, health and physical activity lifestyle). Several explanations for the other half are proposed and practical data-quality tools are provided.

Online Social Networking Sites (SNSs) are immensely popular, especially among adolescents. Activity on these sites leaves digital footprints, which may be used to study online behavioral correlates of adolescent psychological distress and to, ultimately, improve detection and intervention efforts. In the present work, we explore the digital footprints of adolescent depression, social rejection, and victimization of bullying on Facebook. Two consecutive studies were conducted among Israeli adolescents (N = 86 and N = 162). We collected a range of Facebook activity features, as well as self-report measurements of depression, social rejection, and victimization of bullying. Findings from Study 1 demonstrate that explicit distress references in Facebook postings (e.g., "Life sucks, I want to die") predict depression among adolescents, but that such explicit distress references are rare. In Study 2, we applied a bottom-up research methodology along with the previous top-down, theory driven approach. Study 2 demonstrates that less explicit features of Facebook behavior predict social rejection and victimization of bullying. These features include 'posts by others', 'check-ins', 'gothic and dark content', 'other people in pictures', and 'positive attitudes towards others'. The potential, promises and limitations of using digital Facebook footprints for the detection of adolescent psychological distress are discussed.

Argumentation has been recognized as an important classroom activity and as a potentially powerful means for learning complex academic content. However, eliciting and sustaining student-to-student argumentive discourse that is both critical as well as constructive is also known to be notoriously difficult. Whereas previous research has traditionally focused on the cognitive, meta-cognitive and task-related antecedents and conditions for productive student argumentation, in the present work we explore two social-motivational factors that may provide insight into this difficulty, namely students' individual achievement goals and gender. In two separate studies, undergraduate students indicated their intentions to engage in different discourse types when asked to discuss their solutions to a complex topic from astronomy (N = 245, Study 1) or economics (N = 98, Study 2) with a disagreeing peer. In addition to the productive, ideal type of argumentive discourse for learning purposes (i.e., deliberative argumentation), three additional discourse types were targeted that typically ensue, but are considered less productive (i.e., disputative argumentation, quick consensus seeking and private deliberation). The overall pattern of results show that mastery goals (a focus on developing competence and task mastery) are associated with deliberative argumentation and with private deliberation. In contrast, performance-approach goals (a focus on demonstrating competence relative to others) as well as high confidence are associated with disputative peer argumentation. Quick consensus seeking was predicted by higher performance-avoidance goals (a focus on avoiding incompetence relative to others) and lower mastery goals. No consistent gender differences were found. Taken together, the results extend previous work in socio-cognitive conflict settings and emphasize the role of achievement goals in peer argumentation.

We examine how teacher-student communication through social network technologies may support student resilience during an ongoing war (i.e., the 2014 Israel-Gaza war). Based on student responses from open-ended surveys (N = 68), five content categories of emotional support were identified: caring, reassuring, emotion sharing, belonging, and distracting. The mere existence of continuous online contact with teachers also contributed to resilience perceptions. Interviews with 11 secondary school teachers revealed three main purposes for this communication: (a) delivering emotional support to students, (b) monitoring their distress; and (c) maintaining civilized norms of discourse. Practical implications and theoretical contributions are discussed.

In the present study, we examined the effects of feedback that corrects and contrasts a

student's own erroneous solutions with the canonical, correct one (CEC&C feedback)

on learning in a conceptual change task. Sixty undergraduate students received

expository instruction about natural selection, which presented the canonical,

scientifically accepted account in detail. Two-third of these received CEC&C feedback on their self-generated solutions to open-ended test items. Students either received this feedback on their pretest solutions (prior to instruction), or on their immediate post-test solutions (following instruction). Students in the control condition only received the correct canonical answers to the immediate post-test items and compared these with their own solutions autonomously. Conceptual understanding on transfer items was assessed after one week. Results showed that students in the CEC&C feedback conditions outperformed control students. Timing of feedback did not affect learning, however. These findings add to accumulating evidence from different lines of research on the importance of instructional support that explicitly compares and contrasts between erroneous student models and canonical models in conceptual change tasks.

A small but powerful body of evidence shows that certain forms of classroom discussion can produce learning gains that go beyond the topics actually discussed. In a range of countries, students who engaged in dialogue showed better initial learning and retained their learning gains for longer periods when compared to untreated comparison groups. In some cases, students who were engaged in learning through dialogue even outperformed their untreated counterparts. In this chapter, we review the evidence and consider why dialogue might produce these effects, looking at both cognitive and motivational-social explanations. Despite evidence of the surprising and robust effects on student learning, it is rare to find dialogic teaching in the classroom. We propose explanations for the resistance to it, from individual teachers and from the system, and suggest that opening up opportunities for more students to learn through dialogue will require researchers and practitioners to work together in new ways.