Tie formation is known to depend on spatial propinquity, or physical distance between individuals in organizations. Spatial propinquity has been studied extensively at many scales and shown to have a strong effect on tie formation, but measured effects are typically related to an individual spatial scale of analysis. Organizations are often interested in spatial reorganization of individuals to affect the connections between their employees. However, organizations are often able to change the spatial location of individual contributors at multiple scales. For organizations to achieve effective reorganization of contributors, the relative importance of multiple spatial scales must be considered. In this work, spatial propinquity effects are measured at a desk and office level with a novel higher education institution (HEI) dataset to determine the relative importance of office layout and distance between offices on organizational social structure. In this dataset, changes in social structure are observed before and after directed spatial reorganization at the desk and office level. Through analysis of the changing sociometric network measures, we develop insights that HEI leaders can use to encourage collaboration through targeted spatial reorganization.

This study investigates the dynamic evolution of formal and informal leadership in cross-functional teams by leveraging machine learning to build predictive models. Utilizing longitudinal data from 2,845 employees at a major Chinese high-tech firm (2014–2017), the research examines how leadership roles shift over time in response to organizational and relational factors. The authors construct annual collaboration networks and categorize leadership changes into promotion, demotion, or stability across three time periods. Five machine learning algorithms—logistic regression, random forest, gradient boosting, support vector machines, and k-nearest neighbors—were tested for predictive performance, with the random forest model emerging as the most accurate (Accuracy = 0.839, F1-score = 0.865).

To interpret the models, SHAP (Shapley Additive Explanations) analysis was employed, revealing that the most influential predictor of leadership change was an individual's relationship with formal leaders. Strong relationships with formal leaders increased promotion likelihood, while high intra-departmental mobility was associated with demotion. The study highlights that relationships with formal leaders encompass trust, authority, and mutual obligations, which shape performance expectations and influence the likelihood of leadership role changes. These social dynamics, often overlooked in traditional leadership theories, prove essential in predicting leadership trajectories.

By integrating predictive modeling with explainability, this research advances understanding of leadership dynamics beyond static hierarchies, emphasizing the interplay between formal authority and informal influence. The findings offer theoretical insights into leadership evolution in complex, cross-functional environments and practical implications for leadership development and talent management in rapidly changing organizations. This approach highlights the potential of machine learning to inform and optimize leadership strategies in contemporary teams.

Recently, platforms like OpenReview have emerged as widely used open peer-review systems in computing fields. Using OpenReview data from the International Conference on Learning Representations (ICLR) in 2018 and 2019, this study examines the prevalence and structure of academic collaboration within the Machine Learning (ML) community. The goal is to understand how collaboration networks among researchers evolve, and whether diversity relates to paper acceptance or rejection outcomes.

We constructed collaboration networks where nodes represent authors and edges represent co-authorship of submitted papers. Our analysis addresses three questions: (1) What is the prevalence and structure of collaboration in the ICLR community? (2) Does diversity in collaboration (international, interdisciplinary, inter-institutional) associate with paper acceptance? (3) How do collaboration characteristics differ between accepted and rejected submissions?

To answer these questions, we first assess the network structure by analyzing degree distributions and centrality measures, and by applying community detection to identify groups within the network. Next, we conduct temporal analyses to study changes in co-authorship and determine whether collaborations remain stable or fluctuate over time. We also examine the influence of key institutions on network formation. Finally, we compare accepted and rejected papers to investigate how factors such as international and interdisciplinary collaborations may relate to acceptance outcomes.

This study contributes to understanding how collaboration diversity may influence academic success, highlighting potential barriers or biases that impact researchers, particularly those from underrepresented backgrounds or institutions. The findings aim to encourage broader, more inclusive networks that share knowledge and enhance innovation in ML.

Introduction. The CHILD-BRIGHT Network is a national, patient-oriented research network dedicated to child health, involving researchers, clinicians, decision-makers, and partners with lived/living experience. Five service programs and a central office support twelve funded research project teams in integrating implementation science, equity, diversity, inclusion, decolonization and Indigenization, partner engagement, capacity building and knowledge mobilization (KM) into their work. The purpose of this work is to establish a baseline quantification of the network’s social structure with respect to KM, mapping connections among members and their ties to external organizations. Methods. We conducted a cross-sectional, mixed-methods social network analysis (SNA) case study. We collected demographic, relational and attribute data through remote meetings and online surveys. Network properties and visualizations were generated using UCINet and NetDraw softwares. Results. With 97 participants, representing a 45% response rate, findings indicated a core-periphery structure with an average path length of 2.2. Eight key actors reached over 97% of the network, highlighting the importance of highly connected individuals in facilitating KM. Participants reported over 400 external institutional or organizational ties nationally and internationally, but these were not leveraged to support KM or other network mandates/goals. This finding suggests that individuals embedded in multiple groups within the network and connected externally can act as knowledge brokers, enhancing knowledge exchange, uptake, and collaboration. Conclusion. Leveraging strategic connections through key actors and external ties can enhance the CHILD-BRIGHT network’s impact for children and their families nationwide. Future evaluations will assess network evolution and the effectiveness of national KM strategies.