In the grand scheme of things, my interests are interdisciplinary at the intersection of analytics, novel methods, open science and education. My focus is to push the boundaries of methodological advances and unlock the power of big data to bring new insights. In so doing, I explore the temporality, individual differences and heterogeneity as an overarching analytical framework for personalized, equitable and debiasing science. As such, I worked with machine learning, mixture models, network science, sequence analysis, process mining, Markov models, psychological networks, temporal networks, meta-analyses, topic modelling and scientometrics. The main themes of my research can be summarized in the following list with some links to examples:
One of my main areas of research centers around networks and network analysis. I published about networks in collaborative learning, in problem-based learning, in language learning and in computer science education. I also worked on finding best methods for network analysis, performed a large scale empirical investigation of centrality measures and did a meta-analysis on centralities to find out which measures capture students’ achievement. Other threads of network research include intervention based on network analysis, modeling diffusion and social contagion of knowledge, modeling conflict and robustness in networks and group influence on collaborative dynamics and success factors.I also contributed to guidelines on best practices in network analysis, finding future agendas, and addressing the challenges that face network analysis in education.
Temporal networks are a new paradigm in general and have been recently introduced to education. Currently, temporal networks are I am passionate about the different possibilities that temporal networks could bring to modeling the relational time dynamics. I published about the basic principles of temporal networks in education. How temporal networks can early predict students’ performance and how network analysis can help reveal the dynamics of collaborative learning across different learning platforms. My latest work explores the affordances of temporal networks in collaborative learning, the temporal network centralities and the pathways of knowledge construction.
Modeling the complexity of human behavior has always been challenging. The recent introduction of psychological networks has made it possible to model the complexity of human behavior. Recently, I started working with psychological networks to understand how self-regulated process works in language learning, how self-regulated unfolds in collaborative learning and to model the complex interplay between SRL monitoring events as well as to map idiographic processes. Idiographic analytics seeks to derive insights from individual persons and deliver recommendations based on their own data, and therefore, one can have a single person study. Having idiographic analytics would have several privacy and ethical constraints resolved.
Longitudinal research is rare, exhaustive and requires advanced methods. However, there is much to learn from how behavior unfolds in time. I have tackled several of these challenges using very large datasets that span a whole program. I explored the longitudinal trajectories of engagement to show how engagement unfolds over four years and the consequences of early disengagement. I also studied how collaborative roles emerge, persist, transition or evolve over time in a massive study over four years. I also investigated the evolution of learning strategies and the transfer of successful strategies across courses over a full program.
Learning analytics and quantitative research methods in general are on top of my mind and work. I worked on improving predictive learning analytics, applied single-paper meta-analysis to study which predictors capture students’ performance reliably and consistently. I also studied the effect of individual differences on accuracy of predictive models. I also used sequence mining in several studies, multi-channel sequence mining and Hidden Markov Models, Mixed-Hidden Markov models. I also used several person-centered methods including latent class analysis, latent profile analysis and clustering. Relevant to the temporal methods are the process mining techniques which includes frequency based process mining and First-order Markov Models. I have also used Graphical Gaussian models and epistemic network analysis in mapping students strategies.
My love for networks, visualization and scholarly practices have amounted to a new passion for scientometric and bibliometric studies. I mapped the landscape of research on e.g., computational thinking, games, education technology, Koli calling conference and multiple sclerosis. Such work has opened the doors for a whole new book coming soon on computer science education research. Altmetrics is closely related to bibliometrics research, where one can see how the public has embraced, discussed or debated a paper. In this paper, I look at the the entire field of computer science education research across social media