Skip to main content
Go to the home page of the European Commission (opens in new window)
English English
CORDIS - EU research results
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary

The Time is Now: Understanding Social Network Dynamics Using Relational Event Histories

Periodic Reporting for period 4 - TIMEISNOW (The Time is Now: Understanding Social Network Dynamics Using Relational Event Histories)

Reporting period: 2022-08-01 to 2024-01-31

Relational event history data are becoming increasingly available due to new technical developments. These data contain detailed information about who interacted with whom in a network and when. For example, employees wear sociometric badges storing time-stamped interactions between colleagues, classrooms are monitored to observe interactions between teachers and students, and police databases store violent interactions between criminal gangs in city districts. This new type of data has the potential to greatly contribute to our understanding of dynamic social networks by providing new insights about speed, rhythm, duration, and lag in social interactions. However a crucial problem is that statistical tools for analyzing such data are currently underdeveloped. We are therefore unable to exploit this treasure of information, resulting in a limited understanding about the evolution of social relations in continuous time.

To address these shortcomings, this project (i) developed an innovative and advanced statistical framework for relational event history analysis, (ii) implemented the new framework in free and user-friendly software packages (R and JASP) to ensure general utilization among social scientists, (iii) applied the methodology to result in a better understanding of social interaction dynamics in temporal social networks. The software will open many new doors to groundbreaking research to study social interaction dynamics in temporal social processes.
The contributions of this project can be categorized according to the three main objectives of the project:
I. An advanced relational event modeling (REM) network framework has been developed to study complex social interaction behavior using relational event history data.
II. The new framework has been applied to better understand social interaction dynamics in organizational science, education, and criminology, and other fields of science.
III. Free, open-source, and user-friendly statistical software packages (R and JASP) have been developed that allow researchers to apply the framework in their own research.

I. The REM framework encompasses the following novel methodologies:
* Exploratory statistical methods for time-sensitive social processes
...to study how social interaction behavior changes using moving window techniques (Mulder & Leenders, 2019);
...to study memory decay functions using relational event sequences (Arena et al., 2022);
...to identify key drivers of social interaction processes using Bayesian shrinkage priors (Karimova et al., 2023);
...to assess model (mis)fit of REMs using posterior predictive checks (Mulder & Hoff, 2024);
...for events having different types, sentiments, and duration (Meijerink et al., 2023; Meijerink et al., 2024; Arena et al., 2024a).
* Confirmatory statistical methods for testing theories on social network interactions using Bayes factors
...for testing which drivers dominate social interaction processes (Heck et al., 2023, Mulder & Leenders, 2019);
...for testing memory decay parameters in REM (Arena et al., 2023);
...for testing the heterogeneity of social interaction styles across social networks (Viera et al., 2024);
...for making predictions about social interactions in the (near) future (Arena et al., 2022, Lakdawala et al., 2024);
...for testing the factors that affect the formation of ties (Mulder et al., 2024).
* Advanced statistical methods
...to study abrupt changes of social interaction behavior (Shafiee-Kamalabad et al., 2023).
...to study complex social processes caused by latent social structures (Mulder & Hoff, 2024);
...to analyze multilevel relational event models (Vieira et al., 2024a).
* Statistical algorithms have been developed to ensure general utilization of the methodologies including
...Markov chain Monte Carlo algorithms (Mulder & Hoff, 2024) and Hamiltonian Monte Carlo algorithms for learning of Bayesian posteriors of REM parameters (Arena et al., 2024c);
...Meta-analytic approximation techniques for “Big” relational event sequences and sequential fitting based on streams of relational events that are observed in real time (Vieira et al., 2024b).

II. We developed free and user-friendly software that allow researchers to apply the novel framework in an accessible manner:
* The following R packages have been developed (which also include tutorials)
- ‘remify’ for processing, arranging and transforming relational event data (Arena et al., 2024c).
- ‘remstats’ for computes statistics using relational event history data (Meijerink et al., 2023).
- ‘remstimate’ for optimization algorithms for tie-oriented and actor-oriented relational event models (Arena et al., 2024c).
- ‘remx’ for fast meta-analytic approximations for relational event models (Viera et al., 2024).
- ‘bremory’ for modeling memory retention in relational event data (Arena et al., 2024d).
- ‘remulate’ for generating dynamic temporal network data based on relational event models (Lakdawala et al., 2024).
* A REM module (Pfadt & Mulder, 2023) has been developed for the popular software program JASP with an intuitive graphical user-interface (such as SPSS). The module will be part of the next large update of JASP.

III. The application of the novel framework has resulted in a better understanding of social interaction dynamics in various fields of research:
* In organization studies (WP1), we obtained new insights about integration processes of new workers (Mulder & Leenders, 2019), unobserved social structures (Mulder & Hoff, 2024), and key drivers of face-to-face interactions (Karimova et al., 2023).
* In sociological applications (WP2), we learn how social interaction dynamics change in classrooms (Vieira et al., 2024) and among freshmen (Meijerink et al., 2023).
* In criminology (WP3), we learned how complex relations based on rivalries (e.g. enemy of a friend) and retaliations affect can predict when violent interactions will occur and who will be involved (Gravel et al., 2023).
* In communication science, we identified key drivers of communication patterns (Karimova et al., 2023) and how they can abruptly change when emergencies occur (Shafiee-Kamalabad et al., 2023).
* In political networks, we learned how the relations affects future relations among socio-political actors (Arena et al., 2023a).
The main objective of the project was to develop a generally applicable statistical framework for studying social interaction dynamics using relational event history data. The framework includes the following key methodologies:
- Exploratory statistical methods that can be used to better understand temporal features such as pacing, rhythm, and memory in social interaction processes;
- Confirmatory statistical testing procedures (such as Bayes factors) for testing theories on social relations and social structures;
- Advanced statistical methods to study complex interaction dynamics which may be driven by hidden social structures or driven by external factors influencing social relations which may change over time;
- Advanced and fast statistical algorithms to fit complex relational event models in a relatively short amount of time;
Moreover, the novel methodologies are implemented in free and open statistical software packages in R and JASP (with graphical user-interface) to allow social network researchers and practitioners to study social interaction processes in their own research.
The network between communication stations during NASA's famous (but disastrous) Apollo 13 mission.