Technological Forecasting & Social Change 176 (2022) 121461 Available online 31 December 2021 0040-1625/© 2021 Elsevier Inc. All rights reserved. Dynamic group formation in an online social network Reshawn Ramjattan a, * , Nicholas Hosein b , Patrick Hosein a , Andre Knoesen b a The University of the West Indies, St. Augustine, Trinidad b University of California, Davis, CA 95616, USA A R T I C L E INFO Keywords: Online groups Recruitment Matchmaking Network fow ABSTRACT For those seeking to recruit teammates for a specifc purpose, like a project or study group, challenges quickly arise once they have exhausted their social circle. In the wake of the current pandemic, meeting new people that are right for a specifc team is even more diffcult than before due to the lack of in-person events. On social media platforms, users often have large networks of connections but have very few close personal relationships within them. This makes it diffcult to fnd compatible people that share the same goal, and are interested in niche groups on those platforms. We present a scalable framework for establishing small online groups that balance two objectives, making the best group recommendations to users and guiding group hosts to the best users for their group. We illustrate this framework using three use cases. Lastly, we evaluate a serverless implementation using a large social media dataset to simulate a production environment and compare our framework to a network fow approach to solving the problem. 1. Introduction The ongoing pandemic has forced many events and gatherings to be cancelled or moved online. However, many social aspects of in-person gatherings, such as networking, cannot be easily facilitated by digital media. Attending relevant events, classes or conferences is a frequent approach to building a network of compatible and like-minded people. So without real chances to network, assembling teams with a common purpose becomes a challenging social task. For example, fnding suitable co-founders for start-up ventures or potential employees based on a casual conversation at mixers or academic events, or establishing rapport with classmates during practical exercises to form study groups. These are situations where physical gatherings allow for personal con- nections as a by-product of its main purpose and can greatly help with meeting new people if you need a group. While online platforms can create an acceptable stage for the main event, they do not provide the same level of interpersonal possibility. We consider the problem of a user seeking to form a small purposeful group of the most suitable and compatible people. On existing social networks, users often have a large number of friends but very few of them are close personal relationships. The large network of loose and estranged re- lationships is not very useful for fnding group members that share the same interest, purpose and timing. Research done by Ramjattan et al. (2020) shows prior work in this area as well. Consider a student forming a study group. Studies by Dolmans and Schmidt (2006) and Springer et al. (1999) show that there are several cognitive and motivational benefts to small group learning. There are also benefts to presentation, communication and team-building skills. Study groups allow students to take responsibility and beneft from small group learning outside the classroom. Without preexisting re- lationships, forming a study group in a class is close to forming a random assortment. This randomness results in a variety of personality types, learning styles and interests among the members. This variety can negatively affect not only the group’s compatibility, and therefore comfort in engaging discussion, but also the effectiveness of the small group learning since its members learn best in different ways. Even if the students do not know their best learning method and style, work by Chamorro-Premuzic et al. (2007) shows the correlation between per- sonality and preference for learning methods. By creating an online means to form these groups we can allow students to fnd a group with high compatibility both as people and as learners. It allows us to consider a wider set of potential group members and some key variables such as learning style, personality traits McCrae and John (1992), subject comfort and general common interests. We can then form the ideal online study group effciently. For example, a * Corresponding author. E-mail addresses: reshawn.ramjattan@my.uwi.edu (R. Ramjattan), nhosein@ucdavis.edu (N. Hosein), patrick.hosein@sta.uwi.edu (P. Hosein), aknoesen@ ucdavis.edu (A. Knoesen). Contents lists available at ScienceDirect Technological Forecasting & Social Change journal homepage: www.elsevier.com/locate/techfore https://doi.org/10.1016/j.techfore.2021.121461 Received 10 February 2021; Received in revised form 24 December 2021; Accepted 27 December 2021