AI-Enhanced Team Formation in Online Learning Environments
ABSTRACT
Online Education Surge: A Statistical Overview
Mohammed Almutairi
Ph.D. Student Department of Computer Science and Engineering
University of Notre Dame
The transition to onl ine education during the COVID-19 pandemic has significantly disru pted tradition al pedagog ical interaction,
especially in non-verbal cues such as facial expressions and body language, which are crucial aspects of collaborative
learning. Traditional indicators of team synerg y play a role in student engag ement and are often lost in virtual environments.
This l oss can imped e communicati on, hinder team rapport, and adversely affect student motivation and academi c outcomes.
To address these challenges, ou r project proposes an innovative AI framework to enhance team formati on and collaboration in
online learning. Central to this framework is the use of the Up per Confid ence Bound (UCB) algorithm, which optimizes team
composition by considering academic strengths and student preferences, thereby fostering more d ynamic an d effective
collaborative learning en vironments.
UCB’s role involves balancing two critical aspects of stud ent team formation: the p reference for familiarity and addressing
academic di versity among students to enhance th e learning experience. The UCB algorithm takes inpu t from the Non-
Dominan t Sorting Geneti c Algorithm (NSGA) and ach ieves its objectives by iteratively evaluating and updating strategies based
on student feedback. It calculates the u pper confid ence bound for each potential team composition, which includes a measure
of both the expected success of the team and the degree of uncertainty or exploration. This appro ach ensures the repetition of
past successful formations an d the exploration of new combinations, potentially uncovering more effective team dynami cs.
In 2026:
The global market share will reach $336.98 billion.
1
In 2027
The number of students is expected to be 1 billion worldwide.
1
Full-time student enrollment increased 100%.
1
Between 2019 to 2020
0.9
1.3
2.3
4.6
6.1
12
2012 2019 2020
Full-time student enrollment Millions
1
Graduate Undergradue
1
Diaz-Infante et al., 2022 McKinsey & Company.
will successfully reach higher levels, but only if
given enough attention, as face-to-face learning
is the best approach to providing these levels of
attention.
2
90% of Students
2
University of the Potomac, online-learning-vs-traditional-learning
Prof. Diego Gómez-Zará
Assistant Professor Department of Computer Science and Engineering
Concurrent at the Department of IT, Analytics, and Operations
University of Notre Dame
Enhanced Engagement and Collaborative Learning
during online courses, a significant
contrast to traditional classroom
settings. This disparity highlights
underlying challenges in virtual
learning environments, including
motivational issues that stem from
students' perspectives.
3
40% to 80% of Students Drop out
3
Papia Bawa et al., 2016. Retention in Online Courses: Exploring Issues and Solutions—A Literature Review
Preliminary Results
Heat-map demonstrates that the UCB algorithm consistently
recommends high-reward teams to the user. It’s clear from
this visualization that the algorithm has identified a distinct
pattern.
To contrast with the UCB algorithm, this plot illustrates the
teams recommended to the user on a random basis. It’s
evident that there’s no discernible pattern in these random
recommendations.
The chart reveals that the UCB algorithm outperforms
random selection in terms of cumulative rewards. We can
observe that in the beginning both of random and UCB
approach are exploring up to 15 teams selected, then we
clearly see that the UCB started creating pattern to choose
the most reward team.
