Reinforcement Learning

As many of us are near the end of our academic terms, this week I would like to share an article that connects some of the recent generative artificial intelligence (AI) literature with foundational educational theories in higher education. The article is entitled, “Reinforcement Learning (RL) in Education: A Literature Review” by Bisni, et al. (Sep 2023).

[Extremely Brief Background: In the area of psychology, reinforcement indicates that the consequence of an action increases or decreases the likelihood of that action in the future. RL draws inspiration from various psychological theories, including operant conditioning. Conditioning is strengthening of behavior which results from reinforcement (Skinner, 1965). Some researchers are making connections between historical educational RL and Q-Learning (Watkins, 1989). Q-Learning is a RL policy that will find the next best action, given a current state. It chooses this action at random and aims to maximize the reward.]

The research pursues three research questions (RQ):

Does RL actually help in the education field?
If so, what are the applications, and where might we anticipate it being most useful?
What are the considerations, challenges, and future directions of RL in education?

The authors state that the use of RL within the field of education can bring about a significant shift in the way students engage with learning and how teachers [assess, measure and] evaluate progress. The use of RL in education allows for personalized and adaptive learning, where the difficulty level can be adjusted based on a student’s performance. As a result, this could result in heightened levels of motivation and engagement among students. The aim of this article is to investigate the applications and techniques of RL in education and determine its potential impact on enhancing educational outcomes. It compares the various policies induced by RL with baselines and identifies four distinct RL techniques: the Markov decision process, partially observable Markov decision process, deep RL network, and Markov chain, as well as their application in education.

The article shares the potential RL Techniques for higher ed application:

Teacher–Student Framework
Adaptive Experimentation in Educational Platforms
Instructional Sequencing in Education
Modeling Students
Generating Educational Content
Personalized Education through E-Learning
Personalizing a Curriculum

The authors do point out potential RL challenges in higher ed:

Insufficient pertinent learning resources for personalized or adaptive learning
Lack of perspectives on education in AI research
Data selection for AI predictive models
Socio-emotional factors are understudied in AI research
Teachers lack sufficient expertise in AI technologies
Potential ethical and social issues

References

Bisni, F., Wasfi, A., Hayajneh, M., Slim, A., & Abu Ali, N. (2023). Reinforcement Learning in education: A literature review, Informatics, 10 3), 74. https://doi.org/10.3390/informatics10030074