Predicting the Persuasiveness of Influence Strategies from Student Online Learning Behaviour Using Machine Learning Methods

There is a dearth of knowledge on how persuasiveness of influence strategies affects students' behaviours when using online educational systems. Persuasiveness is a term used in describing a system's capability to motivate desired behaviour. Most existing approaches for assessing the persuasiveness of a system are based on subjective measures (questionnaires) which are static and do not allow for automatic measurement of systems persuasiveness at run-time. Being able to automatically predict a system's persuasiveness at run-time is essential for dynamic and continuous adaptation of the system to reflect each individual user's state. In this study, we investigate the links between persuasiveness of influence strategies and students' behaviour in an online educational system for a course. We implemented and tested Machine Learning (ML) classification models to determine whether persuasiveness had a significant impact on students' usage of a learning system. Our findings revealed that students learning data can be applied to predict the persuasiveness of different influence strategies. The implications are that by using machine learning classifiers powered with learning sessions data, online educational systems would be able to automatically adapt their persuasive strategies to improve students' engagement and learning.