Enhanced Learning Experiences Based on Regulatory Fit Theory Using Affective State Detection

Introduction

Intelligent Tutoring Systems (ITS) is used in the field of education for online learning as one of the major application of Artificial Intelligence. In the present era of e-learning, ITS mainly focuses on personalized learning with the help of Artificial Intelligence techniques. A cloud assisted learning models have started to assign a lot of importance to learners’ affective states in learning through the use of Intelligent Tutoring Systems (ITS). The role of ITS is to provide custom learning content to learners through learning models and an adaptation engine (Ramkumar Rajendran, Sridhar Iyer, & Sahana Murthy, 2019). The ITS uses log files in the cloud to create personalized learning models by recording their answers or responses, time taken to respond and the number of attempts to answer, read or respond in the same log files. The ITS also understands the learners’ cognitive states by accessing the information from the cloud server based on the learners’ past knowledge and background already existing in the learner models (Peter Brusilovsky & Eva Milln, 2007).

During this pandemic situation, online learning plays the most vital role in school and college education. When compared to online learning, offline learning have efficient strategies to monitor the students affective state and counsel them to enhance their learning outcomes. Hence, affective state detection based on their performance during online learning is considered to be important. Moreover, providing personalized feedback messages will further help to improve the learners' performance. The learning process is now well known, researched and proven to involve both cognitive and affective processes along with the fact that considering affective states improves learning outcomes (Sidney K. D’Mello et al., 2006). Affective states widely include frustration, monotony, boredom, eagerness and anxiety that are part of the learning model. The ITS detects learners’ frustration by way of self-reporting, recognition of emotions through their faces, mining log files from the server and use of physical as well as physiological sensors (Cristina Conati et al., 2009; Jacob Whitehill et al., 2014; Keith Brawner et al., 2012).

Using a theory-based approach, this research makes use of student interaction with ITS to create a personalized learning model with the help of MoodleCloud. Frustration is defined in theory and those definitions are used along with ITS log files collected from the server to determine learner frustration. This model also helps to infer the factors that contribute to the frustration of learners and also if, when and why frustration occurs in learners during the learning process. The motivational messages are thus delivered from the server as a form of feedback to the learners’ affective states while learning. Frustration is considered as one of the important parameters. It is also argued that sending a different kind of motivational messages from the server related to frustration component leads to increased learning success. Moreover, learning styles also have a great impact on the performance of the learner as discussed in (Jegatha Deborah L, Baskaran R & Kannan A, 2014). Based on the individual learners’ learning style, suitable e-learning contents can be delivered to them which helps in scoring high marks.

This research integrates a theory-based approach called the Learning Management System (LMS) brand-named MoodleCloud for real-time detection of learners’ frustration. Moodlecloud is a web-based freeware LMS primarily used for education in schools, universities and companies since it helps to create personalized learning environment. It stores the learners’ profile information and the entire data of the course in the server. As the information required can be retrieved through the Internet, Moodlecloud is used as a web application. Since the learners’ profile data and the learning contents are stored in cloud, security policies to access the data should be taken into consideration. There are some existing works (Christian Esposito et al., 2021; Ahmad Al-Qerem et al., 2020; Christos L. Stergiou et al., 2021; Mohammed Banu Ali, 2019; Kostas E. Psannis et.al., 2019) in which the authors discuss about the security policies and protocols to reduce the computational complexity in the cloud.