Machine learning - mediated immediacy for synchronous communication in online classrooms

Abstract

Instructor immediacy includes behavioural practices adopted by the instructor in an online classroom that can imbibe a sense of co-presence with students despite being physically separated, and in the process promotes effective communication between individuals, as well as students’ understandings and experiences. Much of the research analysing instructor immediacy in online learning focuses on asynchronous communication (e.g. emails, discussion forums), with less attention paid to its synchronous counterpart (when communicating parties interact in real time simultaneously, e.g. chats).

Question answering (QA), an automated process of answering questions asked by humans, is based on machine learning and natural language processing. As a QA system can ‘learn’, it can thus be ‘trained’ to produce responses that preserves lecturer immediacy, instead of the cold, impersonal responses produced by machine. With an immediacy-equipped QA system taking over the role of an instructor during synchronous communications in an online learning environment, learners will have access to an ‘instructor’ any time they wish.

This translational project aims to explore instructor immediacy as mediated by a trained QA system (or QA-immediacy) in synchronous communications in an online learning environment. Specifically, we strive to determine how QA-immediacy is related to three criterion variables in an online learning environment: affective learning, cognitive learning, and motivation. Furthermore, this project seeks to discover whether online learners are able to differentiate between a real-life instructor and an immediacy-trained QA system. QA-immediacy will be embedded within an SIT-developed online bridging course for Chemistry that incorporates personalised feedback and adaptive online assessment – Chem Quest.

We will use a mixed method research study with multiphase convergent parallel design. The project will be implemented in four phases: (a) Pre-project phase where control groups are established, (b) Phase I in which the QA system is trained, (c) Phase II where instructor-immediacy is applied, and (d) Phase III for QAimmediacy implementation. Analysis of Variance (ANOVA) will be used to test RQ 1 (instructor-immediacy versus QA-immediacy) and RQ 2 (learning outcome of control versus experimental groups). Simple regression analysis will be conducted to determine the degree to which the predictor variable (instructor immediacy) might explain variance in student affective learning (RQ 3), cognitive learning (RQ 4), and student motivation (RQ 5).

QA-immediacy in online learning has not been attempted thus far, to the best of the team’s knowledge. The findings of this project will shape the future of both online learning and machine learning, by shedding light on whether QA-immediacy has the potential to replace a real-life instructor. A well-trained QA-immediacy system is expected to translate the advantages instructor immediacy brings to online classes, while freeing the instructor from synchronously communicating with online learners around-the-clock. At the institutional level, the integration of technology (QA-immediacy in this case) into teaching and learning is in line with SIT’s Smart Campus efforts. Beyond just education, QA-immediacy, if shown to be capable of extending the human touch to otherwise cold and transactional automated services (e.g. chat bots in banking), will also be a step forward in Singapore’s Smart Nation initiatives.