What impacts matriculation decisions? A double-blind experiment via an AI-led chatbot trained with social media data

Abstract

This thesis explores students’ matriculation decision factors via an AI-led chatbot trained with social media data. The novelty of this thesis resides in the following methodological approaches: Firstly, it employs data mining and text analytics techniques to explore the use of topic modelling and a systematic literature reviewing technique called algorithmic document sequencing to identify decision factors from social media to be integrated to the internal model of the AI through a methodological pluralist approach. Secondly, it introduces a chatbot design and strategy for an AI-led chat survey generating both unstructured qualitative and structured quantitative primary data. Finally, upon interviewing 1193 participants around the world, a double-blind true experiment was run seamlessly without human intervention by the AI testing hypotheses and determining the factors that impact students' university choices. The thesis showcases how AI can efficiently interview participants and collect their input, offering robust evidence through an RCT (Gold standard) to establish causal relationships between interventions and their outcomes. One significant contribution of the thesis lies in aiding higher education institutions in understanding the global factors influencing students' university choices and the role of electronic word-of-mouth on social media platforms. More importantly, the research enhances knowledge in identifying themes from social media and literature, facilitating the training of AI-augmented chatbots with these themes, and designing such chatbots to run large scale social RCTs. These developments may enable researchers from a wide range of fields to collect qualitative and quantitative data from large samples, run double-blind true experiments with the AI and produce statistically reproducible, reliable, and generalisable results.