Can Generative AI Reduce the Dropout Rates of Online Learners for Failures in Information System Services?

Instantaneous and on-the-spot online learning is limited by the obstacle of failure in information system services (such as none-response or mismatched Q&A), resulting in consistently high dropout rates. Generative AI, trained using large language models (LLMs), can recognize the content of user-generated (UGC) and offer personalized services and anthropomorphic interactions. However, the effect of applying generative AI to failures in information system services remains unknown. Building upon this, this study aims to investigate the impact of integrating generative AI into online learning platforms on the dropout rates of online learners. Comparative experiments were conducted to examine the relationship between generative AI and the dropout rate of online learning (Study1), as well as their mechanisms (Study2). Additionally, education performance could be influenced by course level and learners’ ability. Furthermore, the moderation of course difficulty (Study3) and learner education degree were considered. The results of the study shed light on whether generative AI can be used to mitigate the negative effects of information service failures, while also laying the basis for understanding the influence of generative AI on user trust.