Assessing the Impact of Large Language Models on the Scalability and Efficiency of Automated Feedback Mechanisms in Massive Open Online Courses

Abstract

The rapid proliferation of Massive Open Online
Courses (MOOCs) offers particular difficulties in providing
timely and high-quality personalized feedbacks associated with
customer interactions at scale. This research examines the gap
which Large Language Models (LLMs) address with focus on
automation in providing timely feedback and the scalability
efficiencies of LLMs in the feedback scope provided in MOOC
settings. Adopting a results-oriented experimental approach to
feedback systems, LLMs like GPT-3.5 and GPT-4 are
implemented across varying course contexts and learning
groups. Their outputs are benchmarked against traditional
systems through semantic similarity calculations, response time
measurement, cost evaluation, and learner satisfaction metrics.
LLMs’ ability to comply with instructor feedback while
improving responsiveness and personalization outpaced
traditional methods in every context analyzed, with satisfaction
scores outperforming pre-set benchmarks across the board.
Learners reported appreciation towards AI responses, citing
enhanced understanding and interaction, overshadowed by
defendable claims of bias, genericity, and flawed constituent
pressure. All in all, the study provides concrete guidance
illustrating the ways in which LLMs reconfigure pedagogical
feedback mechanisms alongside MOOCs, shaping subsequent
shifts in the design and integration strategies utilized in elearning
frameworks across the world.