Abstract:
This paper explores the implementation of large language models (LLMs) for
automatic answer script evaluation, examining their potential to revolutionize
educational assessment. LLMs offer significant advantages, including increased
efficiency, consistency, and scalability in grading processes, as well as cost
savings for educational institutions. However, their deployment also presents
challenges, such as high energy consumption, potential biases, privacy concerns,
and the socio-economic impact on employment within the educational sector. This
study addresses these issues by proposing a comprehensive sustainability plan that
includes optimizing energy efficiency, utilizing renewable energy sources,
enhancing data protection measures, and ensuring transparency and accountability
in automated evaluations. Additionally, strategies for balancing automation with
human oversight and upskilling educators are discussed. The findings highlight
the need for ongoing research to refine LLM applications, ensuring their ethical,
sustainable, and effective integration into educational systems. The implications
for further study emphasize the importance of continuous improvement in energy
use, bias mitigation, data privacy, and the broader impact on educational practices
and employment
