Lan Jiang's Dissertation Defense
PhD candidate Lan Jiang will present her dissertation defense, “Effective and fair language-based assessment with large language models.” Jiang's dissertation committee includes Associate Professor Nigel Bosch (Co-chair), Associate Professor Halil Kilicoglu (Co-chair), Associate Professor Vetle Torvik, and Professor Dong Wang.
Abstract
Timely and accurate text assessment is essential in many contexts, including student learning and scientific publishing. For example, instant evaluation of student responses can facilitate student learning, while assessing reporting quality in scientific articles supports peer review. However, with growing college enrollments and increasing numbers of manuscript submissions, maintaining high-quality assessment has become more challenging. Traditional human assessment is time-consuming, requires specialized expertise, and may suffer from inconsistencies due to fatigue or subjective judgment. Advances in machine learning and natural language processing, particularly large language models, offer new opportunities to automate and support language-based assessment.
This dissertation explores how to construct systematic evaluation frameworks and how large language models can be applied to assess human-produced text in both educational and biomedical contexts, while also examining important bias issues that may arise in automated assessment. Collectively, the studies in this dissertation demonstrate that LLMs can assess human-produced work, such as automated scoring of student responses and evaluation of scientific reporting quality and can contribute to fair assessment. Overall, this dissertation provides benchmark datasets, methodological advances, and practical insights for the automated assessment of text in both scientific articles and student responses.
Questions? Contact Lan Jiang.