Writing/Publications
Exploring fairness, policy, and human-centered design in technology. My work investigates how algorithms impact, and can reshape, social inequalities.
Completed Research
AI in Legal Systems:
Examining Gender Bias and
the Role of UK Legal Frameworks in Addressing It
The 2nd International Conference
on Global Politics and Socio-Humanities
(ICGPSH 2024)
Abstract:
This study investigates gender bias in AI systems used within the legal sector, focusing on risk assessment, facial recognition, and decision-support tools. It examines how historical data, representational gaps, and homogeneous development teams perpetuate inequality, and evaluates how the UK GDPR and proposed DPDI Bill address these issues. Finding current laws insufficient, the paper proposes a proactive legal framework assigning clear responsibilities to policymakers, companies, and users to prevent bias and ensure equitable AI governance.
Judicial AI
Policy Analysis
UK GDPR
AI Governance
Gender Inequality
View presentation
View paper
Algorithmic Bias in the Gig Economy: Policy Pathways for Gender Equity
Social Justice Watcher NGO &
Independent Research Project (EPQ)
Abstract:
This paper examines how the gig economy deepens gender inequality by reinforcing occupational segregation and wage gaps. It identifies two drivers: persistent discrimination and unequal care responsibilities carried over from traditional workplaces, and new algorithmic biases embedded in gig-platform systems. These factors collectively undermine women’s income, safety, and wellbeing. The study proposes solutions that target both structural and algorithmic roots of inequality—strengthening anti-discrimination laws, extending worker protections, integrating women’s experiences into platform design, and expanding social and health safeguards for gig workers. This paper was later adapted and awarded First Prize (top 2%) at the Social Justice Awards, Spring 2025.
Gender Equity
Gig Economy
AI Governance
Algorithmic Harm
View paper
View winning entry
Fairness & Validity in NLP Models: Evaluating LDA, Word2Vec, and BERTopic for Text Analysis
Department of Education, University of California, Santa Barbara Research Mentorship Program (RMP)
Abstract:
Natural language processing (NLP) offers an efficient, systematic, and scalable alternative to manual qualitative analysis. While NLP models are known to introduce algorithmic bias, there is limited understanding of how different model architectures contribute to biased outputs. This study compares the fairness and validity of three NLP models—Latent Dirichlet Allocation (LDA), Word2Vec, and BERTopic—applied to uncovering topics. Fairness is assessed using a framework of representation, aggregation, and learning bias; validity is evaluated through human interpretation, coherence scores, and clustering metrics. We found that all models exhibit sensitivity to representation bias. LDA and Word2Vec are more susceptible to aggregation and learning bias, often emphasizing dominant terms due to reliance on statistical word co-occurrence. In contrast, BERTopic leverages contextual embeddings to generate coherent topics and better capture rare but meaningful terms. While BERTopic performs best in fairness and validity, LDA and Word2Vec remain effective for exploring dominant relationships in large datasets.
Natural Language Processing
Algorithmic Bias
Machine Learning
Computational Social Science
View paper
View presentation
View poster
China’s Digital Economy:
Examining Its Impact on Inequality through the Lens of
Employment, Income Disparity, Innovation, and Competition
The 2024 International Conference on Accounting, Finance,
and Business Administration (ICFBA 2024)
Abriged Abstract:
This paper analyzes how China’s digital economy affects inequality across employment, income, innovation, and competition. While it widens gaps between high- and low-skilled workers and fosters market concentration through network effects, it also enhances job creation and innovation efficiency. The study highlights the need for policies that balance equality and growth by improving labor quality, reducing regional disparities, and reinforcing antitrust regulations.
Digital Economy
Inequality
Policy Analysis
Policy Recommendation
View paper