In the age of the internet, literature is consumed in unprecedented ways. Modern social movements often call upon those of the past through key quotes and references to influential literary works. Quotes can go viral, seen outside of their context by thousands of people and become associated with these movements or rediscovered by new communities. For instance, key figures in post-WWII literature such as author and civil rights activist James Baldwin have had their words re-immortalized within the context of contemporary movements such as Black Lives Matter. Baldwin’s era of literature was one of marked social change and evolution within the literary world that parallels our society today, making it significant to understand how quotes from this period can reappear and spread across social media. To analyze the reception of post-war literature on Twitter, we utilized a dataset of over 40 million tweets quoting or referencing James Baldwin, as well as similar datasets quoting four other influential authors of the time including David Foster Wallace and Kurt Vonnegut. We focused on the patterns of text reuse (i.e., the repetition of known quotes) in tweets from 2006-2023, examining key moments of reception and exploring the context of virality for key quotes. During this context-finding process, we also developed a novel method for conducting self-identified user demographic analysis. We implemented clustering algorithms on both tweets and user bios, supplemented the resulting clusters with manual merging processes, and experimented with various visualization strategies. Our results yielded clear quote usage patterns for certain demographic groups, demonstrating the efficacy of the novel demographic extraction method. These methods can be expanded for further demographic-focused social media research and help us understand how cultural movements evolve today.

This study utilizes design research to explore how storytelling informs the design, usage, and knowledge production of a digital archive repository housing digitized memory objects. Ranging from ao dai to math booklets, these memory objects are grounded by narratives of Vietnamese diasporic identity and experiences shared by community researchers as part of Sarah Nguyen’s Sharing Stories, Sharing Trust (SSST) workshop series. To understand how story-driven approaches translate and transform digital archive design, I draw upon multiple methodologies such as case study analysis of existing community-based applications of digital archives and thematic analysis of SSST workshop discussions (formatted as observational memos). I also draw from user interviews with community researchers using a semi-structured, narrative-driven protocol. These analyses inform the design of a digital repository prototype that foregrounds story-driven design whilst exploring possibilities for the preservation and sharing of Vietnamese diasporic experiences.

This research project focuses on developing and enhancing an AI auditing system to assess diversity and fairness in large language modeling (LLMs) systems. By replicating an existing Python-based audit framework, originally created by my Principal Investigator (PI), this study extends its functionality to specifically evaluate how race and ethnicity are represented in AI-generated outputs related to professional occupations. The enhanced auditing system cross-references race and ethnicity data with job positions to identify potential biases, providing a deeper understanding of whether AI systems (specifically GPT-4) disproportionately associate certain ethnic groups with specific professions. These findings contribute to the ongoing discourse on fairness in AI, offering insights into how LLM models may perpetuate or mitigate biases in career representation. This research is critical for the development of more equitable AI systems that reflect diversity across various social and professional contexts, highlighting the importance of fairness in the deployment and usage of AI technology.

This research examines the statistical interactions of genetic risk scores and behavior data from wearable devices, including physical activity and sleep measures, to predict Major Depressive Disorder (MDD) symptom onset. MDD is a widespread mental health issue, with nearly all indicators of mental health worsening from 2013 to 2023 and 30% or more current children experiencing mental health symptoms. Research shows that lifestyle changes, such as improving physical activity and sleep behavior, can alleviate early-stage MDD symptoms. But, many people are unaware of their genetic vulnerability to MDD, leaving them unprepared for potential challenges. This study uses the Adolescent Brain Cognitive Development (ABCD) dataset, the largest U.S. longitudinal study of brain development and child health. ABCD provides extensive psychometric, demographic, genetic, and wearable data for research. This study uses genetic and wearable tracking data to predict MDD severity and support early interventions. It also investigates how genetic risk levels inform how physical activity and sleep patterns must change to mitigate MDD symptom severity.  This study will calculate polygenic risk scores (PRS) for ABCD subjects and improve prediction accuracy for non-European populations using state-of-the-art bioinformatics tools. Then, this study will utilize mixed effects modeling to analyze additive and interactive effects of PRS, wearable data, and depression severity scores. Lastly, this study will program machine learning (ML) models to provide variable importance and accuracy results. The goal is to create a personalized, data-driven approach to MDD prevention and empower individuals to take proactive steps toward mental well-being based on a comprehensive view of their genetic and behavioral factors.