🔍 Research Projects

Designing LLM-Agents with Personalities: A Psychometrics Approach | 2023.10 – Present

• Presented at American Psychological Association 2024 Convention.
• Explored 5 different personality tests using OpenAI’s large, contextualized embeddings, suggesting new paradigm to analyze psychometric tests.
• Investigated personality assignment strategies using real world data and parametric simulations, allowing controllable, fine-grained and replicable personality assignments to LLM-Agents.
• Collected data from 350 participants and conducted direct comparison for humans and Agents’ response.
• Validated the predictability of Agents’ decision using their personality and comparing that with human participants, showing promising potential to use LLM-Agents as human surrogates in behavioral research.
• Conducted systematic evaluations via Confirmatory Factor Analysis (CFA) and convergent correlations.

Can SEM Fit Indices Distinguish Between CFA and EFA Data Structures | 2022.05 - 2023.08

• Developed R Shiny app to simulate experimental data and produce interactive visualizations; Conducted extensive literature search, review, and replication on CFA and EFA models.

• The manuscript has been submitted and currently under review.

Improving Big Five Inventory-2: Alternative to Likert Scale | 2021.08 – 2023.08

• Analyzed survey responses using Structural Equation Modeling (SEM);
• Compared psychometric properties of four scale formats;
• Recruited and communicated with participants;
• Handled survey distribution; Managed compensation process.
• Used Qualtrics and Human Subject Pool platform to oversee the recruitment and data collection progress.
• Presented at American Psychological Association 2023 Convention.

Socioeconomics and Children’s Imagined Future: A Study using Machine Learning | 2020.09 – 2021.09

• Analyzed 10,000+ essays and socioeconomic data using Machine Learning and Natural Language Processing (i.e., Structural Topic Model).
• Managed / supervised a group of 20 research assistants; Developed protocols and offered technical support.
• Poster Presentation at the Association for Psychological Science 2020 Annual Convention.

Gauging Student Engagement with an Explainable Al interface via Eye-tracking | 2021.11 – 2022.09

• Evaluated personalized explainable AI tutoring system; Analyzed participants’ eye-tracking data and cognitive and personality data using a linear mixed model, demonstrating AI tutoring systems are more effective when tailored to cognitive ability.
• The project the effectiveness of explanation AI in the context of education.
• Presented at IJCAI 2022 Workshop of Explainable Artificial Intelligence