Contributed Poster Presentations: Justice Equity Diversity and Inclusion Outreach Group

AI systems rely on complex big data algorithms that emerge from training processes rather than explicit human programming. This inherent complexity often makes it difficult, even for experts, to fully understand how these systems generate their outputs. This "black box" can lead to misplaced trust on AI, biased decisions, and unjust social outcomes, raising serious ethical and practical concerns. To address these challenges, this paper explores key strategies for enhancing data transparency, algorithmic transparency, explainability, and interpretability. Explainable AI (XAI) techniques, such as feature importance analysis and counterfactual explanations, can help make AI decision-making more transparent. Additionally, hybrid models that combine black-box AI with interpretable components offer a balance between performance and accountability. However, no technical solution is sufficient on its own. Human oversight remains the most critical safeguard, ensuring that a responsible party is always accountable for AI-driven decisions. This is especially crucial in high-stakes domains such as healthcare, finance, and law enforcement, where AI's impact on human lives is profound. 

In the United States, although the gaps in health insurance coverage by sexual orientation have been closing since the implementation of the Affordable Care Act and legalization of same-sex marriage, the LGBTQ+ group continues to report healthcare disparities such as more delayed or unmet care due to cost. There appear to be structural forces at play. This study answers the need to build upon the literature in understanding social contexts attributed to healthcare disparities of LGBTQ+ people using multilevel modeling. In addition, sexual orientation may intersect with gender to impact healthcare disparities. Further, the aforementioned situations would be more problematic for people who have chronic diseases because healthcare is essential to them. Therefore, it is critical to understand healthcare disparities related to sexual orientation and gender for people who have chronic diseases. From the methodological perspective, quantitative intersectional approaches have drawn increasing attention. In this study, I utilized quantitative intersectional methods for logistic regression such as prevalence risk ratios and measures of additive-scale interaction for easier and clearer interpretations of intersectional results and more actionable policy implications for inter-categorical groups using the National Health Interview Survey data. This research aims to answer the following research questions: (1) How is the intersection of sexual orientation and gender related to healthcare disparities? (2) Do social contexts have effects on healthcare disparities? (3) How much of the total variability in an outcome of healthcare disparities is attributable to the social contexts? 

Data are a powerful tool to inform policy and social change, but often personal perspectives can influence interpretation of data. This is especially relevant when data concerns systems, such as prisons, that are understudied, historically not transparent, and mute the voices of those inside. We sought to evaluate the impact of incorporating narrative and storytelling in data reporting on prisons. It is the goal of this project to produce a data report that delivers a more nuanced story that is supported by data analysis and limits misinterpretation due to biases and misunderstanding of underrepresented populations. Using survey data collected in a Vermont prison, two reports will be constructed: one by academic researchers and one co-created with incarcerated individuals and prison staff. Both will be shared with policymakers and focus groups to assess the impacts of narrative on understanding. Phase 1 results will detail the co-creation process and present the final data reports. Phase 2, the focus group process, will be conducted in Fall 2025. We hypothesize that incorporating narrative storytelling into data reporting will offer a better understanding of prison context.