12:30 PM – 1:15 PM | Merten Hall, Room 1204
Merten Hall, Room 1204
Nicotine-Induced Changes in Microglial ATP Responses: Role of Mitochondrial and ER Calcium Regulation
Aya Nusir (College of Science)
Nicotine exposure contributes to cardiovascular and neurodegenerative diseases through its modulatory effects on inflammatory processes. In the brain, microglia serve as primary immune cells whose functions, including cytokine release, phagocytosis, and synaptic maintenance, are critically regulated by ATP-dependent purinergic signaling. While nicotine acts through nicotinic acetylcholine receptors (nAChRs) to influence microglial activity, its effects on purinergic signaling remain unclear. Using human microglia HMC3 cells, we investigated how nicotine exposure (1-10 µM, 72 hours) alters calcium signaling dynamics. GCaMP imaging experiments reveal that nicotine-treated microglia exhibit significantly enhanced ATP-evoked cytosolic calcium responses. Targeted imaging of subcellular calcium dynamics using R-CEPIA1er and RCaMP mito demonstrate that nicotine treatment promotes increased ATP-meditated calcium release from ER stores and alters mitochondrial calcium handling, respectively. Proteomic analysis of nicotine treated cells identified significant changes in mitochondrial protein expression, including alterations in ATP synthases (ATP5MG and ATP5PO), superoxide dismutase 1 (SOD1) and ATP subunit inhibitory factor 1 (ATPIF1). These results suggest a mechanism by which nicotine modifies microglial ATP signaling through regulation of organelle calcium dynamics. These findings provide insight into mechanisms that may contribute to neuroinflammatory signaling during nicotine addiction and smoking-related diseases.
Everyday Perceptions of the Carceral System in the U.S.: Power, Hegemony, and the Media
Jae-Lynn Tabverez Brown (College of Humanities and Social Sciences)
This study explored how various types of media—including social media, television, film, and news—shape perceptions of the criminal legal system and connect to the hegemonic culture of prisons in the United States. It examined how people accept or reject media representations and derive meaning from them. Two focus groups of Northern Virginia residents from diverse backgrounds engaged in guided discussions on media and the criminal legal system. Participants analyzed representations, explored their cultural and social meanings, and used lived experiences to affirm or challenge what they had seen. Through this phenomenological inquiry, they reflected on key themes, including the punish-reform dichotomy, witnessing injustice on social media, “prison porn,” and how public opinion influences the system via mediascapes. Additionally, discussions revealed how relationships counter hegemonic narratives, as individuals judged the system through loved ones’ experiences. These findings open new inquiries into disrupting hegemonic carceral norms in American society.
Results of a digital health intervention for Chinese American dementia caregivers
Yiwen Lee (College of Public Health)
Chinese American dementia caregivers experience high psychosocial distress yet face barriers to culturally appropriate support. Digital health interventions offer promising solutions. This study evaluates the effectiveness of WECARE 2.0, a culturally tailored digital health intervention designed to enhance caregiving skills and improve caregiver well-being. Methods: WECARE 2.0 was assessed through a single-arm trial with 48 Chinese American dementia caregivers (2023–2024). The 7-week intervention, delivered via WeChat, provided culturally tailored multimedia articles 6 days a week, covering caregiving and self-care themes. Participants engaged in group chats and had the option to attend three online meetings for social support. Surveys at baseline, 3 months, and 12 months measured program outcomes. Clinical trial ID: NCT05992467. Results: Participants reported high satisfaction (30/35) and perceived usefulness (22/25). Significant improvements were observed in caregiving mastery (effect size = -0.33, p < 0.05) and positive aspects of caregiving (effect size = -1.55, p < 0.005). While reductions in depressive symptoms and caregiving burden were not statistically significant overall, caregivers with elevated depressive symptoms experienced significant reductions in depressive symptoms (effect size = 0.61, p < 0.05) and caregiving burden (effect size = 0.46, p < 0.05), alongside increased life satisfaction (effect size = 0.40, p < 0.05). Those with limited English proficiency showed greater improvements in caregiving mastery (effect size = -0.44, p < 0.05) and reductions in problem behaviors (effect size = 0.40, p < 0.05). Improvements persisted at 12 months. Conclusion: WECARE 2.0 demonstrated feasibility, acceptability, and positive impacts on psychosocial well-being, particularly benefiting caregivers with high distress and language barriers. Findings highlight the potential of digital health interventions for underserved dementia caregivers.
Discerning the Role of Latent Heat Release in Changes of Observed Extratropical Cyclones over the Atlantic and Pacific Storm Tracks: A Lagrangian Approach
Austin Reed (College of Science), James L. Kinter
Diagnosing how global warming has impacted the evolution and impact of Latent Heat Release (LHR) on Extratropical Cyclone (ETC) development remains an open question. At larger scales, factors such as Arctic Amplification, Hadley cell expansion, and tropical upper-tropospheric warming can lead to opposing shifts in ETC activity that vary by storm track basin. Additionally, more atmospheric water vapor as a result of global warming can have more than one effect: either weakening the number or intensity of ETCs to transport the same amount of moisture polewards, or by intensifying individual ETCs through Latent Heat Release (LHR) as a result of extreme precipitation. Despite previous studies focusing on the role of LHR in ETC development, questions about the timing of LHR’s impact on ETC intensity, as well as the role in development of explosive ETCs remains unclear compared to other factors. This research aims to adopt the Multiple Object Tracking algorithm to classify cool-season ETCs within either the Atlantic or Pacific storm tracks, and then separate the distribution of all ETCs from the most intense. Next, an Ertel Potential Vorticity (EPV) tendency inversion will be used to calculate how the influence of LHR on ETC intensity differs between each basin and for each ETC distribution. This EPV approach can more accurately recover the circulation of the ETC compared to other approaches. Differences in the contribution of LHR for each phase of the ETC lifecycle between early (1980-2001) and late (2001-2022) periods can be taken to understand shifts in LHR related to all and most intense ETC activity for each storm track. Improved process understanding of LHR in ETCs can then foster improved predictability on both weather and climate timescales for damaging ETC impacts like extreme precipitation.
Tracking Human Mobility and Behavior during Pandemics: A GPS-Based Analysis of Movement Patterns and Social Distancing
Naren Durbha (College of Public Health), Janusz Wojtusiak
Understanding human movement and behavior is crucial in controlling the spread of infectious diseases, particularly during global pandemics. During COVID-19, many studies used aggregated datasets like SafeGraph to track population movement. A study was conducted to collect and integrate GPS, WiFi, health, and vaccination status data, along with survey responses on social distancing attitudes from participants at George Mason University. A total of 205 participants consented to share GPS data through iPhone or Android applications, generating 232,423,558 data points. To protect privacy, all identifiable participant IDs were replaced with randomly assigned unique identifiers before analysis. The GPS data underwent multiple preprocessing steps, including removing duplicate records based on ID and timestamp, and eliminating data points at (0,0) coordinates, refining the dataset to 10,950,228 data points. Further processing involved noise reduction and home location detection for each participant. After filtering out noise and establishing home as the base location, the remaining GPS coordinates were matched with SafeGraph Points of Interest (POI) within a 250-meter radius. Participants were assigned to the POI with the lowest cumulative score at a given time. This analysis helps identify the types of locations individuals visited, such as healthcare facilities or entertainment venues, and provides insights into movement patterns across populations. These findings contribute to understanding how mobility affects disease transmission, informing public health interventions and policies.
Parkinson’s Disease Biomarker Discovery: Molecular Pathway and Therapeutic Insights
Nicholas Minster (College of Science)
Parkinson’s disease (PD) is a progressive neurodegenerative disorder for which early diagnostic tools and disease-modifying therapies remain elusive. We hypothesized that integrating transcriptomic and proteomic data can enhance PD classification and uncover biomarker signatures. To test this, we employed a random forest pipeline with feature selection on multi-omics data from PPMI and PDBP covering 1617 samples. Hyperparameter optimization used cross-validation, and the final model was evaluated on both an internal test set and an external validation set. Our best-performing pipeline achieved a strong external validation performance (ROC-AUC = 0.8781). The optimized model included 39 features spanning both gene transcripts and proteins. These results underscore the potential of combining transcriptomics and proteomics to detect PD with high accuracy. Analysis of the biological significance implicate inflammatory pathways, lysosomal pathways, extracellular signaling, and vascular changes. While analysis of drug perturbations connected biomarker profiles with disease modifying drug compounds. In conclusion, our findings suggest that multi-omics approaches can robustly distinguish PD from healthy controls, offering a promising avenue for future biomarker-driven diagnostic and therapeutic development. Further validation in larger, diverse cohorts could pave the way for improved patient stratification and earlier intervention.
Optimization Models for Shelter Allocation and Disaster Management
Md Mahfujur Rhaman (College of Science)
Efficient and equitable allocation of shelters for evacuees is a critical challenge in disaster management, complicated by uncertainties such as fluctuating shelter capacities, travel disruptions, and post-disaster resource availability. This study investigates the following key questions: (1) How can equity constraints be integrated into evacuation planning to ensure fair access to shelters? (2) What are the trade-offs between efficiency (e.g., minimizing travel distances and costs) and equity (e.g., prioritizing vulnerable populations, including the elderly, individuals with disabilities, low-income households, and children) in shelter allocation strategies? To address these questions, we propose an optimization framework that integrates pre-disaster planning and post-disaster adjustments, focusing on the balance between equity and efficiency in shelter allocation, with a particular emphasis on prioritizing vulnerable populations. The model will be tested using synthetic demand scenarios, real-world emergency shelter locations, wildfire risk data, and demographic and socioeconomic data, providing a flexible framework for policymakers to design equitable disaster response strategies. By incorporating uncertainty and fairness considerations, this approach aims to improve shelter accessibility and enhance disaster resilience for at-risk communities.
Leveraging Large Language Models (LLMs) for PDDL Model-Based Task Planning in Multi-Domain Environments
Nafisa Mehjabin (College of Engineering and Computing)
In human-robot cooperative environments, robotic task planning relies on predefined models that often lack adaptability and situational awareness. We have seen a growing demand for automation in service industries, domestic assistance, and industrial settings, and it disproportionately affects communities like individuals with disabilities, low-income workers, and those in high-risk roles. This research investigates whether Large Language Models (LLMs) can enhance autonomous decision-making by generating Planning Domain Definition Language (PDDL) world models while incorporating safety constraints. To address critical challenges in deploying robots in real-world settings by ensuring that AI-driven planning prioritizes human safety, this study evaluates four domains: household, laboratory, restaurant, and warehouse, using OpenAI’s GPT-3.5 and GPT-4 for PDDL generation and safety enforcement. The experiment results showed safety constraint satisfaction rates between 60% and 80%, with GPT -4 achieving superior planning success. An iterative feedback approach further improves safety adherence and reliability. The objective of this study is not only to mitigate the risks associated with robotic automation to ensure these systems are designed with human well-being at their core but also to contribute to AI transparency and accessibility by reducing the expertise required to develop safe task plans for robotic systems.
Online Community Size and its Effects on the Specialization of Linguistic Features
Jordan Kidd (College of Humanities and Social Sciences)
Social media has become a larger portion of language input for today’s population, this in part is due to an increase in social media users since the 2020 pandemic. This means that more and more people are using social media, and with that comes an increase in online-sourced language input within their daily lives. However, not a lot of prior research has been conducted on the effects of online communications on language use. Based on prior research, it is predicted that there is an influence of community size on the specialization of language terms used within online communities. Using a Reddit API crawler, two corpora of were amassed based on subreddits of similar topics, music and sports. Each subreddit was picked based on current relevance in the news (e.g., Olympic related sport) and was webcrawled for 1000 recent posts, with 32 music and 20 sports subreddits chosen. The corpora were given a lexical diversity score to determine the ratio of unique tokens to determine whether community size had an impact on the specialization of terms, as well as measuring the average word usage per comment and post based on community size. It was found that there was no effect in community size on lexical diversity scores, as there was no significant effect on lexical diversity as the communities got larger. Current results are inconclusive and further research needs to be conducted to address the limitations found in the study.
Charisma and its Challenges to Measure It
Judith Rautenberg (College of Visual and Performing Arts)
Charismatic individuals can lead a company or enterprise to success by encouraging and engaging people and developing a vision for the group they are leading. Research in the last decades could measure that charisma is a skill set that can be taught and learned, rather than a gift or talent. Two questions we address in this literature review poster are: Which paths did researchers take in the past to measure charisma? How are the different measurements tied to the definition of charisma? Further, we display an overview of the measurements researchers used, and we connect these approaches to the various definitions of charisma. Most of the current research about charisma is based on Max Weber’s works (1968), but he never defined charisma explicitly (Antonakis et al., 2016). As a result, the research field developed different paths to measure this phenomenon. In our research, we found four different indicators: charisma (i) as a signal (Antonakis et al., 2016), (ii) as a leader-follower imitation skill (Katz-Navon et al., 2023; Paul et al., 2001), (iii) in the eyes of the followers (Conger et al., 1997; Howell & Shamir, 2005), and lastly (iv) as effective leadership (Vergauwe et al., 2017; Hogan & Hogan, 2009). However, only Antonakis et al. (2016) approach charisma as an endogenous phenomenon, all other measurements rely on the opinion of the followers which holds the risk of highly subjective data. In this study, we provide relations among different definitions and establish connections with respect to the data acquired.
Fairness in AI-Based Bruise Detection: The Impact of IoU and Confidence Thresholds
Dharmi Desai (College of Public Health)
Artificial intelligence (AI) is increasingly used in healthcare for injury assessment, yet bias in detection models can increase health disparities. Bruises, a critical indicator of Intimate Partner Violence (IPV), are harder to detect on darker skin tones, often leading to misdiagnoses. Existing fairness metrics primarily focus on classification accuracy but fail to account for the additional complexities of Intersection over Union (IoU) and confidence thresholds, both of which are essential for fairness in object detection models. Our study hypothesizes that fairness disparities in AI-based bruise detection are influenced by both IoU and confidence score thresholds and that different fairness measurement approaches capture bias to varying degrees. We evaluate fairness using Demographic Parity (DP) and Equality of Opportunity (EO) across diverse skin tones, employing two methods: (1) Averaging predictions and (2) Retaining individual predictions. We systematically vary IoU and confidence thresholds to assess their impact on fairness outcomes. The analysis was conducted on a dataset of 11,786 images, evaluating fairness on a test set of 1,767 images. Our results show that higher IoU and confidence thresholds increase fairness disparities, particularly in Method 1, which exhibits more fluctuations in fairness outcomes. In contrast, Method 2 provides more stable fairness results, especially at higher IoU thresholds. Stricter IoU thresholds help reduce fairness disparities when confidence requirements increase, while lower IoU thresholds lead to less consistent fairness outcomes. These findings highlight the limitations of traditional fairness metrics and offer actionable insights for bias mitigation in healthcare AI, particularly in applications requiring object localization.
In Tune with Change
MeHaley Babich (College of Visual and Performing Arts)
As music education strives to be a transformative force in society, this in-progress mixed methods study examines the experiences, beliefs, and attitudes of music educators concerning the intersection of the use of multicultural music and guiding objectives of the Reconstructionist Philosophy of Music Education (RPME). The purpose of this study is to highlight how and why teachers incorporate social change practices in their curricula, through multicultural music and RPME objectives. This research aims to uncover the pivotal role of RPME in promoting multicultural music for social change in the music classroom. The questions used to guide this research are: 1. What are the experiences, beliefs, and attitudes of music educators regarding the guiding objectives of RPME and their role in promoting multicultural music for social change in classrooms? and 2. To what degree do music educators incorporate multicultural music for social change in their curricula? This study will utilize a descriptive exploratory sequential mixed methods design. The qualitative strand will explore music educators’ experiences, beliefs, and attitudes, while the quantitative strand will assess familiarity with RPME objectives and measure integration. By understanding this occurrence and frequency of use, the outcomes of this study could provide advocacy that cements the need for music education classes in K–12 education, advocacy for professional development in RPME, and recommendations for changes to current curricula practices for educators, policymakers, and researchers interested in enhancing the transformative potential of music education.