Mason Journals (George Mason Univ.)
Not a member yet
3256 research outputs found
Sort by
Mandating Diversity: The Impact of California’s SB826 on Female Board Representation
No full textGender diversity in corporate governance draws increasing attention from firms as they consider how to improve board representation. In 2018, California passed Senate Bill 826 (SB826), the first U.S. law requiring publicly traded firms to appoint at least one female director. The repeal of SB826 in 2022 created an opportunity to test whether such mandates lead to structural change that lasts, rather than temporary compliance. We analyze over 100,000 board appointment records from BoardEx from 2002-2025 using a difference-in-differences (DiD) design with firm and year fixed effects. Firms from California are compared to matched control firms in New York, Texas, Florida, Illinois, and Massachusetts, states without board diversity mandates but with comparable governance environments. We find that female board representation in California increased by 4.7% during SB826 enforcement and remained 3.7% higher after repeal, relative to control firms. These changes were statistically significant, suggesting a durable policy effect. We also observe weaker upward trends in control states, implying potential national spillover. Our findings indicate that mandates can trigger structural behavioral shifts in corporate governance, even when time-bound and repealed. Future work will explore whether these changes impact firm valuation, and whether reputational or institutional forces can help sustain progress after the mandate ends
Cross-Asset Momentum Spillover Effects
No full textCross-asset momentum spillover, which describes the trend of the price changes of one asset connected to the other assets, is an important aspect of financial markets. Such a situation is typically thought to arise from nudges or reciprocal interdependencies of the underlying products. Classic linear models do not generalize well over time-series data and do not capture complex, non-linear relationships. In addition, with the delayed information spreading, the linear model is challenged again due to its inability to understand the sentiment of such information. The need for sophisticated computing methods in order to identify and measure individual spillovers is the focus of this investigation. Nowadays, we are able to use daily stock prices for feature engineering and convert them to many momentum indexes such as Rate of Change and Moving Average differences. We develop a deep learning framework utilizing a three-layer Gated Recurrent Unit (GRU) neural network to detect and quantify complex nonlinear dependencies and delayed information diffusion underlying momentum spillovers across paired equities. Momentum-based features—including Rate of Change and Moving Average differentials—are constructed from daily stock returns data spanning 2022 to 2023. The GRU model, trained with a 30-day lookback window, is applied to forecast future returns for stock pairs such as TSLA–F and NKE–ADDYY, which represent competitive and supply chain relationships. Model performance is evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared metrics. While the explanatory power is modest (e.g., R² = 0.0103 for the TSLA–F pair), the model captures weak yet statistically meaningful predictive signals. These findings underscore the potential of recurrent neural networks to reveal subtle cross-asset dynamics in noisy financial environments, offering valuable insights into the mechanisms of momentum contagion in modern equity markets
From Imagery to Insight: Applying Multimodal Large Language Models for Rapid Disaster Impact Assessment
No full textNatural hazards continue to pose persistent and growing threats to infrastructure systems and communities worldwide. In the aftermath of a disaster, rapid damage assessment is critical for directing emergency response and allocating resources effectively. However, first responders and decision-makers often lack reliable, real-time validation of disaster impacts, resulting in delays in deploying critical aid and support. Most current methods rely on manual inspections and individual surveys, which are time-consuming, inconsistent, and difficult to scale across large, affected areas. Recent advancements in artificial intelligence and vision-language models—particularly in Multimodal Large Language Models (MMLLMs)—offer the potential to automate image-based damage evaluation, making the process faster, more objective, and scalable. Drawing on established engineering guidelines for disaster damage assessment—such as those from STEER, FEMA, and ATC—a set of metadata fields was identified. These fields were used to annotate an expert-curated dataset of post-disaster imagery. Using prompt engineering and the Gemini API, automated damage assessments were generated from the imagery. The predicted outputs were then used to evaluate the performance of the MMLLM across post-disaster scenarios. A confusion matrix approach was applied to assess both binary and multi-class classification performance, from which accuracy, precision, and recall metrics were computed. This automated approach to damage analysis demonstrates strong potential to improve disaster response by enabling standardized assessments that can guide resource allocation and rescue operations more efficiently than traditional methods
Unveiling Counterparty Risk: A Multi-Dimensional Assessment of Governance, Regulation, and Compliance Exposure in Global Cryptocurrency Exchanges
No full textCryptocurrency exchanges have become essential infrastructure in the global digital asset economy, but their continued expansion has exposed critical vulnerabilities stemming from opaque governance, fragmented regulatory environments, and risky operational practices. These vulnerabilities have important implications for market participants and regulators, given the reliance of both retail and institutional participants on exchanges for critical functions such as trading access, liquidity, and asset custody. Prevailing methods for evaluating exchange risk tend to emphasize surface-level indicators such as trading volume or user growth, offering little insight into the structural and legal dimensions that underpin systemic vulnerability. To address this gap, this study develops a structured dataset of over 250 active and defunct centralized cryptocurrency exchanges, systematically categorizing each based on governance models, regulatory exposure, product offerings, and geographic reach to produce detailed operational and legal profiles for comparative analysis. Our analysis identifies a clear relationship between jurisdictional permissiveness, limited ownership transparency, and elevated compliance risk. Notably, exchanges situated in jurisdictions lacking enforceable licensing frameworks and organizational transparency are disproportionately represented among platforms offering leveraged derivatives. Taken together, these findings highlight how structural factors such as regulatory permissiveness and clarity in ownership structures can serve as leading indicators of exchange-level risk. By centering structural and jurisdictional characteristics, this study advances a more robust foundation for evaluating exchange-level risk, enhancing the practical relevance of risk assessments for regulators and informing investment decision-making
A Review of Extreme Space Weather Event Impacts on Critical Infrastructure
No full textAs our world becomes increasingly more reliant on interconnected technologies, space weather events, including geomagnetic storms, pose consequential risks to modern critical infrastructures. The May 2024 Gannon Storm, one of the most intense storms in recent decades, disrupted satellite operations, power grids, aviation routes, and high-frequency communications worldwide. This study investigates existing research on the wide-ranging repercussions of the May 2024 Gannon Storm, including direct impacts (e.g., satellite degradation, power grid fluctuations), cascading effects (e.g., communication blackouts, supply chain disruptions), and broader systemic consequences across multiple domains. There exists significant research data to draw upon from NASA’s DSCOVR satellite, NOAA’s SWPC alerts, and magnetometer readings from global observatories. To begin consolidating the available data and quantifying these impacts, our initial study focuses on authoring a novel software based literature review analyzer app, leveraging various tools for analyzing millions of research papers. The app, which is integrated with different search engines, such as Semantic Scholar and Crossref, and is configurable for multiple search categories, provides an automated way to locate, consolidate, and deduplicate matching research papers. The app also retrieves important research papers parameters such as DOI, Abstract, Citations, Fields and Urls. By surveying the current body of literature, this review highlights emerging patterns in the assessment of space weather risks and identifies critical knowledge gaps for future research
Mathematical Modeling of Knowledge Transfer between Students and Mentors to Estimate Ideal Quantities of Mentors for Student Populations Using Optimal Control Theory
No full textA shortage of educators has always been a concern across all educational institutes in the US. In education, the main goal is to progress students adequately and evenly; however, it is inevitable for some to fall behind. To combat this, upwards of 30% of students graciously volunteer their time as mentors. This population is limited; therefore, it is crucial to optimize the mentor assignments maximizing students benefited and minimizing mentors employed. In this project, a compartmental model of differential equations was used to describe the interactions between students and mentors. The compartments of a Susceptible-Exposed-Infected-Recovered (SEIR) model are modified to describe the positive propagation of knowledge. As a result, the Amendable-Learning-Informed-Unlearned (ALIUM) model describes the spread of information, where instead of an Infected category, the Informed compartment holds the population of students that were exposed to information through other students, students in the process of learning, and mentors each with unique transmission rates (β1, β 2, β3). The M variable is used to keep track of the percentage of Informed students that are required as mentors. To optimize the M variable, Optimal Control Theory is carried out using Pontryagin’s Maximum Principle and the Forward-Backward Sweep Algorithm with the aim to minimize the number of tutors necessary and maximize the informed population. Preliminary results show that, with a nonlinear control, 30% of the Informed population must be employed as tutors. Mentors’ high employment rate is needed during the first quarter of the whole learning period, before gradually declining to 0% of the Informed population by the time learning is finished. Future research hopes to explore heterogeneous learning speeds for students. Also, a further step is to shape the model according to real world data using Physics Informed Neural Networks (PINNs). This work aligns with UN Sustainability Goal #4: Quality Education. 
Mechanistic Interpretability Uncovers Biological Hypotheses in a Deep Learning Model for Breast Cancer Histology
No full textPredicting spatial gene expression from tissue images provides a scalable method for molecular profiling, but the deep learning models used are often opaque. This lack of internal transparency hampers clinical trust and obscures biological insights. Our research applies mechanistic interpretability to a state-of-the-art model, iStar. We extracted learned deep features from an intermediate layer and correlated them against traditional morphological traits, spatially-resolved gene expression, and expert cell-type annotations in breast cancer (n=167780 cells). This analysis identified a deep feature (cls45) that robustly represents proliferative invasive tumor cells (p<1e-100), characterized by high metabolic (SCD, r=0.60; FASN, r=0.59) and luminal (FOXA1, r=0.58) genes and a morphology of high nucleus-to-cytoplasm ratios (r=0.39) and small cell area (r=−0.24). A second feature (cls53) represents the immune-stromal microenvironment, corresponding to large, elongated cells (r=0.41) and expressing stromal (CCDC80, r=0.56) and chemokine (CXCL12, r=0.53) markers. Spatially, these features form coherent, mutually exclusive domains (Moran's I > 0.98) and are strongly anti-correlated (r=−0.50), demonstrating that the model learned the fundamental tumor-stroma architecture without explicit labels. By reverse-engineering deep features into biological concepts, we validate the model's reasoning and provide a workflow for generating testable hypotheses (e.g., a functional link between cls45’s specific visual phenotype and metabolic gene upregulation). This converts the predictive model from a black box into an interpretable tool for biological discovery, enabling the scalable identification of novel biomarkers
Understanding the Effectiveness of State-of-the-art Deep Learning Models on Vulnerability Detection
No full textSoftware vulnerability detection is a critical area in cybersecurity, and recent advances have explored the use of deep learning (DL) to automate this process. However, existing DL-based methods often suffer from slow performance on real-world datasets and fail to capture the relationships between functions in large codebases. The original VulBG study proposed a novel approach by using a Behavior Graph Model to extract and connect semantic slices of code, therefore enhancing the effectiveness of baseline DL models in detecting vulnerabilities. Key components of VulBG were reimplemented from scratch, including data loading, slicing, embedding with CodeBERT, clustering via K-means, and graph embedding using Node2Vec, due to limited or incomplete scripts in the original repository. A neural network classifier was then trained using both baseline features and behavior-based graph features. The model was evaluated on real-world C function datasets using standard metrics such as accuracy, precision, recall, and F1-score. The replicated model achieved an F1-score of 0.5791, closely matching the original study and demonstrating improved recall through behavior-based features. This replication confirms that incorporating inter-function semantic relationships via Behavior Graphs can significantly improve DL-based vulnerability detection. It also shows practical changes in reproducibility and suggests potential improvements such as enhanced slicing techniques and model fusion with pretrained embeddings
Triboelectric Nanogenerator for Self-Powered Wearable Sensors
No full textAs the Internet of Things becomes increasingly integrated into our daily lives, their large volume and need for periodic power replacement has highlighted how traditional batteries are unsustainable for the next generation of smart devices. Triboelectric nanogenerators (TENGs), which generate electricity from everyday mechanical motion, offer a promising alternative, but further improvement of the output performance is still required for usage toward practical applications. This study investigates if silica derived from tetraethyl orthosilicate (TEOS) can improve the performance of polydimethylsiloxane (PDMS), a common electron accepting material used in TENGs, through their electron-attracting abilities. A layer of PDMS and three configurations of silica-PDMS composites were tested against human skin in ten repeated trials, recording the maximum voltage in each. By taking the average of the maximums, the preliminary results suggest that adding silica did not improve the performance of PDMS, decreasing the average voltage output by 50% - 88%. Specifically, the configurations with silica particles after gelation decreased the performance the most (76% - 88%), whereas mixing PDMS with silica before gelation decreased the performance by 50%. Thus, more experiments and tests should be conducted in that direction to confirm if silica can improve the performance of PDMS. If optimized, TENGs featuring silica-PDMS composites could offer efficacy with human skin, high flexibility, and biocompatibility, opening the door to wearable electronics, biomedical sensors, and more. Furthermore, their scalability, low cost, and energy generation performance make it applicable to civil engineering applications, such as harvesting electricity from transportation and foot movement, contributing to a sustainable future
The Spillover Effect of Chinese Market Volatility on US Technology Stocks.
No full textVolatility shocks originating in China’s equity markets increasingly reverberate across global technology sectors, yet the magnitude and timing of these spillovers to U.S. firms remain poorly quantified. This study examines how swings in the Shanghai Composite and CSI 300 indices propagate to the share prices of 25 U.S. technology companies that derive at least 10 % of revenue from mainland China. Daily price data spanning 2010 – 2024 are merged with firm-level revenue exposure and analyzed using vector autoregression, Granger causality tests, and event-window techniques around 42 major Chinese market shocks. Results show that a 1 % negative innovation to the Shanghai Composite leads, on average, to a 0.33 % decline in high-exposure U.S. tech stocks within two trading days (p < 0.01), explaining up to 14 % of their return variance, compared with 4 % for low-exposure peers. Impulse-response functions reveal that the effect dissipates after six days, while variance-decomposition indicates that Chinese volatility accounts for 27 % of forecast error variance during tariff-escalation periods. These findings suggest investors systematically price Chinese macro-risk into U.S. technology valuations, underscoring the need for portfolio hedging and corporate diversification of revenue streams