33760 research outputs found
Sort by
Reforms Always Have an Impact: Measuring the Efficiency of Medical Personnel Policy in the U.S. Military Health System
No full textThe U.S. Military Health System is among the largest programs in the federal government and one of the most complex health care delivery systems in the world. The Military Health System costs more than $60 billion dollars annually and employs over 200,000 Active Duty and Reserve medics, civilian employees, and contract support personnel operating on hundreds of military installations worldwide. The system is critical to national security and combat readiness. For over a decade, on-going reforms and perturbations affected the system. One of the most significant was a change in Department of Defense policy that transferred day-to-day control over most Active Duty medical personnel from health care administrators at military medical treatment facilities to operational units. Qualitative, quantitative, and normative tests show that this transfer of personnel reduced the overall efficiency of military health care
Single-Cell Cloning of H5N1 HA-Transfected CHO Cells & Protein Expression Analysis
No full textCell Culture & Transfection: CHO cells were seeded in 6-well plates at a density of 2×10⁶ cells/well and cultured for 18 hours. Cells were transiently transfected with 2 µg of pDNA encoding either H5N1 protein + B3.13-BoIgG1, bovine His-tagged control protein (B3.13-His), or human His-tagged control protein (D1.1-His). Transfections were done using polyethylenimine (PEI) Max at a DNA:PEI ratio of 1:6, according to standard protocols.
Single-Cell Cloning: Cells were single-cell cloned manually into 96-well plates to ensure monoclonality. Cells were maintained under standard culture conditions until sufficient confluency was achieved.
Protein Expression Analysis (Dot Blot) & Further Confirmation by Western Blot: His-tagged control proteins (B3.13-His and D1.1-His) were probed overnight at 4°C with biotin-conjugated anti-His primary antibody (1:10,000 dilution in blocking buffer). Membranes were incubated with HRP-conjugated streptavidin secondary antibody (1:20,000 dilution in 3% BSA) for 2 hours at room temperature. H5N1 fusion protein (B3.13 - BoIgG1) was blocked with 5% milk in PBST and probed overnight at 4°C with HRP-conjugated bovine IgG antibody (1:10,000 dilution). No additional secondary antibody was required. Membranes developed using Immobilon® ECL UltraPlus Western HRP Substrate.Avian influenza, caused by influenza Type A virus, poses significant threats to both animal and human health worldwide. Among these, the highly-pathogenic H5N1 subtype has emerged as a critical concern, causing severe global outbreaks in poultry, substantial economic loss, and ongoing zoonotic infections in humans². In this study, we generated recombinant H5N1 proteins through single-cell cloning of transfected Chinese Hamster Ovary (CHO) cells to verify protein expression. We utilized plasmids encoding influenza H5N1 protein (strain B3.13 with Bovine IgG1) and His-tagged control proteins from bovine (strain B3.13) and human (strain D1.1) origins. Cells were transiently transfected using PEI MAX reagent and subsequently expanded in single-cell-derived clones. Protein expression was screened and confirmed via dot blot and Western blot analyses. Our findings demonstrate a robust methodological framework to produce stable recombinant influenza proteins, essential for further research toward diagnostics, therapeutics, and vaccine development.This work was supported by the National Institutes of Allergy and Infectious Diseases Grant No. 5-R01-AI-146063-05 under the National Institutes of Health. Additional funding was provided by the United States Department of Agriculture and the University of Maryland. Special thanks to all members of the Zhu Lab at the University of Maryland’s Department of Veterinary Medicine for their guidance and support
Essays on Labor and Education Economics
No full textThis dissertation investigates the effect of various labor and education policies on individual decisions and outcomes. Chapter 1 examines how a considerable wage increase affects worker financial outcomes and family labor supply decisions. To estimate this causal relation, we leverage quasi-experimental variation from an exam-based promotion policy and use novel data linking census-like banking and household information to personnel records from a large public organization. We exploit the sharp discontinuity in the allocation of promotions among police officers in Colombia to compare the outcomes of nearly identical workers before and after they receive a 25 percent pay raise. We show that within four years, promoted officers accumulate debt equivalent to 160 percent of their annual pay premium. This response is driven by an increase in the use of installment credit, particularly mortgages and personal loans. As a result, officers allocate only one-third of their monthly wage increase to repay debt. We also document a sharp decline in the employment rate of promoted officers' spouses, an effect that is more pronounced among couples with school-age children. Our findings suggest that households rely on credit instruments to adjust their time and consumption allocation in response to a pay increase.
Chapter 2 studies how signaling skills that are specific to college majors affect labor market outcomes of college graduates. We rely on census-like data and a regression discontinuity design to study the impacts of an academic award granted to top performers in a mandatory nationwide exam in Colombia. The award allows students to signal their high level of specific skills when searching for a job. These students earn 7 to 12 percent more than nearly identical individuals lacking the signal. This positive return persists five years after graduation. The signal mostly benefits workers who graduate from low-reputation colleges, and allows workers to find jobs in more productive firms and in sectors that better match their skills. We rule out that the positive earnings returns are explained by human capital. The signal favors mostly less advantaged groups, implying that reducing information frictions about students’ skills could potentially shrink earnings gaps.
Finally, Chapter 3 examines the impact of college financial aid on tuition prices and how colleges respond to changes in their revenues. We exploit exogenous variation from a large-scale program in Colombia, where loan recipients were only allowed to enroll in high-quality colleges. Using a difference-in-differences approach, we find that tuition at these institutions increased by nearly 6.9 percent following the program’s introduction. Additionally, we provide evidence that high-quality colleges maintain selective admission practices, even as demand from high-achieving students rises. This response is likely driven by concerns about reputation and education quality, as evidenced by a constant student-to-faculty ratio despite a growing student body. Colleges achieve this by hiring additional faculty, including instructors with doctoral degrees
NEW EFFICIENT ALGORITHMS FOR NESTED MACHINE LEARNING PROBLEMS
No full textIn recent years, machine learning (ML) has achieved remarkable success by training large-scale models on vast datasets. However, building these models involves multiple interdependent tasks-such as data selection, hyperparameter tuning, and model architecture search-that can lead to nested objectives when optimized jointly. These nested objectives arise because each task both influences and depends on the others. This dissertation aims to develop efficient algorithms to tackle these challenging nested problems in machine learning.
In the first part, we formalize nested ML problems as bilevel optimization tasks and presenting efficient algorithms with theoretical guarantees that solve them. Then, in the second part, we extend to the federated/distributed learning context, examining how algorithmic designs must be adapted to meet the challenges of that environment. Finally, in the third part, we cover challenges with hierarchies in the distributed learning setting including data cleaning, network pruning and constrained problems
Bridging the Gulf of Evaluation in Human-AI Interaction for Knowledge Workers
No full textFor over half a century, user interfaces have served as the primary medium through which humans interact with software systems. To describe this interaction, researchers~\cite{hutchins1985direct, norman1986user} introduced a seven-stage action model encompassing goal formation, intention, action specification, execution, system response, interpretation, and evaluation. Central to this model are two critical challenges—referred to as the Gulf of Execution and the Gulf of Evaluation. The Gulf of Execution represents the gap between a user’s goal and the means available to achieve it within the system, while the Gulf of Evaluation describes the gap between the system’s perceived state and the user’s goals.
The emergence of AI-powered interfaces has reshaped this interaction landscape. Unlike traditional deterministic systems, AI-powered interfaces often exhibit dynamic and unpredictable behaviors~\cite{amershi2019guidelines}, prompting a re-examination of these Gulfs. On the one hand, the Gulf of Execution has narrowed: users can now articulate goals through natural language commands, leveraging large language model (LLM)-powered applications, rather than manually navigating complex menus~\cite{jiang2022discovering, wu2022ai}. On the other hand, the Gulf of Evaluation has widened: AI-generated outputs can be inaccurate or untrustworthy. For example, object detectors may incorrectly classify pedestrians on the road~\cite{hoiem2012diagnosing, simhambhatla2019self}, and LLMs have both intrinsic (contradicted by the source) and extrinsic (not supported by the source) hallucinations~\cite{ji2023survey, liu2023trustworthy}.
The burden of evaluation largely falls on human knowledge workers, who apply knowledge to engage in non-routine problem solving and develop products and services~\cite{janz1997knowledge}. Recent studies on knowledge workers have shown a wide adoption of AI and the practice of evaluating AI-generated results before use~\cite {woodruff2024knowledge, mckinseySurvey}. Therefore, it's important to develop human-centered AI systems to bridge the Gulf of Evaluation for knowledge workers. To bridge the Gulf of Evaluation in the above two dimensions-inaccuracies and lack of trust-we must re-imagine human-centered AI systems beyond chatbots and design novel human-AI interactions. Therefore, the overarching research question of the thesis is: \textit{How do we design human-centered AI systems to bridge the Gulf of Evaluation in human-AI interaction for knowledge workers?}
The thesis addresses this research question by introducing techniques to reduce inaccuracies and foster trust for representative knowledge workers. More specifically, to reduce inaccuracies, we developed \textsc{TutoAI}~\cite{chen2024tutoai}, a cross-domain framework for AI-assisted mixed-media tutorial creation on physical tasks. We present an approach to identifying, assembling, and evaluating AI models for creating mixed-media tutorials from instructional videos, along with an interface for creators to refine AI-generated components. To enhance trust, we developed an AI-assisted visual analytics tool called \textsc{COALA}~\cite{chen2025comparing} for a multilingual collaborative writing dataset. We contribute several interpretable techniques, including interactive clustering, textual pattern explanations and dedicated data visualizations to foster trust among communication researchers. To thoroughly evaluate machine learning models and build trust before deploying them in high-risk applications, we developed \textsc{Safeguard AI} for safety experts —a visual analytics tool powered by AI agents that reveals model inaccuracies and ensures regulatory compliance. Collectively, these systems highlight how human-centered techniques can effectively bridge the Gulf of Evaluation for diverse knowledge workers
Uncovering, Understanding, and Mitigating Social Biases in Language Models
No full textSocial science research has highlighted stereotypes and discrimination based on race/ethnicity or gender, often inferred from names, exacerbating social inequality. Contemporary natural language processing (NLP) systems, including large language models (LLMs), trained on extensive but potentially biased corpora may inadvertently perpetuate these biases. This dissertation aims to uncover, understand, and mitigate social biases in NLP systems, particularly through the lens of analyzing biases related to first names.
The concept of counterfactual fairness serves as a guiding principle, where model predictions should ideally remain consistent despite name substitutions that preserve the original semantic meaning. We leverage name substitution to investigate biases in NLP systems, offering several advantages. First, the automatic generation of diverse instances through name substitution streamlines bias detection without requiring manual data creation. Second, examining model behavior across different first names in an open-ended space can reveal biases that are not captured by pre-defined diagnostic tests. Lastly, using first names for bias identification aligns with real-world applications concerned with individual fairness.
Using name substitution as our primary technique, we study three types of biases in NLP systems: stereotypes about personal attributes, occupational biases, and biases about romantic relationships. Stereotypes about personal attributes emerge when a model infers someone's personality from inputs describing social interactions. Occupational biases encompass both hiring discrimination and gender-occupation stereotypes. Romantic relationship biases include heteronormative assumptions and prejudice against interracial couples.
To study stereotypes about personal attributes, we introduce a framework that uncovers model biases in social commonsense reasoning tasks and show that both demographic associations and tokenization artifacts contribute to observed disparities. For occupational biases, we demonstrate that LLMs can exhibit discriminatory patterns in simulated hiring tasks and stereotypically associate gendered names with gender-dominated professions. We further analyze these patterns by studying the contextualized embeddings and propose a consistency-guided finetuning method to mitigate such biases. Finally, in the domain of romantic relationship prediction from conversations, we find evidence of heteronormative bias and underprediction of romantic relationships for couples involving Asian names.
In sum, these contributions offer a comprehensive examination of first-name-based biases in language models, providing insights into their underlying mechanisms, and present actionable mitigation strategies. This work takes a step toward developing fairer, more interpretable, and more inclusive language technologies
THE INFLUENCE OF THEATRICAL DESIGN ON THE FORMATION OF GENDER AND SEXUALITY IDENTITIES
No full textThrough a review of queer theory and Performance theory, this thesis gains insight into the ways that Academica has critically linked design and the formation of queer and trans identity. Starting this conversation is a grounding of the definitions of queer and trans within the site of queer and trans theory. Looking at works of Butler, Sedgwick, Berlant, and Foucault, to define the queer and trans subject to works then as the starting point from which the thesis begins to expand into performance and design. This thesis then starts drawing parallels, using Muñoz and Bulter’s enacting of performance as a site from which gender is constructed (taking care to note that performativity is rooted in the quotient and not necessarily a function of performance, but can be applied to such). Then I explore scholarship that thinks through theatre design critically, starting with the works by Jane Collins and Andrew Nisbet. Then the thesis adds these two elements together to explore the ways in which design can impact that identarian project, as well as explore ways in which it does not, by examining a case study, Circle Jerk. I then expand on the linkages between design in live performance to how people embody and express gender by examining another case study Playing Wolve
ASSESSING THE EFFECT OF AMINOETHOXYVINYLGLYCINE AND 1- METHYLCYCLOPROPENE ON PREHARVEST FRUIT DROP AND CRACKING, QUALITY, AND RELATED GENE EXPRESSION OF ‘AMBROSIA’ AND ‘FUJI’ APPLES IN THE MID-ATLANTIC
No full textPreharvest fruit drop, cracking, and advanced maturity prior to commercial harvest can promote significant losses in important apple cultivars such as ‘Ambrosia’ and ‘Fuji’ grown in the mid- Atlantic. Aminoethoxyvinylglicine (AVG) and 1-methylcyclopropene (1-MCP) are ethylene inhibiting plant growth regulators (PGR) that can alter preharvest fruit drop, cracking, maturity and quality by impeding ethylene biosynthesis and perception, respectively. However, integrative studies examining the impacts of such PGR treatments in ‘Ambrosia’ and ‘Fuji’ are lacking in the mid-Atlantic region. The objectives of this work were to, first, characterize and compare the effects of different rates of preharvest ethylene inhibiting PGRs on fruit drop and cracking, ethylene production, physicochemical properties, skin color, and transcript accumulation of associated genes in 'Ambrosia' and ‘Fuji’ apples throughout ripening on-the- tree, and second, to use multivariate analysis to detect statistically significant correlations amongst evaluated variables. Our results show that both full-rate AVG (130 mg a.i. L−1) and 1-MCP (150 mg a.i. L−1) significantly reduced preharvest fruit cracking with respect to control fruit in ‘Ambrosia, but not in ‘Fuji’, suggesting a cultivar-specific effectiveness under mid-Atlantic conditions. Furthermore, amongst all treatments, full-rate AVG (130 mg a.i. L−1) displayed the lowest ethylene production and transcript accumulation of ethylene -related genes, ultimately delaying ripening. 1-MCP (150 mg a.i. L−1) and half-rate AVG (65 mg a.i. L−1) treatments exhibited an intermediate performance between full-rate AVG and control fruit, presenting a reduction in ethylene production and transcript accumulation of ethylene-related genes when compared to the latter. Regarding color, apple skin blush and anthocyanin-related gene expression was lowest in full-rate AVG (130 mg a.i. L−1) treated fruit, explaining why this fruit met the minimum requirement of 50% red blush coverage one week later than all other treatments in both assessed cultivars. Correlations amongst the assessed features were also identified. This work specifies a framework for the use of ethylene inhibiting plant growth regulators under mid-Atlantic environmental conditions
UNEARTHED: RESILIENT COMMUNITIES FOR THE UNHOUSED IN LAS VEGAS
No full textHomelessness encompasses various forms, including unsheltered, sheltered, inconsistent, or at-risk, each presenting unique challenges. Unsheltered individuals, such as those living in the tunnels under the Las Vegas Strip, face severe risks due to exposure to harsh elements, contaminated water, and limited access to food, medical care, and educational resources. These conditions can lead to severe health issues or death. Addressing this crisis requires a holistic approach. A proposed solution is the development of a supportive housing campus that provides essential services and resources aimed at transitioning individuals out of homelessness. This campus would offer safe, stable housing and access to medical care, food, education, and supportive services through four distinct stages. By bringing this vulnerable population above ground into a safe environment, the campus would mitigate health risks and foster pathways to permanent housing and stability, ultimately improving their quality of life and health outcomes
PALS 2025 : Deliverable 8
No full textThis Presentation is the compilation of the Final Boards presented at the end of the PALS Takoma Langley Green Isle Project in Spring 2025. This includes the perspectives and results from the Prototype Team, Community Engagement Team, and Masterplan Team of the ARCH408 course.https://drive.google.com/file/d/15qIrKPouYlteVbF9KG2yRFOBdClyERhD/view?usp=drive_lin