UTSA Runner Research Press (Univ. of Texas at San Antonio)
Not a member yet
6846 research outputs found
Sort by
AI in the Information Environment: Content Moderation, Adversarial Manipulation, and AI Security
Get PDFThis multi-essay dissertation examines the dual role of Artificial Intelligence in the information environment as a tool for content moderation and as a target for adversarial manipulation. The information environment, defined as the collection of individuals, organizations, and systems that interact through the collection, processing, dissemination, and use of information, has been significantly influenced by AI systems that now contribute to shaping, moderating, and filtering digital content. As AI assumes greater responsibility for detecting harmful content, misinformation, and unsafe imagery, it has simultaneously become vulnerable to exploitation by malicious actors seeking to bypass safety mechanisms.
The four essays in this dissertation trace the evolution of AI-enabled content moderation systems and their vulnerabilities as the field has progressed from traditional computer vision to modern large language models. Beginning with the CNN era's focus on image classification systems, the research first addresses adversarial robustness in visual content moderation. As vision-language models emerged, the work evolved to leverage these multimodal capabilities for enhanced moderation explanations and reasoning. With the rise of large language models as prominent content moderation tools, the research then investigates fundamental vulnerabilities in LLM safety mechanisms. Finally, the dissertation examines how these same powerful language models can be weaponized against real-world content moderation systems, completing a comprehensive view of the adversarial landscape across the progression of AI technologies.
Together, these essays illuminate key challenges at the intersection content moderation, adversarial manipulation, and AI security. By analyzing vulnerabilities across both visual and textual domains and developing defensive measures against sophisticated attacks, this research contributes to building more robust AI systems in the information environment. The work emphasizes the importance of a dual security approach that not only studies how these systems can be exploited but also designs defenses that detect and mitigate these vulnerabilities, ultimately promoting safer interactions in our digital ecosystem.Information Systems and Cyber Securit
Examination of the Top Three Traumatic Experiences Among United States Service Members and Veterans with Combat-Related Posttraumatic Stress Disorder
Get PDFMany trauma-focused psychotherapies for posttraumatic stress disorder (PTSD) focus on the most distressing trauma. However, military personnel are often exposed to multiple traumatic experiences. This study aimed to evaluate and categorize the top three traumatic experiences identified by United States (U.S.) military service members seeking treatment for PTSD and compare frequency of trauma types by demographic/military characteristics. Active duty service members and veterans (<i>N</i> = 110) with PTSD identified and ranked their top three most distressing experiences. Behavioral health professionals classified experiences according to one categorical and four dichotomous classification schemes. The categorical scheme included life threat to self, life threat to others, aftermath of violence, traumatic loss, moral injury by self, and moral injury by others. The Life Threat to Self classification represented the largest portion of categorical experiences (43%). Most experiences were dichotomously classified as military-related (86%), combat-related (70%), non-sexual (91%), and trainability (versus futility; 71%). Women were more likely to report sexual traumatic experiences and less likely to report military- and combat-related experiences. Military occupational specialty, number of deployments, time in military, active duty status, and marital status were also associated with different classification rates. There was noteworthy variability in types of experience across top three traumas, especially among certain subpopulations.Psycholog
Social Isolation and Psychological Well-Being of Older Adults: A Longitudinal Examination by Race/Ethnicity and Gender
Get PDFThis two-part study examines how social isolation affects psychological well-being in later life, focusing on subjective well-being (SWB) and loneliness, and how these associations vary by race/ethnicity, nativity, and gender. Using 2008–2018 data from the Health and Retirement Study, Study 1 analyzes the link between overall social isolation and SWB. Study 2 breaks isolation into five subdimensions—marital status, contact with children, other family, friends, and social engagement—to assess their associations with loneliness. Both studies use mixed-effects models and test for subgroup interactions. In Study 1, social isolation was linked to lower SWB, but this association was weaker for Black and Hispanic (US- and foreign-born) adults compared to Whites. Foreign-born Hispanic adults consistently exhibited higher SWB at all levels of isolation. Among women, Black women were less affected by isolation, while among men, US-born Hispanic men showed resilience. In Study 2, overall social isolation was significantly associated with greater loneliness. Being unmarried and infrequent contact with friends were most strongly predictive of loneliness. Subgroup differences revealed that Black women were less affected by marital status isolation, while foreign-born Hispanic women were more impacted by limited contact with children and less affected by reduced social engagement. US-born Hispanic men experienced stronger effects from reduced social engagement. Findings highlight the multidimensional nature of isolation and the need for culturally and gender-sensitive interventions to support psychological well-being in diverse aging populations.Sociolog
Advancing Neutral Instrument Calibration: Development of a Molecular Beam Facility and Velocity Filtering Technique for In-Situ Mass Spectrometers
Get PDFTo characterize surface-bounded atmospheres and the evolution of airless bodies in the solar system, it is essential to understand the sources, losses, and composition of the exosphere. Mercury presents a particularly challenging environment for in-situ measurements due to its tenuous exosphere and the significant presence of spacecraft-originating background signals. Strofio, a neutral mass spectrometer onboard the ESA-JAXA BepiColombo mission, is designed to directly sample Mercury’s exosphere. Accurate calibration and effective background suppression are critical to the success of such instrumentation.
At the start of this work, the Strofio instrument was rendered non-functional due to a launch-induced failure involving a critical electrode (D5). The exact configuration of the damaged component was unknown, preventing proper operation. A comprehensive diagnostic campaign was conducted that involved simulations, laboratory experiments, and iterative flight tests to resolve the present state of the instrument. This process resulted in the successful identification and restoration of a configuration that meets the original operational requirements of the instrument, enabling full scientific functionality.
Following instrument recovery, attention was turned to optimizing Strofio's sensitivity to exospheric neutrals by suppressing the dominant background signals arising from spacecraft outgassing. To address this, a velocity filtering technique was developed and refined through a combination of Monte Carlo simulations and laboratory tests. The resulting configuration was shown to selectively remove low-velocity (thermal) neutral particles and enhance the signal-to-background ratio. Experimentally, the optimized filter configuration reduced background levels by a factor of 40 to 60, significantly improving the likelihood of detecting true exospheric particles at Mercury.
Recognizing the limitations of simulation-based calibration, a molecular beam facility was designed and constructed to replicate the conditions expected in orbit around Mercury. This custom-built facility is capable of generating neutral beams with controlled velocities ranging from 1 to 8 km/s. It is specifically tailored to simulate the 3 km/s relative velocity experienced by incoming particles due to the orbital motion of Strofio. This setup supports the calibration and simulation of in-situ neutral mass spectrometers and serves as a flexible platform for developing future instruments for planetary missions.
This thesis presents an integrated approach to advancing in-situ neutral particle detection, combining instrument recovery, background filtering, and molecular beam development. The technical contributions made through this work have significantly enhanced the readiness of Strofio for scientific operations and have established new methodologies for neutral instrument calibration. These developments will support not only the current mission to Mercury but also future efforts to explore airless bodies and investigate the origins and evolution of planetary exospheres.Physics and Astronom
THE FINANCIAL REPORTING, CORPORATE GOVERNANCE, AND ECONOMIC CONSEQUENCES OF THE HOLDING FOREIGN COMPANIES ACCOUNTABLE ACT (HFCAA)
No full textThe full text of this item is not available at this time because the author has placed this item under an embargo until May 15, 2030.Since the Public Company Accounting Oversight Board (PCAOB) started its international inspection program, it has been unable to inspect auditors in Chinese jurisdictions completely. In 2020, amid growing political and regulatory tensions, the U.S. Congress passed the Holding Foreign Companies Accountable Act (HFCAA) that threatens to delist companies that do not comply with the PCAOB’s inspection regime. Using a difference-in-differences method, this paper uses the enactment of the HFCAA as an exogenous shock to examine the effect of the PCAOB inspection access on U.S.-listed Chinese Companies, particularly Commission-identified issuers. Results show that Commission-identified issuers face some corporate governance and economic consequences of the enactment of the Act. Commission-identified issuers experience a decrease in institutional ownership, as well as an increase in the cost of debt. They also adopted a compliance strategy and switched auditors post HFCAA, which is likely a strategy to avoid being delisted. This resulted in an increase in audit fees. However, I fail to find evidence that the enactment of the Act had any effect on financial reporting quality.Accountin
Investigating the Role of Imperfections on the Mechanical Response of Architected Materials
No full textArchitected materials exhibit specific mechanical behaviors by leveraging the relationship between material properties and geometric configurations often not found in nature. Architected materials have been widely studied for different uses and applications including aerospace, biomedical, control systems, and soft robotics systems. Recent technological advances in additive manufacturing have simplified process such as prototyping, manufacturing of small and large-scale components, and the fabrication of medical devices. The recent rise of additive manufacturing has also facilitated the fabrication and exploration of architected materials. However, architected materials produced through additive manufacturing are highly susceptible to imperfections such as warping, waviness, layer inconsistencies, and precision /consistency issues. These imperfections impact the overall quality and mechanical behavior of the fabricated architected material. Understanding these imperfections and their influence on the mechanical response of architected materials is crucial for maximizing their potential and recognizing their current limitations.
The primary goal of our research is to perform uncertainty quantification (UQ) of the mechanical response and to understand the relationship between these imperfections and the mechanical behavior of architected materials. Specifically, through additive manufacturing, experimental testing, and finite element simulations, we aim to quantify the uncertainty associated with the effects of imperfections on the mechanical behavior of the architected material. Within this context, our research focuses on the role of imperfections in the buckling and post-buckling response of architected materials.Mechanical Engineerin
Impact of K on the Basicity and Selectivity of Pt/m-ZrO2 Catalysts for Methanol Steam Reforming with co-fed H2
No full textThis study investigates the effect of potassium (K) promotion on Pt/m-ZrO<sub>2</sub> catalysts in methanol steam reforming (MSR), revealing critical insights into reaction pathways and catalyst performance. While increasing K loading reduces catalytic activity, it selectively enhances the hydrogen-producing formate dehydrogenation and de-carboxylation pathway. Structural analyses using HR-TEM and DRIFTS show that higher K concentrations block Pt sites and promote agglomeration, reshaping catalytic behavior. Notably, the 3.1% K-promoted catalyst achieves high stability at 358 &deg;C, with a CO<sub>2</sub> selectivity exceeding 80% and minimal methane formation, outperforming the unpromoted catalyst in terms of CO and CH<sub>4</sub> selectivity. Temperature studies further demonstrate reduced CO selectivity at higher temperatures, highlighting distinct advantages of K-doped catalysts. These findings underscore the role of K in enhancing surface basicity and its impact on formate interaction, offering valuable insights for optimizing MSR catalysts and advancing hydrogen production technologies.Biomedical Engineering and Chemical Engineerin
Anion radical mediated photocatalysis: Exploring the photophysical properties and the reducing ability of 4,7-Di(thiophen-2-yl)benzo[c][1,2,5]thiadiazole
No full textPhotoexcited anion radicals derived from organic chromophores have gained significant attention over the past decade as photocatalysts in light-mediated organic transformations due to their strong redox potential in doublet excited states (DES), along with visible to near-infrared (NIR) absorption and emission properties. This report introduces a novel anion radical photocatalyst based on the donor-acceptor-donor (D-A-D) motif, namely 4,7-Di(thiophen-2-yl)benzo[c][1,2,5]thiadiazole (TBT). The study explored the photophysical behavior of the TBT anion radical (TBT–•) and its DES (*2TBT–•) for reducing various aromatic acceptors. Photophysical studies revealed that TBT–• shows absorption and emission bands in the visible-NIR region in its doublet ground state (DGS). Picosecond time-resolved transient absorption (TA) spectroscopy showcased that upon photoexcitation, TBT–• gives rise to visible and NIR TA bands due to the formation of *2TBT–•, that decay with τ = 250 ± 6 ps. The time-correlated single photon counting (TCSPC) method was utilized to probe the reduction of various aromatic acceptors by monitoring the quenching of *2TBT–• lifetime. The Stern-Volmer analysis provided quenching rate constants (kq). At the same time, cyclic voltammetry and excited-state energy (E0,0) estimated the potential for electrochemical half reaction *2TBT–•/TBT to be -2.78 V vs. SCE, highlighting TBT–•as a promising photoredox catalyst for reducing highly stable aromatic acceptors like aryl halides.Chemistr
Bridging Technology Readiness Levels (TRL) and AI Maturity
No full textTechnology Readiness Levels (TRLs) provide a structured framework for assessing technology maturity, originally developed by NASA and now widely applied across industries. This poster explores how TRLs can be utilized to evaluate AI and data-driven technologies, particularly those showcased at the UTSA Los Datos Conference. By mapping AI advancements onto the TRL scale—from initial concept (TRL 1) to fully operational deployment (TRL 9)—we offer researchers and industry stakeholders a standardized method to gauge readiness and guide innovation.
The TRL framework fosters collaboration between academia, industry, and government by establishing a common language for technological maturity. This enables a smoother transition of academic research into real-world applications, ensuring responsible and efficient AI development. For example, ChatGPT currently resides at TRL 7, reflecting successful prototype demonstration in an operational setting. However, if adapted for real-time medical diagnostics, its TRL would decrease, indicating the need for further validation in clinical contexts.
Originally used for space exploration, TRLs now inform diverse fields, including defense, healthcare, and autonomous systems. By applying TRLs to AI, we can better assess development stages, mitigate risks, and accelerate the deployment of trustworthy AI solutions.Information Technolog
Comparative Analysis of Machine Learning Models for Smart Valve Force Injection Attacks in Nuclear Power Plants
No full textThis research investigates the effectiveness of machine learning models in detecting force injection attacks on smart valves used in nuclear power plants. Force injection attacks, which digitally manipulate valve control signals without causing physical damage, pose significant cybersecurity risks to reactor safety and cooling systems. To address this vulnerability, three machine learning approaches; Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Random Forests (RF) were comparatively analyzed. Custom datasets were developed using valve operational parameters such as Reynolds number, valve position, and flow coefficient, simulating both normal and attack scenarios for ball, butterfly, and globe valves across multiple sizes. Each model was trained and tested under both in-range (normal operational variation) and out-of-range (anomalous behavior) conditions. Results indicated that SVM consistently achieved the highest detection accuracy, especially under out-of-range conditions, while ANN and RF demonstrated complementary strengths depending on feature availability and valve type. Notably, the flow coefficient emerged as a critical indicator of abnormal valve behavior. The findings highlight that machine learning-based monitoring systems, particularly those leveraging SVM, offer a practical and robust solution for early detection of actuator-level cyber-physical attacks in nuclear infrastructure. This work bridges mechanical engineering, cybersecurity, and data science, contributing a systematic evaluation framework and emphasizing the need for cross-disciplinary innovation to enhance critical infrastructure resilience against evolving cyber threats.Electrical and Computer Engineerin