1393 research outputs found
Sort by
Bridging Cognitive Psychology and Natural Language Processing: A Bias-Detection Framework for Large Language Models
Get PDFBias in large language models (LLMs) poses substantial risks to fair and accurate information processing, particularly in high-impact contexts such as news dissemination and content moderation. These models often learn and unintentionally amplify systematic biases, ranging from confirmation and negativity to anchoring and partisanship, present in the data on which they are trained, thereby distorting public discourse and potentially fueling misinformation. Building upon theories in cognitive psychology and AI-decision-making, the Cognitive Bias in Artificial Intelligence Theory (CoBAIT) becomes the interdisciplinary theoretical foundation of this work.
Drawing on both cognitive bias theory and a design science approach, this dissertation develops a novel CoBAIT-Informed Fine-Tuning (CIFT) bias-detection artifact that integrates short, text-based “context snippets” during LLM fine-tuning. The framework is empirically evaluated using the BABE dataset, which contains over 3,700 sentences from politically oriented news articles annotated by expert raters. By prepending domain specific cognitive bias language known to evoke those heuristics in humans, such as confirmation or negativity bias, to each sample, the LLM gains a clearer “mental model” of what constitutes biased language. When compared to a baseline DistilBERT model, results show an approximate 1–2% improvement in F1 score and a notable reduction in false positives, indicating that the enhanced models are more effective at discerning biased and neutral content. Although the number of false negatives increases, this trade-off proves beneficial in domains where the cost of overclassifying bias far outweighs the risk of occasionally missing biased text, such as where impartiality is particularly important.
Beyond demonstrating technical feasibility, this dissertation contributes to the literature by bridging psychological theory, design science methodology, and advanced Natural Language Processing (NLP) practices, showcasing how explicit cognitive bias frameworks can yield measurable improvements. The findings have implications for large-scale platforms such as social media networks, where small percentage gains can translate into thousands or even millions of accurately identified biased posts, and for smaller, specialized communities, where higher precision and transparency are critical to sustaining trust. Future directions include extending these techniques to multilingual contexts, more complex bias typologies, or otherLLM architectures (e.g., GPT or Llama), thereby refining how cognitive principles are harnessed to build more equitable and reliable AI-driven content analysis systems
Aeolus-DS: Dust-Aware AI Decision Support for Coccidioidomycosis (Valley Fever) — A Design Science Research Framework Integrating Aerosol Remote Sensing, Land Disturbance, and Clinical Sentinel Signals
No full textThis paper presents Aeolus-DS, a design science research (DSR) artifact that integrates aerosol remote sensing (MAIAC AOD; dust fraction), mesoscale meteorology and soil moisture (ERA5), land-disturbance telemetry (construction, off-road vehicle activity, nightlights) and clinical sentinel signals (syndromic ED chief complaints, pneumonia rule-out) into a dust-aware, AI-driven early warning and decision-support system for Coccidioidomycosis (Valley Fever). Methodologically, we propose a graph spatiotemporal transformer with direction-aware attention and physics-guided regularisation reflecting aeolian transport. Using county-week panels (2014-2024) for the U.S. Southwest (AZ–CA–NV), Aeolus-DS improves now-casting mean absolute error (MAE) by 18% and two-week area under precision-recall curve (AUPRC) by 21% over strong baselines (XGBoost, LSTM). Role-based “action cards” translate probabilistic forecasts and uncertainty into targeted mitigations (site watering cadence, temporary grading pauses, N95 staging, clinician test prompts). We evaluate predictive skill, calibration, runtime, interpretability and stakeholder usability, and discuss governance, ethics and portability to other dust-borne mycoses in climate-stressed regions
Harnessing Large Language Models for Passive SCADA Security Risk Assessment: A Case Study
No full textSupervisory Control and Data Acquisition (SCADA) systems play a vital role in critical infrastructure, but outdated protocols and delicate operational requirements undermine their cybersecurity. Active scanning poses a risk of disruption, which drives the need for passive methods. This paper investigates using three large language models (LLMs) to assess SCADA risks by analyzing Wireshark captures of network traffic without interfering with system operations. Tested on a Siemens S7 -1500 PLC scenario, the proposed framework processes traffic data and produces risk reports non-intrusively. The framework effectively identifies vulnerabilities, assesses protocol-specific risks, and generates structured risk reports as an alternative to active vulnerability scanning. Per IEC 62443 standards, this method strengthens SCADA security while maintaining operational continuity
FastTree-Guided Genetic Algorithm for Credit Scoring Feature Selection
No full textFeature selection is pivotal in enhancing the efficiency of credit scoring predictions, where misclassifications are critical because they can result in financial losses for lenders and exclusion of eligible borrowers. While traditional feature selection methods can improve accuracy and class separation, they often struggle to maintain consistent performance aligned with institutional preferences across datasets of varying size and imbalance. This study introduces a FastTree-Guided Genetic Algorithm (FT-GA) that combines gradient-boosted learning with evolutionary optimization to prioritize class separability and minimize falserisk exposure. In contrast to traditional approaches, FT-GA provides fine-grained search guidance by acknowledging that false positives and false negatives carry disproportionate consequences in high-stakes lending contexts. By embedding domain-specific weighting into its fitness function, FT-GA favors separability over raw accuracy, reflecting practical risk sensitivity in real credit decision settings. Experimental results show that FT-GA achieved similar or higher AUC values ranging from 76% to 92% while reducing the average feature set by 21% when compared with the strongest baseline techniques. It also demonstrated strong performance on small to moderately imbalanced datasets and more resilience on highly imbalanced ones. These findings indicate that FT-GA offers a risk-aware enhancement to automated credit assessment workflows, supporting lower operational risk for financial institutions while showing potential applicability to other high-stakes domains
Navigating intent-based networking: from user descriptions to deployable configurations
No full textNetwork automation development has accompanied network evolution due to its significant role in speeding up and simplifying network operations. Emerging networking and computing paradigms such as information-centric networks, next generation networks, cloud, and edge computing and recent innovative technologies, such as the Internet of things (IoT), enabled novel network services (such as the Internet of Vehicles (IoV), context-aware applications, virtual reality, and augmented reality) that demand complex configurations and management. Intent-based networking (IBN) is a promising networking paradigm that provides abstract and autonomous network management. IBN promises to simplify configuring networking devices, allowing network engineers and service providers to focus on providing the expected services and continuously verifying that the network operates within the desired status. An IBN process starts by expressing the user requirement in a high-level descriptive format. Then, the IBN system translates these requirements to a low-level deployable format in a process called intent translation. In this work, we formally define the intent translation process and propose a generic intent translation system. Furthermore, we review the research on intent translation published between 2018 and 2022. We analyze and classify the proposed intent translation schemes and discuss the challenges and recent trends in intent translation
Analyzing the Effectiveness of Information Security Compliance on Cloud Based Small and Medium Enterprises (SMEs)
Get PDFAs cloud computing becomes integral to business operations, security compliance is crucial for the success and sustainability of Small and Medium-Sized Enterprises (SMEs). While cloud adoption offers cost efficiency, scalability, and operational flexibility, it also introduces regulatory and security challenges. Unlike large enterprises with dedicated compliance teams and substantial budgets, SMEs often lack the financial resources and expertise to implement security frameworks effectively. Compliance with industry standards such as ISO 27001, SOC 2, NIST 800-171, and PCI DSS is particularly difficult, yet non-compliance can result in severe financial penalties, reputational damage, operational disruptions, and heightened cyber risks. Despite growing regulatory requirements, many SMEs perceive compliance as an expense rather than a strategic investment. This perception leads to reactive approaches, where security issues are addressed only after they arise, increasing risk exposure and placing SMEs at a disadvantage in industries where compliance, data privacy, and cybersecurity transparency influence customer trust and business partnerships. Given these challenges, there is a need to examine how security compliance efforts impact SME business performance, not only in terms of risk reduction but also in financial stability and operational resilience.
Existing research primarily focuses on the technical and regulatory aspects of security compliance, yet there is a gap in understanding its influence on SME business outcomes such as financial sustainability, customer trust, and operational resilience. While some studies explore barriers to compliance, few empirically examine the correlation between security compliance efforts and key business performance indicators in cloud-based SME environments. This study addresses that gap by employing a quantitative research method to analyze the relationship between security compliance strategies and SME business success.
A structured survey was conducted among SME leaders, IT security managers, and compliance officers to collect data on compliance challenges, risk management efforts, security investments, and industry best practices. The study evaluates which compliance measures contribute most to SME competitiveness, cost efficiency, and scalability. Key areas of focus include regulatory compliance reviews, structured risk assessments, employee security training, incident response preparedness, and vendor risk management.
Findings reveal that security compliance is not merely a regulatory obligation but a strategic enabler of business success. SMEs that invest in proactive compliance experience cost savings, stronger customer confidence, and improved scalability in cloud environments. Regular compliance audits and structured risk assessments are identified as strong predictors of business stability, while security training programs directly enhance incident response efficiency, reducing downtime and financial losses from security breaches.
From a theoretical perspective, this study demonstrates how regulatory preparedness and compliance-driven security strategies influence SME business performance. Empirical evidence reinforces the understanding that compliance initiatives not only strengthen risk mitigation efforts but also contribute to financial performance and market competitiveness.
Beyond academic contributions, this study provides practical insights for SMEs, policymakers, and industry regulators. It highlights the need for SME-friendly security compliance frameworks that balance regulatory requirements with cost-effective implementation. Regulatory bodies can leverage these findings to develop tailored compliance guidelines, while SME leaders can use them to prioritize compliance investments that enhance both security resilience and business sustainability. Ultimately, this study reinforces that security compliance is not a barrier to SME growth but a driver of trust, operational efficiency, and long-term success. By shifting the perception of compliance from a financial burden to a business enabler, SMEs can fully leverage cloud-based infrastructures while safeguarding sensitive data, reducing cybersecurity risks, and achieving greater financial stability
Natural Language Processing Applications in Medical Data
Get PDFNatural Language Processing (NLP) is the utilization of Artificial Intelligence / Machine Learning in understanding human language. NLP is increasingly being applied in the realm of healthcare as it can make information processing highly efficient through data summarization. In this modern day of technology, data is increasing at an unprecedented pace, and AI can assist in organizing unstructured information. In this use case, Named Entity Recognition, which is the classification of words, can recognize key terms in medical information. This study provides in depth research on the foundations of Natural Language Processing for healthcare use cases. Furthermore, comparison of AI models were put in place to apply information gained in order to see which models are most effective in Named Entity Recognition for the purpose of healthcare term. Specifically, open-source models like BERT and SpaCy were fine-tuned to process medical texts. In some cases, synthesized data was created to generate more results in a controlled environment.https://scholar.dsu.edu/research-symposium/1058/thumbnail.jp
Is Al-Driven threat detection an effective substitute for current threat detection architectures?
Get PDFThis research evaluates the use of Artificial Intelligence (Al] in the development of cyber defense systems.https://scholar.dsu.edu/research-symposium/1056/thumbnail.jp
Quantifying Dynamic Response in Web Application Honeypots
Get PDFHoneypots are intentionally vulnerable computers designed to attract and analyze malicious activity. While an effective cybersecurity tool, research quantifying the effectiveness of various honeypot web services that respond dynamically is limited.https://scholar.dsu.edu/research-symposium/1054/thumbnail.jp
A Teacher Apprenticeship Pathway in a Rural, Midwest State: Perspectives of Teacher Apprentices
Get PDFThe national teacher shortage makes it challenging for principals and superintendents to hire certified teachers. To address this problem, a university in a rural state in the Midwest partnered with their state’s agencies to develop a teacher apprenticeship pathway (TAP) for 78 paraprofessionals working in the state’s public, non-public, and tribal schools. The TAP provides an organized pathway to earn a teaching degree. This study reveals the perceptions of the participants after completing their first semester. The results reveal that most experiences are positive, yet they desire more communication and help with time management. The results of this study can be useful to principals and superintendents who may be partnering with stakeholders to begin a TAP