Master of Science, Miami University, 2024, Computer Science and Software Engineering

Drawing inspiration from lateral thinking, synergetics, psychology, creativity, and business, this research project employs an interdisciplinary approach to investigate the research process which drives innovation in the field of artificial intelligence. This research project explores methods for harnessing the synergy present in the latest, neuro-symbolic paradigm of artificial intelligence, while noting similarities between the first two waves of AI and dual process theory. It attempts to integrate unconventional, yet potentially promising interdisciplinary ideas into a proof of concept, including creative tools and techniques like the Six Thinking Hats, methods of psychotherapy, including cognitive behavioral therapy and internal family systems, as well as principles related to conflict resolution and ``tensegrity". The proof of concept is a hybrid semantic search system for research papers in computer science, constructed using a process of rapid prototyping and iteration, with special consideration for evaluating how more modular, interpretable, and human-centric approaches to system design can help narrow the gap between cutting-edge AI research and ethical, practical application in business. This research is conducted with the hope of opening the research field to greater creative possibility, as well as deliberate action towards creating more sustainable and human-centric artificial intelligence systems.

Master of Science, Miami University, 2024, Computer Science

Causal maps are graphical models that are well-understood in small scales. When created through a participatory modeling process, they become a strong asset in decision making. Furthermore, those who participate in the modeling process may seek to understand the problem from various perspectives. However, as causal maps increase in size, the information they contain becomes clouded, which results in the map being unusable. In this thesis, we transform causal maps into various mediums to improve the usability and accessibility of large causal models; our proposed algorithms can also be applied to small-scale causal maps. In particular, we transform causal maps into meaningful paragraphs using GPT and network traversal algorithms to attain full-coverage of the map. Then, we compare automatic text summarization models with graph reduction algorithms to reduce the amount of text to a more approachable size. Finally, we combine our algorithms into a visual analytics environment to provide details-on-demand for the user by displaying the summarized text, and interacting with summaries to display the detailed text, causal map, and even generate images in an appropriate manner. We hope this research provides more tools for decision-makers and allows modelers to give back to participants the final result of their work.

PhD, University of Cincinnati, 2022, Engineering and Applied Science: Aerospace Engineering

This research falls in the field of obtaining new high-performing simple architectures that demystify explainable AI. We narrowed down the scope of work focusing only on regression tasks. The ultimate goal is to provide an algorithm that not only makes the internal functioning of its logic transparent to human understanding, but also produces better results than the current state of the art. This type of technologies will play a fundamental role in all engineering during the next century, especially in those areas where the human supervision and quality assurance is unquestionable. In this framework of XAI, fuzzy logic is leading the development of algorithms that meet these transparency requirements. The linguistic and interpretable nature of fuzzy inference systems, together with their ability for supervision and direct integration of expert-knowledge, make them the perfect candidates for certifiable and trustworthy AI. Therefore, in the architectural design of the algorithms covered in this research, we will leverage the tools that fuzzy logic has to offer.

Doctor of Philosophy (PhD), Ohio University, 2020, Educational Research and Evaluation (Education)

The purpose of this non-experimental study was to examine the concerns of preservice teachers in reference to the Amazon Echo, powered by Alexa, as measured by the Stages of Concern Questionnaire (SoCQ) (George, Hall, & Stiegelbauer, 2006). The study participants were preservice teachers at a Midwestern University enrolled in a technology course (n = 124). This researcher utilized a pair of two-way multivariate analysis of variance MANOVA which were conducted to determine if there was a statistically significant difference in means between the dependent variables of the preservice teachers' Stages of Concern (stages 0-6), and the independent variables of gender, projected grade to teach, and teaching geographical area. Results from the SoC Questionnaire are as follows: Preservice teachers would like to receive more information regarding the Amazon Alexa as a tool for teaching in formal and informal settings. The two-way MANOVA models did not reveal conclusive results. Contributing factors for these results may be low statistical power and small effect size which may have affected the overall models results. Implications for preservice teachers' training, suggestions for further research, and the limitations of this study are discussed.

Bachelor of Fine Arts (BFA), Ohio University, 2019, Theater

Inspired by continuing developments in artificial intelligence, this creative thesis comprises an overview of conversational artificial intelligence and an exploration of how education and entertainment overlap in the broad genre of science plays. I argue that the combination of education and entertainment can be mutually beneficial, and that this relationship should be exploited to increase public accessibility to information. Moreover, this thesis documents the process of writing, producing, directing, and designing a new work entitled Time to Think: A Short Play about Ignorance & Bliss. The original script, including marginal director's notes, is appended.

Master of Computing and Information Systems, Youngstown State University, 2012, Department of Computer Science and Information Systems

Gaming artificial intelligence must look intelligent, real intelligence is a near impossible goal to achieve because of the number of CPU cycles it requires. Using Multiple Need Smart Terrain, AI can look intelligent without requiring the large number of CPU cycles it requires for such intelligence. It can be used to manage a character's needs and direct them to which objectives are most profitable for them. Unlike normal smart terrain that only pays attention to one need at a time, this algorithm can look at any number of needs and check which objective would be best to meet that need. This way, the AI can have little actual intelligence, the terrain tells it where to go. With this technology, AI can appear intelligent and keep the cycles required to a minimum.

Master of Arts in English, Youngstown State University, 2024, Department of Languages

The rise of Generative Artificial Intelligence (GAI) technologies accessible to everyday users has attracted significant attention, including in education. For instance, ChatGPT attracted over 1 million users in less than a week after its release, marking one of the fastest-growing forms of AI. These technologies have the potential to transform the products and processes of writing, especially those of L2 writers who face challenges with composing. However, scholars and instructors have raised concerns about the potential ethical issues surrounding their use, especially in cases of accusations of cheating or plagiarism. At the same time, less is known about the perspectives of students, including international and L2 students, who have the most to lose in instances of accusations of lack of academic integrity or plagiarism. To respond to this gap, my study uses sociocultural theory to examine multilingual university students' perceptions of ChatGPT as a scaffold for writing academic essays. Participants were 11 international students enrolled in a developmental composition course for undergraduate L2 writers at a mid-sized U.S. university in the Midwest. Data collection included a classroom intervention utilizing ChatGPT, a pre-intervention questionnaire, a post-intervention questionnaire, and semi-structured interviews. Overall, findings include that these students' perceptions are divided, and individual students may be torn about how useful ChatGPT is. Specifically, in various areas, students rated ChatGPT as less helpful than what other scholars have found (e.g., word-, sentence- and some discourse-level scaffolds; Sumakul, 2023). Additionally, students' perceptions of using ChatGPT as a scaffold were not as positive as getting feedback on their writing from a peer. However, students also became less concerned about the accuracy and trustworthiness of ChatGPT after being exposed to it. Theoretical and pedagogical implications are discussed.

Master of Science in Computer Science, Miami University, 2024, Computer Science and Software Engineering

Adversarial training (AT) is a defense technique used to increase the robustness of neural networks. AT generates adversarial examples that maximize the loss to the model and then adjusts model parameters to minimize that loss. Previous AT methods typically use only a single attack to perturb adversarial examples that maximize loss, and ignore the roles that different image-classes play in determining final robustness. These techniques are thus unable to properly explore the perturbation space and cannot target specific weaknesses of the training data. As a result, they train models with diminished robustness. This thesis proposes class-based adversarial training, which increases the robustness of AT by using a variety of attacks that target the weakest image-classes of the dataset. We designed and implemented two novel algorithms within this category: the Various Attacks (VA) technique and the Advanced Adversarial Distributional Training (ADT++) technique. Using a novel testing framework created to better examine model robustness across a variety of metrics, we conducted a series of experiments on two benchmark datasets. The results demonstrate the superiority of the VA and ADT++ frameworks over state-of-the-art adversarial training methods.

MARCH, University of Cincinnati, 2024, Design, Architecture, Art and Planning: Architecture

Automation is a term loosely thrown around within contemporary society, spoken to either spark hope for a better tomorrow or sow seeds of fear. For as long as mankind had the stability to stop and think, they have asked a simple question. Why? Is there any intrinsic purpose that might befit existence? The question has been asked thousands of different ways, and answered countless more. Our purpose is to seek leisure, to seek pleasure. To serve the common good, to praise god. Perhaps there is no purpose at all, and to act on carnal impulses would have no noticeable effect on the cosmic meat-grinder of our universe. The question has often been rhetorical, regardless of the answer people have for the most part continued the trudge forward out of implicit or explicit necessity. But for the first time in history, a plausible future where any personal exertion is superfluous has revealed itself. The recent exponential rise in artificial intelligence exposes just the tip of the iceberg in regards to what could be possible in the not-so-distant future. Impossible structures built by scores of tireless metal men, entire cities sprung from the ground without a single human raising their finger. The bounding box of science fiction is capsizing to expose an underbelly of truth. This paper, through notional conversations with architects, will follow in the tradition of iterating upon the oldest of questions. If man invented a machine that could do anything, would man have any reason to persist?

MPH, University of Cincinnati, 2024, Medicine: Epidemiology

BACKGROUND: Every year millions of people in the United States seek dental care in emergency departments (ED). The annual cost of these visits has risen over the past decades and now accounts for over a billion dollars of healthcare spending. ED treatment often consists of antibiotic therapy supplemented with analgesic medication without any definitive dental treatment. Extreme Learning Machines (ELMs) are a type of artificial neural network which has shown tremendous ability to create classification models to solve various problems in medicine. This study aims to assess if an ELM could accurately predict the necessary treatment regimen for a patient presenting to the ED for a non-traumatic dental condition (NTDC). METHODS: This was a retrospective cohort study was utilizing electronic health records (EHR) data from UC Health EDs in Cincinnati, Ohio. To be included subjects must have presented to a UC Health ED between October 2nd, 2015, and January 2nd, 2024, with a primary diagnosis code of K02.9, K04.4, K04.7, or K08.9, and must have been at least 18 years of age at the time of their visit. The primary predictor variables in the ELM model were sex, age, if the afflicted area was swollen, if the afflicted area was reddened, the stage of the symptoms, air involvement, systemic signs of infection, elevated blood pressure or heart rate, and drug allergies. The primary outcome variable was the discharge plan for the patient, classified as either simple or complicated management. ELMs were then created using sine, triangular basis, tan-sigmoid, sigmoid, and hard-limit activations functions. All activation functions were tested with 50, 75, 100, 125, 150, and 175 neurons to test their accuracy. RESULTS: The final study sample included 501 subjects (52% male and 48% female). The mean age of the sample was 40.29 ± 14.35 years of age. 57% (287) of the subjects required simple management of their symptoms with the other 43% (214) requiring complicated management of (open full item for complete abstract)

PhD, University of Cincinnati, 2024, Engineering and Applied Science: Computer Science and Engineering

This work explores the explainability of features used for classification of malicious binaries in machine learning systems based on semantic representations of data dependency graphs. This work demonstrates that explainable features can be used with comparable classification accuracy in real-time through non-parametric learning. This work defines operational semantics in terms of data dependency isomorphism, and quantifies the network structure of the graphs present in static features of binaries. This work shows that a bottom-up analysis holds across levels in the architectural hierarchy, and can be performed across system architectures. This work shows that semantic representations can be used for search and retrieval of malicious binaries based on their behavior. This work shows that unknown vulnerabilities can be predicted through descriptions of structure and semantics.

Master of Science, Miami University, 2024, Electrical and Computer Engineering

In wireless communications, traditional base stations provide network connectivity to users. Static base stations, however, require significant time and money to construct and are therefore not suitable for remote areas and disaster scenarios. An alternative uses mobile base stations attached to UAVs. Such UAV-based communication networks can be rapidly deployed and adapt to their environment. The goal of this research is to position the UAVs to maximize user coverage. One approach treats UAVs as independent agents and uses multi-agent reinforcement learning to design policies that move the UAVs to positions that increase coverage; each UAV, however, must train its own policy and optimality is not guaranteed. Instead, we consider two centralized approaches to place the UAVs. The first uses centralized reinforcement learning to design a joint policy over all UAVs, but training the policy is not computationally tractable for large problems. The second approach uses mixed-integer optimization to find the UAV positions that maximize user coverage. While this yields the optimal solution, the computational time does not scale well with the problem size. Therefore, we first group users into clusters and then optimize UAV positions with respect to the clusters. The number of clusters trades off computational time with optimality.

Master of Computer Science, Miami University, 2024, Computer Science and Software Engineering

In the field of Artificial Intelligence, the usage of intelligent agents has increased significantly. The term intelligent agents refers to a computer program that doesn't need any human interference or help to make decisions working towards a goal. While the external controller of the agent may impose goals, it is up to the agent to come up with a plan to achieve them. This thesis introduces a framework using Answer Set Programming (ASP) and the policy specification language AOPL to enforce policy compliance in autonomous agents through penalization while considering goal prioritization as well. A systematic penalty system assesses and addresses violations by severity, enhancing safety and reliability. Validated across scenarios including self-driving agents and service robots, the framework demonstrates adaptability and effectiveness. This work advances autonomous agent governance, promoting safer AI integration in everyday applications and setting the stage for further research on scalable penalty mechanisms.

Doctor of Philosophy (Ph.D.), Bowling Green State University, 2024, Media and Communication

While scholars from a wide range of disciplines have analyzed thematic development, iconography, narrative, characterization, and animation style of Japanese anime, the music of anime programs is largely ignored or trivialized. This dissertation fills the gap in critical intercultural communication and media studies research by examining original anime soundtracks and their roles as narrative devices. Anime is explored as a site of global cultural resistance, while maintaining articulations of gender and cultural ideals in their stories and reflected in the lyrics of their theme songs. Employing critical intercultural communication, critical media studies, Affect Theory, with textual analysis and rhetorical criticism, this dissertation analyzes how music is intrinsic to the narrative and an expression of cultural values in anime. Analysis focuses on Hibike! Euphonium (2015-present) by Tatsuya Ishihara and Naoko Yamada, from the studio of Kyoto Animation, a slice-of-life drama involving the coming-of-age stories of high schoolers in a competitive concert band, and Vivi -Furoraito Aizu Songu- (2021) by Tappei Nagatsuki and Eiji Umehara, produced by Wit Studio, which follows an autonomous Artificial Intelligence (AI) programmed to entertain humans with her voice, and who discovers her humanity through music while trying to save the world from destruction. Each anime illustrates how musical scores, lyrics, and instrumentation are incorporated into narratives of gender, agency, culture, and humanity. The dissertation also analyzes compositional style, structure, instrumentation, and lyrics encoded with hegemonic messages and constructions of gendered, raced, and cultural distinctions. It provides a critical analysis of how music is used as a narrative tool in media and communication studies involving anime and how the rhetorical messages encoded in texts, via lyrics and instrumentation, are forms of intercultural communication of Japanese anime viewed by a Western aud (open full item for complete abstract)

Doctor of Philosophy (Ph.D.), Bowling Green State University, 2024, Statistics

Deep learning, neural network, has been penetrating into almost every corner of data analyses. With advantages on computing power and speed, adding more layers in a neural network analysis becomes a common practice to improve the prediction accuracy. However, over depleting information in the training dataset may consequently carry data noises into the learning process of neural network and result in over-fitting errors. Neural Network has been used to predict the future time series data. It had been claimed by several authors that the Neural Network (Recurrent Neural Network) can predict the time series data, although time series models have also been used to predict the future time series data. This dissertation is thus motivated to investigate the prediction performances of neural networks versus the conventional inference method of time series analysis. After introducing basic concepts and theoretical background of neural network and time series prediction in Chapter 1, Chapter 2 analyzes fundamental structure of time series, along with estimation, hypothesis testing, and prediction methods. Chapter 3 discusses details of computing algorithms and procedures in neural network with theoretical adjustment for time series prediction. In conjunction with terminologies and methodologies in the previous chapters, Chapter 4 directly compares the prediction results of neural networks and conventional time series for the squared error function. In terms of methodology assessment, the evaluation criterion plays a critical role. The performance of the existing neural network models for time series predictions has been observed. It has been experimentally observed that the time series predictions by time series models are better compared to the neural network models both computationally and theoretically. The conditions for the better performances of the Time Series Models over the Neural Network Models have been discovered. Theorems have also been pro (open full item for complete abstract)

Doctor of Philosophy, Case Western Reserve University, 2024, Accounting

Audit firms are dedicating significant amounts of time and resources to develop artificial intelligence (AI) systems that will assist auditors with a wide range of tasks and are expected to improve audit quality. However, existing research demonstrates that individuals tend to exhibit bias in favor of human generated outputs compared to algorithm generated outputs, a phenomenon termed as algorithm aversion. Prior research also demonstrates, to a lesser extent, the existence of algorithm appreciation - people relying more on algorithmic advice compared to human advice. The occurrence of these two very related, yet divergent theories is highly relevant to the audit setting. As auditors begin to interact with AI systems on a regular basis, unsubstantiated aversion or appreciation towards algorithmic advice will be not only be a waste of resources, but also decrease audit quality. Theory suggests that additional information about an advisor may help advice takers assess the advisor and the quality of the provided advice. I conduct an experiment to examine whether the provision of input information and type of advisor affects user reliance on advice. Results indicate that the provision of input information and advisor type do not significantly affect user reliance on advice. Results also indicate that perceived advisor expert power and advisor trustworthiness mediate the relationship between advisor type and reliance on advice.

Master of Computing and Information Systems, Youngstown State University, 2024, Department of Computer Science and Information Systems

The field of coding education is rapidly evolving, with emerging technologies playing a pivotal role in transforming traditional learning methodologies. This thesis introduces Dr. Lego, an innovative framework designed to revolutionize the assessment and understanding of block-based coding through the integration of sophisticated deep learning models. Dr. Lego combines cutting-edge technologies such as MobileNetV3 (Howard, 2019), for visual recognition and BERT (Devlin et al., 2018), and XLNet (Yang et al., 2019) for natural language processing to offer a comprehensive approach to evaluating coding proficiency. The research methodology involves the meticulous curation of a diverse dataset comprising projects from the LEGO SPIKE app (LEGO Education, 2022), ensuring that the models are subjected to a broad range of coding scenarios. Leveraging the dynamic educational environment provided by the LEGO SPIKE app (LEGO Education, 2022), Dr. Lego empowers users to design and implement various coding projects, fostering hands-on learning experiences. This thesis delves into methodologies aimed at enhancing coding education by exploring model integration, data generation, and fine-tuning of pre-trained models. Dr. Lego not only evaluates coding proficiency but also provides cohesive and insightful feedback, enhancing the learning experience for users. The adaptability of the framework highlights its potential to shape the future of coding education, paving the way for a new era of interactive and engaging learning experiences.

Master of Arts (M.A.), University of Dayton, 2024, Communication

Drawing upon organization-public-relationship (OPR) theory as a framework, this study examined how the public's perception of AI-driven chatbot characteristics impacts their perceived relationship with the organization. Specifically, this study examined whether the chatbots' social presence, conversational tone, and interaction quality were related to OPRs regarding trust, satisfaction, and commitment. A quantitative survey via Qualtrics was conducted at a medium-sized private university in Southwest Ohio. The study revealed that participants' perceptions of the AI chatbot's social presence and conversation tone had a moderate, significant, and positive correlation with perceived trust, satisfaction, and commitment to the organization. However, the perceived interaction quality of chatbots had a weak but significant positive correlation with participants' perceived trust, satisfaction, and commitment to the organization. The result also indicates that while social presence and conversational tone influence customer perception, the quality of a chatbot's interaction (solving problems and providing individual attention) is a parallel key driver of trust, satisfaction, and commitment. The study discussed the implications of the findings and suggested the potential for future research.

Master of Science in Computer Engineering, University of Dayton, 2024, Electrical and Computer Engineering

This study aims to incorporate some advantages of computed tomographic data into the chest X-ray deep lung segmentation paradigm. We do this by training a deep convolutional neural network on chest radiographs (a.k.a. X-rays) with manually drawn ground truth and an identical network on radiographs digitally reconstructed from computed tomographic data with ground truth generated for the given computed tomographic image using an automated morphological 3D lung segmentation algorithm. The resulting twin-network ensemble generates pairs of lung image segmentation labels for chest X-rays: 1) a “traditional” segmentation of the lungs encompassing the apparently low-density tissue and 2) a novel, “full” lung segmentation encompassing an expanded view of the lungs' position in a chest X-ray including those regions obscured by the heart, ribs, and viscera, in essence, a 2D projection of any portion of the 3D lung. These networks perform consistently, with mean Intersection-Over-Union scores of > 90% and > 95%, respectively, across five trials. By subjective analysis, the proposed lung segmentation approach shows satisfactory ability to generalize onto genuine check X-ray images. The proposed technique's high performance and robustness establish a precedent for applying computed tomographic data to automatic chest X-ray segmentation and present an opportunity to further refine existing computer-aided detection and diagnostic tools by considering the full lung.

Doctor of Philosophy, The Ohio State University, 2024, Computer Science and Engineering

Autonomous agents are used across multiple tasks and domains to improve processes. These agents are often trained using extensive amounts of data or examples. However, expecting autonomous agents to have perfect knowledge of the environment or specific tasks is unreasonable. Therefore, developing capabilities within agents to identify gaps in their knowledge and resolve discovered gaps is one way to increase agent flexibility during training and execution. To this extent, we propose three crucial steps to eliminate knowledge gaps (KGs) in autonomous agents: detection, identification, and resolution. We refer to these three steps as KGDIR processes. To study and understand how to formalize and develop KGDIR processes, we turn to the tasks in the Artificial Intelligence (AI) community. Using these tasks, we study how to: Detect the existence of a knowledge gap that prevents the successful completion of the task. Identify which type of knowledge gap the agent is experiencing. Resolve the knowledge gap by selecting and executing the best template strategy to complete the task. Implementing these KGDIR processes will prevent agents from incorrectly completing tasks and enhance their robustness. This, in turn, will significantly increase the reliability and usability of autonomous agents, offering a promising solution to the challenge of knowledge gaps.