Abstract
Artificial Intelligence (AI) is an extensively capable tool that can help or hinder individuals depending on how it is used. It has become prevalent within the tools regularly utilized at work and in life. AI can be beneficial in education, as seen in my technical capstone project. One feature of the project is a course chatbot that can answer student questions about course concepts and logistics. If enabled, the Retrieval-Augmented Generation (RAG) feature allows the chat to also leverage information from uploaded course content and cite the files where the information was retrieved. This helps students get a detailed explanation on topics and the ability to have their follow-up questions answered and confusion cleared. It is also advantageous in assisting students in sorting through the information in course files. On the flip side, generative AI presents as a friendly, affirming resource; however, when used for companionship or casual conversation, it can have largely negative impacts on individuals. These themes are explored throughout this Undergraduate Thesis Portfolio.
The AI-Smart Classroom Initiative (ASCI) is a system developed by University of Virginia professors Nada Basit, Mark Floryan, and John “Robbie” Hott, to foster better understanding between students attending office hours and course instructors and teaching assistants. For my capstone project, I am working on a team to further the system to help make it easier to monitor student progress within courses, improve the existing course chatbot, and promote course engagement. Besides advancing existing features, user interface changes were implemented to improve user satisfaction and new features were launched onto the system, such as Discord Activity tracking and course archival. The main tools available in the system are an office hours queue, a course-specific local large-language model (LLM), administrative tools for monitoring learning status, and extra-credit quest opportunities for students. The system synchronizes data from Piazza and Gradescope for up-to-date information about course logistics and student status. Course content can be uploaded to the system as well for the chatbot to utilize.
My STS Research Paper focuses on the disadvantages of AI use for companionship given the rise of applications for AI companionship and neural impacts from extended generative AI usage. OpenAI’s ChatGPT is the AI tool focused on in the paper. The usage of AI for casual conversation by individuals is troubling because of the negative psychosocial impact, reported statistics on allegedly mentally unstable conversations with ChatGPT each week, and cognitive offloading that occurs from AI use. Through the connection of known knowledge on different brainwaves’ functionality and the neural impacts from relying primarily on AI for tasks that were identified through brain imaging, the effect AI has on emotional regulation and responses is realized. The accessibility of generative AI tools versus the accessibility and use of mental health resources is discussed to consider how technological disparity factors into the negative effects of generative AI use.
Working on these projects simultaneously highlighted the importance of implementing proper safeguards for AI. Such powerful tools can easily be taken advantage of, so ensuring that they are only capable of performing their intended tasks is important. The ASCI system has lots of restrictions in place to ensure that only course related questions will be answered, and the chatbot is not able to simply provide code or answers to class assignments. OpenAI as a company updated their policies in August 2025 to reduce sycophancy and discuss procedures for handling mentally concerning conversations between users and ChatGPT. I was really reminded that although the invention and increased accessibility of new tools and technologies can be exciting, proper safety measures must be taken beforehand and the possible negative long-term effects need to be considered.
