HoosInAction: A Student Volunteering Application; Biased AI: How It Manifests and How to Mitigate It

Though my STS research and technical project are not directly related, the exploration of my STS research could potentially help implement its contents into a future extension of my technical project. My technical project proposed the creation of a user-friendly website serving as a centralized hub for all volunteering opportunities offered by the University of Virginia and its organizations. On the other hand, my STS research explored the manifestation of bias in Artificial Intelligence and potential ways to mitigate it. Therefore, future implementations of my technical project could incorporate the use of Artificial Intelligence, more specifically personalized AI sorting algorithms in the main feed, while also adhering to the methodology discussed in the STS research to decrease the probable appearance of bias.
The University of Virginia does not have a digital space where students can convene and become involved in their community. Instead, students resort to the painstaking task of navigating numerous websites with the promise of social involvement. Therefore, my technical project proposed the creation of a user-friendly website that would hold a centralized database containing all volunteer opportunities offered by the school and its associations. This project was achieved through the development of wireframes to visualize the functionality of the application and through the creation of an ER diagram to demonstrate the layout of the database that this project concept utilized. Desirable outcomes include a less frustrated student demographic and increased volunteer engagement.
In my STS research, I explored how bias manifests in AI as well as proposed potential methods to mitigate it. Three main reasons that lead to this unfair manifestation in AI are as follows: 1) AI training under flawed, undiversified datasets, 2) human error in selecting proper attributes considered by the AI algorithm when making decisions, and 3) the inherent biases that are transported into the AI due lack of diversity among its creators. Given the prominence and future of AI in modern society, it is imperative that engineers rid AI of its bias. In regards to flawed datasets, a possible solution would be randomization, A/B experiments, and the use of data distribution frameworks. In terms of human error during development, training sessions and bringing an extra layer of neutral experts to make implementation decisions could potentially aid in reducing AI bias. Finally, regarding the lack of diversity among developers of AI, the best option would be to diversify the workforce in charge of AI. As an extra preventative measure, it is recommended to validate predictions, or results, returned by an AI with a neutral entity, such as an expert in the subject matter; thus, aiding in detecting any bias before the system is released to the public.
A lot was learned from doing both my technical project and my STS research paper. The contents explored in my technical project exposed me to UX and UI design as well as database design in terms of the necessary preparatory steps before developing an application. Meanwhile, my STS research paper exposed me to the present issue of bias in AI in a world where AI has risen in presence significantly. Moreover, delving into this topic has enabled me to familiarize myself with techniques aimed at diminishing this problem, posing very valuable knowledge. The ethical significance behind my work is the importance of attaining a fairer implementation of AI given how diverse our civilization is. It would be unethical to continue to ignore this problem and allow for further perpetration of biases into the many communities that make up our society.

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisor: Rosanne Vrugtman, Briana Morrison
STS Advisor: Richard Jacques
Technical Team Members: Angie Loayza Bernuy