Abstract
Working at a global investment firm like T. Rowe Price means following very strict rules about who can access what data. Managing Active Directory (AD) groups was difficult because there was not a single automated, centralized interface, but rather a single text file on a physical server that had to be manually updated and verified, which compromised security and efficiency. To address these security needs, I developed myTRPCompliance, a full-stack solution that streamlines Active Directory (AD) group management for compliance tracking by consolidating scattered processes into a single tool. I chose to focus my STS research on Algorithmic Fairness in Hiring, even though the two projects are not clearly related. This research argues that automated hiring systems obscure judgments about merit by embedding historical social biases into "unbiased" algorithms. As a student recently entering the job market, I wanted to understand how institutions use technology as a shield to provide an ethical distance from the choices their algorithms make every day. Ultimately, this demonstrates that STS perspectives are vital to engineering because they challenge us to look beyond the code and recognize that software is never a purely objective tool.
The technical portion of my thesis produced myTRPCompliance, a full-stack solution designed to streamline Active Directory (AD) group management for compliance tracking. I implemented a React frontend with a Express and Node.js backend that utilized PostgreSQL for secure data handling. The implementation successfully made a UI where AD group tasks, such as adding, deleting, and validating memberships, were fully automated. By replacing manual entries with a centralized interface, the project reduced the potential for security vulnerabilities and system downtime. The project achieved high standards of code quality through a test suite refactoring that reduced the codebase by 64% while keeping 100% test coverage. I also optimized backend SQL queries to ignore timestamps for accurate data retrieval and updated the frontend to ensure correct API payloads, preventing common group-creation errors. By moving away from manual verification and building a centralized full-stack tool, I was able to create a system that makes the whole compliance process much more reliable. The biggest takeaway for me was seeing how automation can directly stop access creep and keep the infrastructure secure enough to pass strict regulatory audits. In my STS research I examined how automated hiring systems define and obscure judgments about merit, specifically focusing on the biases embedded in proprietary "black box" models.
I argue that these automated tools are pushed by corporations as neutral forces of progress, but are actually social constructs that encode the biases of the institutions that build them. I applied the theory of technological solutionism to show how developers treat the deeply social problem of employment discrimination as a "simple technical bug" to be optimized, while technological fatalism explains how organizations adopt these tools as if they are powerless to resist an inevitable trend. I then explain how this allows for “ethical distance” for corporations to avoid direct accountability. By evaluating academic replications I discovered that AI "success" is often defined by how well a machine can mimic human subjectivity, effectively scaling individual recruiter preferences into a strict algorithmic standard. The significance of these findings lies in proving that "merit" is not a neutral technical metric but a social construct, and that without radical transparency and third-party auditing, we risk merely automating the inequality of the past.
Considering the technical, organizational, and cultural elements of these projects simultaneously has opened my perspective as a software engineer. The success of my technical project was not just about the quality of my code, but also about balancing strict financial regulations with business needs. This sociotechnical approach reveals that engineering solutions are deeply shaped by the institutional values they serve. My STS research reinforced this by showing that ignoring cultural biases in training data allows technical solutions to scale systemic inequality. Ultimately, this synthesis promotes ethical responsibility by replacing the myth of technological neutrality with the realization that every design choice is a series of human choices. I now understand that a successful engineer must balance technical skills with an awareness of their societal impact. Moving forward, I am committed to a practice where technical performance goes hand in hand with social responsibility, ensuring that the software I build is as ethically sound as it is efficient.
