Automating Contract Reports with Power Apps; Examining Circumstances Around Discriminatory Machine Learning Hiring Models and Why Companies Have Permitted Their Use

My technical capstone and STS research both examined the implementation and use of automated software that was built to increase productivity and save time. In the capstone project, I documented my experience working on my summer internship at SimVentions. I was assigned the task of coding, developing, and testing an automated system that would eventually be used by other employees in the future. In my STS research, I focused on examining the biases that automated machine learning hiring models can produce. In the case that these hiring models develop bias as a result of improper training and validation, these systems have the potential to discriminate against marginalized groups and directly affect their livelihoods. During the summer, I interned at SimVentions, an employee-owned US defense contractor located in the Fredericksburg area. My assigned primary task was to finish the development of a Contract Reporting Tool App which would expedite the process of entering certain information on document forms. This application was developed to solve the issue of employees repeatedly having to enter commonly shared data by hand, which could be automated to save time and resources. I was initially given a pre-existing prototype created by my supervisor who did not have enough free time to finish the project on his own. Accompanying the prototype was a list of set requirements expected for the initial release of the application. After completing the development of the application, I worked with a small project team to beta-test the system and facilitate the deployment of the application. Through rigorous testing of edge cases and uncommonly encountered scenarios, I was able to iron out any existing bugs and unintended errors in the system so that I could proceed to the last step of development. My final task was to advance the application as either a Power Platform Solution or a Model-Driven App based on additional research and briefing from the team and chose the former. In my STS research project, I examined the use of machine learning models in the hiring process of companies and the biased results such systems have the potential to produce using the ethical framework of machine ethics to guide my research. Machine learning is a subset of AI that teaches applications using pre-existing data, and the framework of machine ethics attempts to ensure that such systems remain moral in their decision making. As the practice has become more common in recent times, I sought to answer my proposed research question: why have companies in the US chosen to permit the use of biased and unverified machine learning hiring systems as part of the employment process? To accomplish this, I first conducted a literature review of the types of machine learning biases these models can produce, past scenarios in which machine learning hiring models discriminated, and how far the US legal system has gone in monitoring these technologies. After this, I documented the methodology by which I conducted my literature review and gave my reasoning for researching the points of interest discussed earlier. In my analysis, I conducted a case study regarding the lawsuit against iTutorGroup, which involved their automated candidate program discriminating against applicants over a certain age range. Throughout my analysis, I argued that companies have been using biased systems due to a lack of technical and ethical knowledge from higher-ups and because there is a lack of legal and financial reasons for companies to ensure their models are unbiased and validated. Finally, I suggested potential future work for programmers, employees, and lawmakers to ensure these machine learning hiring models remain fair and equitable to all. Working on these projects simultaneously has allowed me to better understand the reasoning for using automated systems in the workplace from both an inner and outer perspective. In my technical capstone, I could see first-hand the benefits of the automated system I was developing and how it could be used to increase company efficiency. In my STS research, I examined the alternative perspective of an executive pushing for these automated systems and the economic and legal reasons for their actions. Additionally, the lack of transparency in many machine learning hiring models affected how I developed my assigned automated application. In case future employees were curious about how the code functioned, I made sure to document my code well with comments explaining functions and my reasonings for choosing certain implementations. I can say that both of these projects were worthwhile experiences that I hope will continue to influence how I develop automated systems in my future work.