Abstract
This Undergraduate Thesis Portfolio brings together two projects that examine the growing role of algorithms and artificial intelligence in modern society. My technical Capstone project, Artificial Intelligence: Creating an Agent to Automate Research, focuses on building an AI agent that can complete a practical research task for a startup by finding company leadership information, scraping web pages, extracting relevant data, and organizing the results into a usable format. My STS research paper, Code and Conflict: The Role of Algorithms in Political Polarization, examines how social media algorithms shape political information environments and contribute to polarization by influencing what users see and how they engage with content.
Although these projects focus on different contexts, they are connected by a shared interest in how algorithmic systems make decisions, organize information, and affect real-world outcomes. In my Capstone project, I studied algorithms from the perspective of a builder: I designed a system that used generative AI, search tools, scraping tools, and prompts to complete a useful business task. In my STS research paper, I studied algorithms from the perspective of society: I analyzed how recommendation systems and engagement-based ranking models can influence public discourse, political behavior, and democratic culture. Together, these projects show both the promise and responsibility of algorithmic technologies. They demonstrate that AI systems can make work faster and more efficient, but also that their design choices must be understood within broader social, ethical, and institutional contexts.
Capstone Project Summary
My Capstone project focused on creating an AI agent to automate research for a San Francisco-based AI startup specializing in job recruiting. The company needed a more accurate and scalable way to gather board of directors and executive leadership information for thousands of companies worldwide. Existing centralized sources often contained missing or inaccurate information, so the goal of my project was to build an agent that could search the web, identify reliable company websites, extract leadership information, and organize the results into a CSV file.
To build this system, I used LangChain to create an AI agent with multiple tools, including a Google Search API, a web scraper, a headless browser API, and carefully written system prompts. The agent first searched for a company’s official leadership or board of directors page, then used a headless browser to scrape the page content, and finally used the OpenAI API to extract the relevant names and return them in a consistent format. I also created a small local web application where users could input a company name or list of company names, run the agent, and view the results. The agent achieved a high success rate after fine-tuning, with an approximate 95% success after testing on Fortune 500 companies.
STS Research Paper Summary
My STS research paper examines the question: How do social media algorithms contribute to political polarization? The paper argues that social media algorithms are not neutral tools, but sociotechnical systems shaped by the priorities of engineers, corporations, users, and regulators. Using the Social Construction of Technology framework, the paper analyzes how different stakeholders define and influence algorithmic design, especially around the concept of “engagement.”
The paper finds that social media algorithms contribute to polarization by shaping what users see, how often they see it, and how they interact with it. Across platforms such as Twitter/X, YouTube, Facebook, and Instagram, algorithmic systems can amplify partisan content, reinforce existing beliefs, create feedback loops, and intensify emotional divisions between political groups. However, the paper also recognizes that algorithms are not the only cause of polarization. User behavior, pre-existing preferences, follow networks, and broader media environments also shape political information exposure. This makes polarization a sociotechnical problem rather than a purely technical one.
Concluding Reflection
Working on both the Capstone project and STS research paper at the same time helped me understand algorithms from two different but connected perspectives. Through my Capstone project, I learned how powerful AI agents can be when they are given the right tools, prompts, and structure. I saw how an algorithmic system could save time, automate repetitive research, and create real business value. At the same time, my STS research paper pushed me to think beyond whether an algorithm works and consider what effects algorithmic systems have once they are deployed into society.
The most important insight I gained from completing both projects is that technical success is not the same as responsible design. In my Capstone project, accuracy, reliability, and hallucination reduction were central concerns because the agent’s output needed to be trusted by users. In my STS research paper, similar concerns appeared at a much larger scale: social media algorithms also organize information for users, but their outputs can shape political attitudes, emotional division, and public discourse. This connection helped me see that even when algorithms are built for efficiency or engagement, they still carry social consequences.
Together, these projects strengthened both my technical and critical thinking skills. The Capstone project helped me become more comfortable building with new AI technologies, integrating APIs, designing workflows, and testing outputs. The STS research paper helped me evaluate algorithms as part of broader systems shaped by incentives, stakeholders, and social values. If I had worked on only one of these projects, I might have understood either the technical side or the societal side of algorithms, but not both together. Completing both showed me that strong engineering requires more than building systems that function correctly; it also requires understanding the people, institutions, and communities affected by those systems. This portfolio reflects that combined lesson: algorithms can be powerful tools, but their value depends on how carefully and responsibly they are designed, deployed, and evaluated.