CareData: Deriving a Representative Dataset with Benchmarks for Machine Learning in Healthcare; AI Healthcare Politics: An Analysis on How Industry Practices Perpetuate Racism 145 views

In today’s digital age, rapid modernization across industries has focused on optimizing processes and sometimes even removing the human element. Particularly in healthcare, the reliance of professionals on newly developed tools that harness artificial intelligence through algorithms is a decision that could end up causing more harm than good. These tools are built on flawed data and trained based on inequitable problem framings, leading to skewed decision-making that reinforce existing disparities in healthcare. The dissonance between the intended and actual effects of healthcare algorithms demands immediate action in order to address discriminatory practices. Through my technical report, I offer a possible remedy for one possible source of racism in healthcare algorithms. Closely related, my STS research paper analyzes the current healthcare industry’s landscape and how it promotes racially discriminatory practices. By using an STS framework, I propose a new perspective on the issue at hand and how to address it in a systematic manner. In my technical paper, I outline a procedure on creating an unbiased healthcare dataset for future healthcare algorithms to be trained on so that they extract relationships from an accurate representation of the population. In addition, I also suggest the creation of various benchmarks that validate whether outputs are equitable and fair. The dataset targets shortcomings of data commercialization along with modern methods that attempt to ignore or correct holes in the data without addressing the cause. By avoiding artificial corrections that misrepresent the data, the methodology I propose will offer a way to resolve unique healthcare challenges faced by underrepresented groups to promote inclusive healthcare tools. In my STS paper, I examine the closely intertwined relationship between healthcare algorithms and racism. Using the framework of technological politics, I demonstrate how these two concepts will always involve each other due to an intrinsic systematic flaw in our nation. Instead, those that develop healthcare algorithms must seize the powerful political force behind their creations and use it to correct an industry that has historically never provided equal healthcare for all. After an overview of the current state of racism in healthcare along with an examination of the array of reasons for skewed healthcare algorithms, I offer my perspective of how the industry needs to progress and change in order to produce a resounding difference that rids racial prejudice from healthcare.

BS (Bachelor of Science)

healthcare; artificial intelligence; dataset; benchmark; technical politics; racial bias

School of Engineering and Applied Science Bachelor of Science in Computer Science Technical Advisor: Briana Morrison STS Advisor: Travis Elliott Technical Team Members: Albert Huang

English

All rights reserved (no additional license for public reuse)

2025-05-07