Data Visualization: How Live Visuals Can Help Predict and Track System Failures; My Body My Data: A Case Study of South Korea’s MyHealthWay and Its Transformation of the Health Information Exchange System to Reduce Data Fragmentation

The Ethical Responsibilities Behind Collecting and Processing Data The United States processes 53.4 trillion megabytes of data every year and has more than 10 times the data centers as any other country around the world. The data processed differs across a wide range of industries, but ultimately cements that actors within the US are responsible for both a significant portion of data existing in the world, and that there is a high level of data availability. My technical research explored how data visualizations may be leveraged to predict and track failures to maintain consistent performance in systems that transmit and process global data. Sociotechnically, my research delved into why the aggregation of large datasets was so difficult in the first place. In the United States regulations like HIPAA, otherwise called the Health Insurance Portability and Accountability Act of 1996, blocks medical data sharing. However, South Korea release of its health information exchange (HIE)-unifying platform provides a robust example of how transforming HIE systems is conducive to more successful exchange of medical data to aggregate viable datasets. My technical report discusses how I developed a suite of interactive dashboards using Kibana and the Elastic Stack to monitor and predict failures within a messaging module as part of a summer internship. These dashboards tracked real-time metrics, including transactions per second, latency, error rates, and message retry attempts. I used regex-based bash scripts to process logs and extract data for analysis within these dashboards. The dashboards visualized trends such as rising error rates or unusual latencies, offering predictive insights into potential system failures. As I returned to intern with the same company, I was fortunate to see my work being implemented and validate the ability of live visualizations to predict errors and monitor system-wide performance. In my STS research, I use South Korea’s release of a unified health information exchange (HIE) platform to better understand how limited medical data sharing can be assuaged in the United States. For complex medical AI tools, models need more data to perform reliably. However, legal barriers like HIPAA foment data fragmentation in the US’s HIE system as actors are motivated to prioritize protecting their data rather than investing in sharing with other actors. The three factors that allowed South Korea to succeed was its government’s position as a regime outsider, the influence of landscape developments like COVID-19, and the government’s careful comparison of its platform to a highway. Using Frank Geels’s framework of the Multilevel Perspective (MLP), South Korea’s government functioned as a regime outsider to the HIE system, meaning they were able to use their vast regulatory power to pressure private healthcare actors to invest in system transformations reducing data fragmentation. Furthermore, South Korea’s quick and aggressive response to COVID-19 created a precedence for the government to seize more control of private health information in sacrifice for the greater good, so actors were less likely to resist the government’s control of the HIE-unifying platform. Lastly, the government described its platform as a “highway” that emphasizes its efficiency and implies that sensitive data will not be stored, it will only transmit much like cars travel on a highway without stopping. Using South Korea as a case study provides guidance on how systems failing to uphold their ethical responsibilities may transform and improve. While my technical report and STS research were done separately, the works collaborated to make a more enriching product. Developing the dashboards in my internship gave me a greater appreciation for both the importance and complexity required to process expansive amounts of data. My technical experience in conjunction with the meteoric rise of AI tools, fostered an interest in exploring how actors in the sociotechnical system influence data collection and its subsequent processing. In Ethics and Engineering, Martin and Shinzinger substantially draw upon duty ethicist Immanuel Kant, a famed ethicist who argued that duties are intrinsic to ethical action. At my internship, I helped fulfill the company’s ethical responsibilities to maintain its reliability to provide crucial financial services to their customers. By developing a dashboard capable of visualizing errors and system latency, I contributed to maintaining my employer’s ethical responsibility. Similarly, medical AI tools are directly dependent on the data aggregated through HIE systems as a means of training models to respond equitably. Therefore, the HIE system has an ethical responsibility to collect data in a manner that fairly represents all socioeconomic groups in the United States. By exploring the deeper complexities surrounding data collection and processing, my project highlights how engineering organizations have succeeded in championing ethical responsibility, and how sociotechnical failures can be corrected by taking example from more successful analogous systems.