Software Engineering with Machine Learning: Productizing an NLU-based Model; Ethics of Artificial Intelligence Applications: Algocracy and its Threat to the Rule of Law

With the rapid innovation in Machine Learning (ML) and Artificial Intelligence (AI), while some ML and AI applications have received lots of public attention, some other fields, such as the legal field and the field of governance, while lesser-known, have also been affected by such innovations. The corresponding technical project is an example of such developments in ML application in the legal field. The project is a legal technology project in which a machine learning system application was built for users to find similar published patents using either the patent number of a published patent or the document of an unpublished patent. The STS research portion is inspired by the potential impact of this type of ML application as it seeks to investigate the societal and ethical implications of ML and AI applications in the legal field and the field of governance.
Specifically for the technical project, its goal is to improve the efficiency of patent engineers of the company – the legal professionals who prepare patents to publish – but its users can be anyone ranging from company executives making strategic decisions to regular employees looking to publish their patents. The ML application developed provides a tool to the users such that it could be used to discover similar patents to an input patent, improving user efficiency when processing patent documents, exploring similar patents or making strategic decisions on intellectual properties. Due to the sophisticated nature of this project, I worked with a machine learning scientist who worked on the machine learning model aspect of the project, such as model training and parameters tuning. On the other hand, my job as a software engineer was to productize a given machine learning model into a consumer-facing application for a similar patent search in the form of a web application, prioritizing strong security, user- friendliness, and scalability. For its direct use cases, this tool is able to speed up the average time it takes for someone to discover similar patents so that they can perform revisions or make decisions faster when filing for a new patent. Indirectly, this tool may have strong societal influence as it is a convenient way to discover patents that can lead to more innovations and encourage technological advancements. According to the reports from the internal use of the final application, it was able to significantly improve the efficiency of the company’s legal professionals and patent engineers by reducing their patents search time by over 20 percent. Company executives and legal consultants also benefited as the tool enables them to make faster strategic decisions.
For the STS research portion, I focused on discussing and investigating how the AI applications in governance, or algocracy in short, can impact the rule of law. The background information regarding how crucial and significant it is to uphold and enforce the rule of law in modern societies is provided along with a brief overview of how most AI applications work behind the scenes. The methods of investigation are historical case studies combined with an analysis using the Actor-Network Theory, an STS framework. Specifically, real-world cases where uses of AI applications in governance such as AI software in surveillance, fraud detection, and criminal justice from the United States, the Netherlands, and Russia, are examined and analyzed to understand the ethical implications and their relations to the rule of law. The paper concludes with the argument that such cases of AI applications significantly and evidently undermine multiple principles of rule of law and voices concerns regarding how future algocracy developments could further such erosion.

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisor: Rosanne Vrugtman, Daniel Graham
STS Advisor: Joshua Earle
Technical Team Member: David Xue