Abstract
The growing integration of artificial intelligence in software development introduces social and ethical concerns. My capstone project focuses on enhancing efficiency through automation, while my STS research examines how AI code assistants disrupt traditionally held notions of software ownership.
In my technical capstone, I addressed the inefficiencies and complexities associated with the manual generation of financial documents within Amazon Nazir’s platform. The existing process made it difficult for new and existing customers to migrate their financial records onto the system. To streamline this process, I designed and implemented an automated workflow system that operates alongside the existing Nazir platform to backfill and generate financial documents automatically from minimal user input. I built the solution using a distributed cloud architecture, using AWS Lambda services to run tasks such as input validation, document creation, and status tracking, and AWS Step Functions to organize these components into a single end-to-end workflow. The resulting prototype enabled users to rapidly generate and backfill estimates, budgets, and capital requests into the Nazir system in minutes rather than through the lengthy manual process, significantly lowering the barrier to adoption and migration.
I used AI code assistants like Amazon Q while developing this software to save time with coding and assist with debugging. While they helped boost my productivity, I also found limitations in their knowledge in certain areas of development. My reliance on AI code assistants for development brought up larger social issues of these technologies: while they can increase productivity, they raise issues of potential software license infringement of generated code.
This motivated my STS research surrounding how AI code assistants are changing the discourse surrounding software ownership. This issue is significant as open-source licenses historically have been used to form ownership of software by using obligations such as attribution and reciprocity when using someone else’s software. Using discourse analysis, I examined these licenses, corporate policy statements from AI companies, and developer discourse in online forums. This analysis is appropriate by going beyond the legal ramifications of AI-generated output to instead ask how AI companies are shifting the notions of ownership in software through the language used. This research revealed that AI companies define ownership as belonging to the prompter of the output, whereas open-source licenses define ownership based on the user’s adherence to the licensing agreement. Developers express concern regarding the legal risks of AI-generated code and also have mixed opinions regarding whether the user should be the model prompter, the creator of the open-source code the model was trained on, or in the public domain.
My capstone and STS projects show the complexities of AI integration in software development. While AI tools can significantly improve efficiency, they create uncertainty about who owns the output generated by them: the model prompter, the model company, or the creators of the open-source code the model was trained on. As AI continues to play a larger role in development, it is critical to address these ethical challenges to ensure sustainable and responsible technological advancement.
