A Framework to Regulate AI-Assisted Lobbying; Impact and Countermeasures of AI-Powered Mis/Disinformation on Society

Originally, as shown in the prospectus, the primary motivation of pursuing the technical capstone and STS thesis was to conduct a technical assessment of AI’s ability to lobby, and to explore impacts/regulations, respectively. However, due to time constraints, the actual technical assessment did not fit the typical capstone deadlines and instead is now an unrelated symposium presentation. Therefore, the technical capstone was changed to the STS topic (impact/regulations of AI lobbying) and the STS thesis paper needed a new topic. When exploring how AI could affect the political landscape, it became clear that AI-powered misinformation/disinformation not only was affecting politics, but society in general. Therefore, I picked that as my STS topic, and now both the capstone and STS thesis share a connection through AI affecting the political landscape, and society in general, by extension.

The capstone project addresses the topic of AI models being utilized in lobbying, which is a concern given that special interest groups could use privatized, advanced models to further their influence in Congress and reinforce underrepresentation of the public’s interest. To resolve this, the project’s proposed design is to identify key issues in AI lobbying and develop a suitable framework solution. To that end, the issues identified include inequality in access to AI tools and development, as well as microlegislation. To counter this, the proposed framework included lobbying disclosure with enforcement, stronger data protection laws, and enforcement of equitable access to AI tools/development. The solution will be evaluated based on it’s adherence to the stated principles at the start the proposed design (using a panel of legal experts), and an analysis of AI lobbying misuse cases after the framework is implemented into law. Although the anticipated result is that the framework will adhere to the principles and is effective, future work 3 includes testing such framework, and, if successful, expanding upon it to counter new lobbying issues.

The STS research paper explored the impact of AI-powered misinformation and disinformation on society, as well as considered potential countermeasures. Using case studies and historical analysis, a list of societal impacts was compiled, specifically in the fields of politics, finance, business, law, and individual safety/well-being. Then, by applying the Actor Network Theory (ANT), in which all living and nonliving entities are considered actors on a network, it was possible to understand how different actors interacted to cause the impacts mentioned. Through this, it was possible to propose field specific countermeasures, which prevented or regulated certain interactions between actors. As a result, the paper accurately summed up the intersection between science, technology, and society in this specific issue, as well as provided a specific framework (ANT) in which effective countermeasures were developed.

Working on these two projects simultaneously was interesting because it challenged my understanding of the relationship between pure technical subjects and STS topics. Had I completed the two projects separately, I might have utilized two separate issues for convenience, but by doing both simultaneously I had to search for the connection between AI lobbying and AI-powered misinformation/disinformation, one of which being the integration of AI in politics. This was necessary because it helped streamline research for both projects, as sources could be reused for either paper or provide a better understanding of the topic for my own understanding. As a result, I was able to create a more cohesive portfolio that demonstrated my research as a 4th year student within the Computer Science department.

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Briana Morrison

STS Advisor: Bryn Seabrook

Technical Team Members: Husnain Choudhry