Proposing an Unsupervised Genetic AI Algorithm to reduce Facebook polarization and increase user acceptance of contradictory beliefs by identifying 10+ parameters (related topics, contents & user location) through analysis of Facebook’s post suggestion algorithm and web layout.
Team Members: Haider Gillani, Lydia Etherington, Nour Elaifia
Supervisor: Mike Cho - Senior Associate at Kakao Ventures
Role: SWE Algorithm Analyst
Company: Kakao Ventures, the most active seed stage VC in South Korea, with over 190 portfolios, managing 3500+ assets, and AUM of $300M (330B KRW)
Location: Seoul, South Korea
In this project, the primary goal was to analyze and improve Facebook's existing post-suggestion algorithm with a specific focus on mitigating the issue of echo chambers and ideological polarization. Working with Senior Associate of Kakao Ventures, Mike Cho, our team delved deeper into the inner workings of Facebook's algorithm, studying over 10 parameters like related topics, geographical location, and content relevancy across more than 100,000 Facebook posts using Python and PHP to understand how content is shared with users.
Leveraging Natural Language Processing (NLP) techniques such as Topic Modeling and Sentiment Analysis, we identified prevailing trends of echo chambers and ideological polarization. These echo chambers were shown to limit users' exposure to diverse viewpoints and foster extremism by reinforcing pre-existing beliefs.
To combat these issues, we present an innovative solution in the form of an unsupervised genetic algorithm. This algorithm incorporates two critical parameters - the Similarity Ratio and Similarity Index - designed to balance user preferences and exposure to new ideas, promoting cross-cultural communication and understanding.
The findings and proposed solutions were presented to an international audience of experts from over 75 countries, initiating a constructive discourse about enhancing social consciousness on social media platforms. This project helped enhance our proficiency in skills data analysis, AI algorithm design, and Natural Language Processing to harness the influence of computer networks in delivering potential solutions for complex socio-technical problems.