SBIR Phase I: This project aims to “predict global elections in real-time through the integration of artificial intelligence, network theory, and big data science. By harnessing the power of advanced machine learning models and analyzing vast amounts of publicly expressed opinions on social media, the team offers accurate forecasts of election outcomes. This approach has the potential to disrupt the conventional polling industry, which faces growing uncertainties and challenges such as declining response rates and inherent biases in sampling.

The research objectives entail tackling critical research and development challenges, including predicting voter turnout, effectively sampling rural areas with limited online coverage, filtering out bots and fake news sources, inferring the preferences of undecided voters, adjusting sample weights on a state-by-state basis, addressing the opinions of individuals not active on social media, and mitigating social desirability bias (where respondents conceal their intention to vote for controversial candidates).

The anticipated technical results involve the development of a transformative machine learning architecture built upon Graph Neural Networks. The framework enables optimized resource allocation and significantly improves the precision of predictions. Ultimately, the results will empower decision-makers with reliable real-time information, facilitating informed choices, and enhancing the resilience of the democratic process.”