AI2030 Responsible AI Tools Marketplace

Key Takeaways

The paper analyzes approximately 50,000 tweets surrounding the January 26, 2021 Red Fort incident in India, creating a dataset (tractor2twitter) that tracks social media activity before, during, and after the event to understand how misinformation spread during a real-world crisis.
The researchers used Explainable AI (XAI) methods to classify tweets into three distinct categories—disinformation, misinformation, and opinion—rather than treating all false or misleading content the same way, which allows for more nuanced understanding of different types of problematic content.
The study demonstrates the importance of archiving and analyzing social media trails following major events, suggesting that examining the "bread-crumbs" of online activity can reveal how false narratives develop and spread before they potentially contribute to real-world incidents.
By employing explainable AI rather than "black box" models, the research prioritizes transparency in how the system identifies and categorizes problematic content, making it possible for researchers and stakeholders to understand why specific tweets were classified as disinformation versus misinformation versus opinion.
The work highlights the connection between social media misinformation and real-world consequences, suggesting that responsible AI approaches to content analysis can help identify warning signs and understand the role of false information in mobilizing offline events.

The paper analyzes approximately 50,000 tweets surrounding the January 26, 2021 Red Fort incident in India, creating a dataset (tractor2twitter) that tracks social media activity before, during, and after the event to understand how misinformation spread during a real-world crisis.
The researchers used Explainable AI (XAI) methods to classify tweets into three distinct categories—disinformation, misinformation, and opinion—rather than treating all false or misleading content the same way, which allows for more nuanced understanding of different types of problematic content.
The study demonstrates the importance of archiving and analyzing social media trails following major events, suggesting that examining the "bread-crumbs" of online activity can reveal how false narratives develop and spread before they potentially contribute to real-world incidents.
By employing explainable AI rather than "black box" models, the research prioritizes transparency in how the system identifies and categorizes problematic content, making it possible for researchers and stakeholders to understand why specific tweets were classified as disinformation versus misinformation versus opinion.
The work highlights the connection between social media misinformation and real-world consequences, suggesting that responsible AI approaches to content analysis can help identify warning signs and understand the role of false information in mobilizing offline events.

Tracking Peaceful Tractors on Social Media -- XAI-enabled analysis of Red Fort Riots 2021