Project Description
FactShield โ A Unified Fact-Checking Orchestration Platform
Overview
FactShield is a digital trust platform designed to combat misinformation and foster safe online participation. By orchestrating distributed fact-checkers, leveraging AI for automated predictions, and enabling government oversight, FactShield creates a scalable and trustworthy ecosystem that enhances digital confidence globally.
Key Contributions
๐ Detection & Protection
- Integrated LLM-powered automated fact prediction to quickly assess incoming misinformation requests.
- Supports real-time fact-check requests from platforms like Facebook/Instagram via open APIs.
- Reduces the spread of harmful misinformation by combining AI + human review.
โ
Trust & Integrity
- Fact-checker orchestration system ensures multiple reviews with weighted trust scores before final verdicts.
- Establishes a transparent process where validated facts are stored and reused, reinforcing information authenticity.
๐ฅ Inclusion & Empowerment
- Creates job opportunities for independent fact-checkers, strengthening community-driven trust networks.
- Government dashboards provide oversight, ensuring inclusion of verified and diverse fact-checker groups.
๐ก๏ธ Governance & Responsibility
- Governments can add and monitor fact-checkers while retaining accountability.
- Requesters (platforms/organizations) receive structured, auditable results to enhance responsibility-sharing.
๐ Digital Literacy & Enablement
- By orchestrating fact-checking into a central hub, FactShield helps communities trust verified information.
- Encourages awareness and participation by allowing citizens to see transparent review processes.
Hackathon MVP Achievements
- Built APIs for request management, manual reviews, and fact-checker onboarding.
- Developed workflows for LLM-assisted predictions and manual review escalation.
- Designed UI flows for fact-checkers, government dashboards, and requesters.
- Implemented real-time status updates when facts are validated.
Impact
FactShield directly addresses the hackathonโs call by:
- Strengthening trust & integrity online through structured verification.
- Safeguarding vulnerable populations from misinformation by enabling scalable and transparent fact-checking.
- Providing a practical and implementable plan that balances automation, human review, and government oversight.
- Leveraging public datasets and fact-check archives to continuously train and improve validation models.
Data Story
Fact Shield โ Data Story & Impact
Datasets Used
- Situation and Outlook for Primary Industries (SOPI)
Usage
For this project, we used the Situation and Outlook for Primary Industries (SOPI) dataset as testing data to validate our misinformation detection model. The dataset provided a reliable, government-published source of factual information, which was essential for evaluating how well the model could distinguish between true and misleading claims.
Our approach combined the Gemini-1.5-Flash LLM with Retrieval-Augmented Generation (RAG) and Qdrant as the vector database. By embedding and indexing the SOPI data in Qdrant, the model was able to retrieve accurate context when assessing claims, ensuring that fact-checking outputs were grounded in trusted evidence.
This data story highlights how verified government datasets can serve as a backbone for building trustworthy AI-driven fact-checking systems.
Ethical, Privacy & Trust Considerations
- Privacy: No personal data is collected; only public datasets and claims are processed.
- Inclusivity: Open marketplace allows NGOs, journalists, and certified fact-checkers to participate globally.
- Fairness: Weighted consensus ensures no single reviewer dominates outcomes.
Impact
- Governments: Ensure trusted datasets are used for truth validation.
- Platforms: Automate first-level fact-checking for rapid misinformation response.
- Fact-Checkers: Gain visibility and new opportunities through orchestration.
- Communities: Reduced misinformation exposure, leading to safer and more confident digital participation.
