Inspiration
Students are constantly making decisions that shape their future—applying for scholarships, joining programs, responding to school announcements, pursuing opportunities, and navigating their communities. Yet those decisions are often made with incomplete, conflicting, or rapidly changing information.
We noticed that most tools focus on answering whether a claim is true or false. But in real life, students rarely need a binary answer. They need context. They need to know what is verified, what remains uncertain, and how confident they should be before acting.
We built Scoop because students shouldn't have to stop growing just because information is imperfect. Instead, they need a way to move forward with confidence.
What it does
Scoop is an AI-powered trust and decision-support platform that helps students, parents, and community members evaluate uncertain information.
Users can submit a claim, screenshot, rumor, announcement, or message. Scoop analyzes the information, retrieves supporting and conflicting evidence, traces source origins, identifies missing context, and generates a confidence-calibrated assessment.
Instead of returning a simple verdict, Scoop explains:
- What is known
- What is unverified
- What evidence supports the claim
- What evidence contradicts it
- How confidence was determined
- What action the user should consider next
Scoop also includes Watch Mode, which continuously tracks evolving claims and updates users when new evidence changes confidence or context.
How we built it
Scoop uses a multi-agent AI architecture designed around transparency and uncertainty.
The system begins by parsing a user-submitted claim and retrieving evidence from credible sources such as official announcements, school communications, public service directories, and news sources.
Multiple specialized AI agents then evaluate the information:
- A Reasoner agent analyzes supporting evidence.
- A Skeptic agent searches for contradictions, missing context, and weak assumptions.
- A Constitutional Review agent enforces trust and transparency principles.
- And more agentic capabilities
The outputs are combined through a confidence-calibration layer that produces an evidence-backed assessment rather than a true-or-false verdict.
To encourage accountability, Scoop includes a delayed feedback system that asks users whether the investigation was helpful and allows problematic results to be reported for review.
Challenges we ran into
One of the biggest challenges was resisting the temptation to build another fact-checker.
Early versions focused heavily on determining whether claims were true or false. We realized that this approach created overconfidence and failed to represent uncertainty honestly.
Another challenge was designing a system that could explain trust instead of simply generating answers. Building source lineage analysis, confidence calibration, and meaningful disagreement between AI agents required us to rethink how information should be evaluated.
We also spent significant time balancing transparency with simplicity so that users could understand the reasoning process without feeling overwhelmed as well as the loading times behind the reasoning, which we tried to solve with langchain response streaming.
Accomplishments that we're proud of
- Designing an Information Ecosystem Analysis approach rather than a traditional fact-checking system.
- Building a multi-agent architecture that actively challenges its own conclusions.
- Creating Watch Mode to track how trust evolves as new evidence appears.
- Implementing constitutional safeguards that prioritize transparency and uncertainty.
- Designing a system that helps users reason about information instead of blindly trusting AI-generated answers.
- Keeping humans in control of final decisions for high-stakes situations.
What we learned
We learned that trust is not a binary state.
Information exists on a spectrum of certainty, and helping people understand uncertainty is often more valuable than giving them a definitive answer.
We also learned that responsible AI requires more than model accuracy. Transparency, calibrated confidence, source visibility, and human oversight are equally important when people are making real-world decisions.
Most importantly, we learned that users do not always need an AI to decide for them, they often need an AI to help them think more clearly.
What's next for Scoop
Our vision is to make Scoop a trusted decision-support platform for students and communities everywhere.
Future plans include:
- Expanding Watch Mode into a real-time trust monitoring system.
- Improving source lineage and citation graph visualization.
- Personalizing investigations based on user interests and community context.
- Building stronger human-review workflows for high-impact claims.
- Developing community trust networks that help users collaboratively evaluate evolving information.
Ultimately, we want Scoop to become the place where people go not to find answers, but to understand whether information deserves their trust before they act on it.
Log in or sign up for Devpost to join the conversation.