The Problem
In this age of rapid information exchange, the rise of fake news and misinformation poses increasingly pervasive threats to the digital landscape, often accepted without verification and reinforced within social networks that echo similar values and opinions.
The prompt
We provided the Fact Checker AI Agent with a clear prompt, defining its role, task, inputs, confidence intervals, and the methodology for calculating the confidence score.
The AI calculates the score by assigning a binary confidence rating to each identified pain point, then averaging these scores to determine the overall confidence.
Lastly, the AI evaluates accuracy and verifies sources by compiling a structured table that maps each claim to its corresponding references, ensuring transparency and traceability.
The process
The MD FactFarm browser extension streamlines data analysis by extracting video titles, transcripts, and user comments from a YouTube URL via API. Leveraging Fetch.ai’s AI agents, one processes the gathered data alongside a user prompt and communicates with the Fact Checker AI Agent to efficiently consolidate and verify information.
The pitch
Our pitch highlighted the issue of accepting information at face value and presented a solution addressing two key pain points: Convenience - easy access to fact-checking via a Chrome extension, and Empowerment - fostering critical thinking for confident navigation of the digital world.
Retrospective
Reflecting on Judging Day, we’re proud to have pitched our product and launched it successfully, earning first place in Fetch.ai’s sponsorship prize. This recognition highlighted the hard work and innovation that went into meeting our product goals.
Along the way, we tackled significant challenges, such as refining agent interactions within the constraints of web architecture (i.e., navigating issues like latency, scalability, and data consistency) and determining the best approach for calculating confidence scores using binary ratings. While we’re thrilled with our results, given more time, we’d focus on further enhancing the precision of our scoring methodology.