Show HN: Jabble – Collaborative Fact-Checking with Fallacy Detection
jabble.xyz
we are three researchers (Felix, Jonathan, Johannes) and over the past years, we've been doing research on polarization and misinformation. An earlier project from us is Quality News (https://news.social-protocols.org/). This is our first prototype for a platform for collaborative fact-checking.
What we have so far:
- automated detection for rhetorical fallacies in posts. This is done using gpt-4o-mini. We might release that as a library. Let us know if you're interested.
On the roadmap:
- calculate fact check results taking all the votes and comments in a thread into account
- comment ranking by measuring how mind-changing a comment is. We'll decide this by analyzing voting patterns using bayesian statistics.
- game theoretical reputation system with incentives for honesty - bias correction for votes with matrix factorization (similar to Community Notes)
- crowdfunding of fact checks (money distributed to contributors based on the effectiveness of their comments and votes)
We're aware that the true/false vote buttons are not ideal for all the content people will post. But we think it's a start. If you have ideas in this direction, please let us know.
We appreciate any feedback!