Candour

Candour enables messenger users to quickly fact check information they find online.

Comment

Inspiration

I think we have all experienced the divide that misinformation has caused in our society. And misinformation continues to spread despite efforts by fact-checkers. With Candour, we might be able to reduce the friction from the process of fact-checking and make it more accessible to everyone.

What it does

Acquires data from IFCN (International fact-checker signatories) such as AltNews and Politifact and allows messenger users to cross-check information they find online.

How I built it

Flask mostly, and a lot of coffee.

Challenges I ran into

  • Not having enough resources, this whole project is dependent on having a really large database of misinformation. I would've loved to do something beyond matching image hashes for comparisons but that would probably require better infrastructure.

  • Matching text-based queries with content from online articles. I ended using an approach where I extracted keywords from each article and joined them into a string and then used a Sequence Matcher to calculate similarity scores. Although it's not the best solution, It gave me better results than the other approaches I tried.

Accomplishments that I'm proud of

  • The data acquisition and parsing module is pretty cool in my opinion. It can parse most types of news into a specific JSON structure which makes it easier to query.
  • I was really dreading making the video, but I actually had a lot of fun doing it.

What I learned

  • I hadn't really tried full-text search on MongoDB before, I would probably use it more in the future.
  • Really liked page-scoped user IDs as a system of preserving user privacy

Testing the bot

The FactCheck database won't get updated until the judging period ends to abide by the rules. Here are some things you can ask candour till then. Fact checks were obtained mostly from IFCN signatory Altnews (Altnews.in), but the framework can be extended to work with any other fact-checking portal.

Here are some examples of misinformation for you to try out.

Questions

  • Is the Whatsapp forward about the lockdown protocol true
  • Does disaster management act prohibit coronavirus posts
  • Does magnet therapy treat diseases

Facebook Post Links

Tweets

External links

Page Link

https://www.facebook.com/candour.framework

What's next for Candour

Get more sources, as of now Candours fact checks are a little too India focussed because those are the sources I went with. I think supporting local languages might also be a good idea. Improving the matching algorithm might be something that could make a lot of difference.

Share this project:

Updates

posted an update —

Update 1

Based on the interactions I've had with the community, and the feedback that I've received from friends, I've decided to add the following two features in the next version of candour.

  • Top 5 Misinformation
  • How to fact check

Top 5 Misinformation

I will be tagging each fact check document that I acquire with a location tag. Users can ask candour to give them the top 5 most popular articles of misinformation based on their geographical area. The popularity score will be calculated based on the number of times other users from their area have searched for it.

How to fact check

Asking candour 'how to fact check' or anything similar will tell users basic information about how fact checks should be conducted. It will also allow them to play a mini-game where candour shows them content and asks them to fact check it. This would give people a realistic sense of their fact-checking ability and help them work on it.

The next version of candour will be published when the judging period of this hackathon ends. I'm open to all feedback, feel free to reach out to me on twitter -> @tabish_imran8

Log in or sign up for Devpost to join the conversation.