Thoughts on detecting fake news
In recent times, it has become difficult to trust content you read. I think below steps at sharing platform may provide users with some insights into credibility of content
Source of content ( when content is shared from external sites)
There are sites which publish satirical/fake news or have reputation of publishing unverified content. One approach can be to associate site with following:
- Trust score — on scale of 0 to 10. 0 means mostly fake, 5 means no information (neutral) and 10 means that it usually reports correct news. Lower the trust score, less trustworthy the site.
- Label about contents — ex: news, satire, rumours etc
This can help users with some pointers about the content. It may not solve the problem but it may provide more information to users.
Users who generated the content
For content to be available on platform, someone needs to introduce the content. If it is fake, it will probably be generated by fake users or users who have done so in past. Associating trust score with every user may help. So, if new content is introduced in platform and users who initially shared it has low trust score, we may flag content.
Sharing relevant data from internet
One may provide links from trusted sites related to content user is reading. This may reduce inertia in users to fact check the content. It enables users to make more informed decision.
Detecting which content should be manually fact checked
Manual fact checking can be another option. But with all things manual, it can’t just keep pace with digital world. With millions of posts/tweets created every hour, you just can’t check every one of them. One can associate popularity score with the content, which can be function of number of users who can be influenced by particular post/tweet. Content with high popularity score can be fact checked first and with low popularity score can be dropped.
What are your thoughts on solving this problem?
