[anigbrowl]

That fails if you do a first-order implementation because you have no way of knowing whether the person voting/sharing had already read the article, but not necessarily shared it at the time of reading. As a simple example, I subscribe to The Economist and prefer reading the paper version. If I see a story from there cited by someone else, I don't need to read it again to recommend it.

Also, depending on the content and source it may be possible to infer the value or significance of something from the headline alone, eg when one has specialist domain knowledge and sees a headline one knows to be an accurate summary of recent developments in that field.

[privong]

> That fails if you do a first-order implementation because you have no way of knowing whether the person voting/sharing had already read the article, but not necessarily shared it at the time of reading.

It might still hold for a large group of people, though, and be valid in a statistical sense? Which is what they're really after.

[anigbrowl]

It could definitely help as long as it's appropriately weighted rather than being a binary choice. It'll be interesting to see how it pans out.