This weekend, I built a prototype for a kind of news aggregation I’ve had on my mind for a while, which I’ve called Panorama.

The idea is fairly straightforward: look for all possible articles on a specific story, across many different sources, then put all the headlines in one place, in chronological order. That’s pretty much it.

For my proof of concept, I picked Michael Flynn, for a few reasons:
1. His brief tenure made it easy to scope down the time range
2. It’s not an ongoing controversy (Turkey lobbying aside), and a relatively self-contained story with a beginning an end is easier to grok
3. Responses to the scandal are mostly polarized

The notion is that, by placing all news sources together, trends will emerge. It should be possible to identify an agenda, for bias to be (more) self-evident when placed in the context of news as a whole.

In terms of methodology, some things worth making note of:

  • This was all collected by hand. There are certainly errors.
  • I did a lot of googling and did my best to find every relevant article on the Washington Post, Breitbart, and the New York Times from the election until now. Other sources are just articles I found along the way.
  • I did not include syndicated articles – AP, Reuters, UPI, etc.. These articles don’t really represent what I’m interested in. Also, it turns out that Breitbart buys almost every syndicated article, no matter how redundant, so there are hundreds of these articles just for the last month of news on Flynn. I assume this is for SEO purposes, since it makes Breitbart much more likely to show up on any given search on a topic.
  • Just getting reliable publication times can be a huge pain. For all of these, I had to open up the source for a timestamp – sometimes I could get an ISO string, other times I’d have to convert from a plaintext time.
  • Some articles just don’t have accurate publication times recorded, e.g. there are articles about¬†Flynn’s resignation showing up before¬†he’s actually resigned.

Deeper thought on this experiment to come later.