I want a go away-button
By Bjørn Borud
In the early days of the web building a search engine was mostly about completeness and recall. To make sure you had the content and help users find it.
Now, 30 years later, we have the opposite problem. Search engines and content services have what you are looking for, but you won’t find it because it drowns in content you are not interested in.
I want a junk filter
Every time I want to use Netflix, HBO Nordic, Amazon Prime, or any of the other streaming services available in Norway, I end up browsing for a while, then giving up. The streaming services mostly try to sell me on content I’ve browsed before, but that I’m not interested in. More often than not we end up watching the news and then I read a book or head down to my office/lab at home to work or tinker.
I need a mechanism to filter out the junk. I’m not interested in superhero movies. I don’t want to see movies that feature Tom Cruise or Nicholas Cage. I don’t need to see movies I’ve already seen (that’s what a history feature is for).
Ditto for search. For instance, who really wants hits from Pinterest when performing a web search?
APIs
I recently checked to see if Netflix had an API so that I could make my own searching, ranking and filtering system for finding content that might interest me on Netflix, but it turns out they’ve shut it down. There are some half-way alternatives, but that’s not really the point here: why not offer APIs that allow third party developers to enrich and improve the service?
What is the worst that could happen? That someone succeeds in making a system that makes users use the service more and some marketing exec can’t claim this as their own accomplishment?
Learn from existing systems
Ever since I briefly worked on Google Reader back in the day, I’ve wanted to add a flexible per-user scoring system to a content curation system. Because most systems don’t do this sensibly.
And I wouldn’t even have had to invent one since Lars Magne Ingebrigtsen already did that for USENET News. The Gnus newsreader had a very good scoring system. Every post would start off with some default score. Then scoring rules are applied either adding or subtracting points. Minus 500 points if the author is annoying. Add 1000 points if the posting is on an interesting topic. If the final score falls below the “expunge limit” the post is never shown. The rest is sorted in order of priority.
I’d recommend having a look at Lars Magne’s work and think of how this kind of thinking can be applied to other services.
You can either build it outside the platforms in question, by publishing APIs to content metadata, or you could build it into the service itself and expose an API that allows develoeprs to have a crack at developing interfaces for managing filtering and sorting rules.