2 ms·
It seems to me that the problem here is not that likes mean the wrong thing, it is that the developer wants to post-process likes to create a metric that means
by btilly 13y ago
It seems to me that the problem here is not that likes mean the wrong thing, it is that the developer wants to post-process likes to create a metric that means something and it isn't working out to what it desired.
The simple solution is to post-process likes in a different way. See http://www.evanmiller.org/how-not-to-sort-by-average-rating.html http://www.evanmiller.org/how-not-to-sort-by-average-rating.... for a suggestion that makes them more like a rating, despite the fact that some people use them differently than others.
But we can do other things. Here is a fun one. If a person likes 2 different photos, we can say that is a connection between them of strength 1/log(# photos liked by that person). The total connection between any two photos is the sum of the connections. From each photo we can now find other photos that are in some sense likely to be liked by people who like the first. People enjoy following connections like these.
This system is similar in principle to the one that I wrote about at http://bentilly.blogspot.com/2011/02/finding-related-items.html http://bentilly.blogspot.com/2011/02/finding-related-items.h.... Which drove a noticeable sales bump.
So the challenge is not to remove likes as a mechanism because it doesn't give you the information that you want. It is to find ways to process the data you get from it to pull out useful information.
- robotmay 13y agoThanks for the links; I'll definitely be reading through those later (as this field is very interesting to me)! I think I may have confused matters somewhat by comparing directly to a 'Like', which is more like a 0-1 rating system, rather than to a combination of Liking and Sharing. How far it goes towards each of those is something I'm experimenting with :) > Here is a fun one. If a person likes 2 different photos, we can say that is a connection between them of strength 1/log(# photos liked by that person). The total connection between any two photos is the sum of the connections. From each photo we can now find other photos that are in some sense likely to be liked by people who like the first. People enjoy following connections like these. I can't wait until I have more data to play with and I can start experimenting with ways of connecting photos like you say; I've not often had the chance to try out some of these ideas due to the limited size of datasets I've had in the past.