YouTube blogs today about how their ratings system is broken, and not doing what they thought it should be doing. The nub is: people tend to rate either 1 star or 5 stars, with the majority rating only 5 stars. This indicates that people tend to rate when they like a video enough, or less commonly, when they dislike it enough. The comments, of course, open up the whole can of worms around whether to use like/dislike indicators only likes, both like/dislike and ratings, and all sorts of other scale systems.
Which totally misses the point.
YouTube is a cultural resource. As a freely usable platform for expression, passing the time, making jokes, commentary, finding & sharing culturally meaningful events - and so on - YouTube generates enormous cultural capital. But the ratings systems don't capture that cultural capital - they merely capture one indicator of interestingness. The debate around ratings systems & scales is essentially one of classificatory accuracy: how do we get people to tell us how much they like something, so we can make a good judgement on how large masses of people like that same thing? The elephant in the room - the one not being addressed - is what exactly is the point of rating? Who benefits? What is YouTube trying to identify? A common enough answer is: the videos that are interesting. But that doesn't hold up: YouTube currently has 4 systems for new video discovery: network-based delivery, a viewing-trail based similarity & recommendation system, a curated set, and a current-activity view. None of them really use the rating system (or at least not visibly so.)
Answer 2: to separate the wheat from the chaff. But that doesn't stand up on scrutiny, either. Even if people had perfect agreement on ratings scale (they don't), any ratings system would still suffer from selection bias. Also, accurate classifications are only one of several use cases: for selecting a video to watch when having to select between multiple similar choices (picture quality, possibility of interestingness). Most of the time - especially with Facebook sharing and embedding and so on - the portability of YouTube videos means that they are in a specific context, and either watched if they seem interesting, or not at all. (In which case the static screenshot might be a much better signal of interest).
But neither of these have much to do with cultural capital: knowing what gets people attention and their engagement.
Culture is created through interaction. Stands to reason, then, that cultural capital - the amount of attention something gets - should also be measured through interaction. Here, then, based on thinking about the interactions one has with videos on YouTube (and taking into account the fact that YouTube videos are not really social objects within YouTube), are a few measures of cultural capital
- no. of times favorited
- no. of times embedded
- no. of times linked to / blogged about
- no. of times replied-to
- no. of times remixed
- no. of times removed due to copyright violations
- no. of times downloaded
- no. of variations uploaded
- no. of times re-uploaded after removal
- no. of times commented on per view
- no. of times added to a playlist
Of course, for some of these we're going to need object descriptors beyond just IDs and URLs, but Google & YouTube have a bunch of smart engineers, and I'm sure they can figure something out, eh?
