Permanent Links


What should be the topic for the next Impossibly Stupid poll?

A Town Square Poll Space

Tech Corner

See Also

[ICO]NameLast modifiedSizeDescription

[PARENTDIR]Parent Directory  -  
[TXT]README.html2014-11-29 11:26 5.9K 
[   ]info.json2014-12-09 01:41 40  
[   ]tags=52k2015-03-26 18:28 0  
[   ]tags=apple2015-03-26 18:28 0  

How many one star songs are in your music collection?

Personal rating systems are impossibly stupid. That goes not just for music, but for photos or movies or food or books. It goes for anything that you rank on an arbitrary scale that nobody else will see. Ratings are an excellent example of something that has high value in a social context, but essentially no value (or even a negative value) on a personal level. Let's look at a few reasons why that is.

Cherry picking

Your collection of whatever is not comprehensive. It simply can not contain items you've never been exposed to and likely does not contain items you tried but don't like. If you were stupid enough to go in and rate what is left over, you're probably going to give everything an above average rating (3 or more stars, if it's a 5 star scale), meaning half the scale is wasted right off the bat.

Odds are good that a high percentage of items in your collection are your "5 star favorites". At least this week. When you find a new song or a new artist that catches your ear, what becomes of those old favorites? Do you have to manually go through your entire collection and change all the ratings so that they incorporate how you feel about this week's "cherry"? Or does it become just another in a growing list of undifferentiated 5 star favorites?

Your social network doesn't care

It'd be different if you were a movie critic that published your preferences, all of your ratings for everything you've rated. Then you'd have something that can be cross referenced and compared and meta-ranked. You then would have some items that only got one star, and your rating system would thus be less personal, but far more useful.

But even then, what you give a low/high rating to has questionable use to anybody else unless you specifically rate items for an audience. Just because you give something 5 stars in your personal collection doesn't mean I'd give the same thing 5 stars in my personal collection. Even if I agree with all of the other ratings you've given to other things (i.e., I think you're a 5 star reviewer!), I might think you're way off in this one instance. When that is not taken into account (it is pretty much never taken into account), the rating system remains more about you than the item you're rating.

Your software doesn't care

It might pretend to care, but the odds are that your software that went out of its way to support the rating feature doesn't actually do anything useful with it. Maybe it'll allow you to sort based on that attribute! Ooooo; really useful . . . At best, all it probably does it let you play highly rated items more often when it's in some sort of shuffle mode. As discussed above, that's of questionable utility when you have to manually readjust the relative ratings all the time. In fact, what most people want from a "ratings" system, at least for music, can be accomplished automatically if you take an impossibly stupid approach.

You owe me, Apple!

Anybody who wants to jump ahead of Apple on this feature should probably pay me to manage the implementation of what follows. Anybody at Apple who just wants to avoid a lawsuit before making millions off this idea, simply send me a check for $52K. I don't think that's unreasonable.

The obvious replacement for ratings is already there. The software knows what the play count of an item is. Consider that a starting position. Something I've played 30 times (Song A) is intrinsically more highly rated than something I've played only 5 times (Song B). The software also already knows the date a song was added. Consider that a velocity. If Song A has been in my library for 6 months, that gives it about the same velocity as Song B if it has only been in my library for a month.

The next factor to consider is, naturally, acceleration. This can be only very roughly determined based on when the song was last played. If Song A was last played 2 months ago, clearly it is decelerating relative to the new Song B. That information alone is very useful in determining an automatic rating for a song, but I'd highly suggest maintaining an entire history for when a song gets played. With that little bit of extra effort, you'd essentially eliminate the busywork of keeping songs rated, and you open yourself up to the possibility of the next potential breakthrough.

Mood Music

I'm amazed that this concept has escaped music companies for so long. Mood Music. It's a phrase that everyone recognizes, but is all but invisible to things like iTunes. They've given us blanket concepts like genre, but that doesn't capture my personal connection to my music

Each time we listen to a song, we are in (or looking to be put in) a mood. I listen to certain songs more when I'm in a festive mood around holidays, when I'm in the mood to exercise, when I'm being romantic, when I'm feeling like a badass just walking down the street, and all sort of other particular moods that I've found match particular songs in my library.

Don't make the mistake of thinking I'm talking about playlists. Again, like ratings, that is something you have to manually maintain. Playlists are an approach that works backwards; you can only guess at my mood based on the playlist I have painstakingly constructed. What I'm talking about is adding the ability to specifically give the music player my mood, and that becomes context for the songs I'm playing. That is information from which you can construct a new playlist. Or possibly never have to construct a playlist again. What kind of time savings is something like that worth?