History:Subjective vs. Objective Data

From MusicBrainz Wiki
Revision as of 18:24, 11 November 2011 by Reosarevok (talk | contribs) (Subjective vs. Objective Data moved to History:Subjective vs. Objective Data)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Status: This page is part of the set of MusicBrainzPhilosophy pages, which discuss the rationale behind metaphysical decisions/suggestions in regards to MusicBrainz data.

So far, MusicBrainz has concentrated on recording objective facts - which song appeared on which album, who sang on it, when it was released, and so on. However, some information that would be worth recording is not so objective. Some things are a matter of opinion; but no less useful because of that. An example is genres. Is "Black Metal" really a distinct sub-genre of "Heavy Metal"? And if so, which of the two covers the band Slipknot? There will always be disagreement over the details of the fine distinctions, and one person's opinion is as valid as anothers. But there is no doubt whatsoever that Slipknot is a heavy metal band, whereas The Backstreet Boys are not. So there is some objective reality there, there just aren't any fine lines.

Virtually every other general-purpose music site out there makes some attempt to categorise releases by genre, even if it's as minimalist as separating "classical" and "pop" music. FreeDB has genres. Allmusic.com has genres. Amazon has genres. It would be an incredibly useful facility for browsing the database, and it's clearly technically possible, so why not?

There are several problems that I can see, and I think they apply not only to genres, but to any kind of subjective data:

  • Someone needs to put their foot down and just decide which genres will be represented and which won't. This will inevitably piss off people with a different opinion, which, as already noted, is just as valid as anyone else's.
  • The list of genres, which for a database as ambitious as MusicBrainz will inevitably be very long, will be a lot of work for developers to maintain - especially since the classification will tend to shift over time, with genres gaining and losing prominence, or being revised as a style develops; and of course new genres are continually being invented.
  • Musicians are an ornery lot, and they frequently deliberately challenge whatever boundaries society imposes on them. Tell a musician that he's playing southern rock, and chances are he'll go straight into the studio to record a rap album, just to prove you wrong. Most new genres are created because a musician was deliberately trying to cross over boundaries, or push a genre in a new direction. This makes any classification scheme inherently futile.
  • Because the information is subjective, it's unstable. What's classified as "heavy metal" today might be classified as "traditional rock" tomorrow, either because the meaning changes over time, or simply because as more users get added to MusicBrainz the prevailing opinion changes. This "churn" in the data means that clients need to process bigger changelists than they otherwise would have to. Ideally, we should be able to enter all the data for an album and know that that's it, the data will never change.

These are arguments against including subjective data, but they aren't knockout blows. Eventually we're going to want to include some subjective data, simply because it's too valuable to users of the site. But subjective data will always have a higher usefulness hurdle to clear before implementation can be considered a good idea.

Some technological changes to MusicBrainz might make subjective data easier to record. See GenreAlternative and SurvivalOfTheFittest.