A really hard question
I’m sitting in the back of the third day of the conference here in Chicago, and a fairly animated question has just been asked of Changbom Park, who has been working with a set of galaxies classified by an automatic routine. (You can find some of their results in this paper). I’ll need more thinking time to properly blog the debate (and the responses we’re getting to Galaxy Zoo) there’s one quote which sums it up :
We don’t classify zebras as horses because they look the same – we use the colours.
The equivalent galaxy argument is whether we should classify according to shape (which we call morphology to confuse people) and then look at colour, as Steven does or whether a ‘true classification’ would take into account all the available information. The consensus seemed to be the former, but then I’m biased…
It’s a very good question and if we stay with the analogy: both the zebra and the horse are part of the family “equidae”, which share a similar evolutionary heritage. The difference between the colour — whether it’s a horse or a zebra — could be viewed like spiral galaxies, many are blue, some are red. Both pieces of information are important.
Is there ever only one way to classify, I don’t think so. There are dozens of different ways to classify anything depending on what you are trying to achieve.
I will draw big analogy to an other of my interests Birds: You can classify by family, by what they eat, where they live, colour, migratory or not, size, solitary or flocking and so on. Each classification is useful in its own way – its not more important to consider diet against colour, it depends on if you are measuring.
So I believe there are many attributes one can classify of any object be it bird or galaxy, it is then up to the context to determine the relative importance of those atributes.
In some sense doesn’t the classification
scheme depend on the goal? I often find
the idea of a “right” way to classify things
somewhat disturbing. These are the ways different humans classify things. I prefer
to leave the definition of the “right” or
“true” way to the god(s).
As far as “all available information” is
concerned, every scientist should know that
not all available information is significant
for all purposes. If that were true, wouldn’t
one want to use all four forces to model the motion of the solar system?
Maybe it’s time to take a leaf from the Biology textbook: nothing in biology makes sense except in the light of evolution (credit: Dobzhansky, 1973).
And so it will surely become in extra-galactic astronomy one day too … the classifications (plural) of galaxies without some sort of acknowledgment of the relevant astrophysics is just a form of stamp collecting (cue the famous Rutherford quote), a point which JayT has already made, indirectly.
Maybe I should revive my project idea (a version of GZ using GALEX+POSS images as inputs), but add something about FIRST or HIPASS?
I guess the question, based on the zebra/horse analogy, is, “Why do we do that?” The answer is that we knew a lot about horses before we (western Europeans anyway) ever saw a zebra. So the classification scheme was heavily biased to begin with. At the same level of bias we classified spiral galaxies as spiral nebulae, and had no idea at all about elliptical galaxies.
In the Galaxy Zoo, we use the Hubble classification system. We could, I suppose, extend to the de Vaucouleurs system, though it’s a bit harder to teach to the amateurs. Or we could use the Yerkes/Morgan system and classify based on integrated color. But I think the creators of the Galaxy Zoo chose to hold to the Hubble classification system because it’s simple and fairly robust. It also fits pretty well with our (lack of) knowledge about the details of galaxies. There’s no point in having a too tightly defined classification scheme when we don’t yet really know if there’s any fundamental difference between an E0 and an E7 other than line of sight appearance.
Apart from the question of classification based on form or colour, I am most interested to know the comparison between automated classification techniques and human observation as we are doing in GZ. In other words, which is considered to be the better method ? Has it been settled for good that human observation is superior to automation as regards recognizing patterns ?
Nereid – we’ve talked about the possibility of adding other wavelengths including GALEX data. It needs careful thought to work out what would be feasible, and what classifications are important.
Joseph – We need to do a proper comparison with automatic classification methods, and I spent sometime talking to Steven about this yesterday. We’re convinced that Galaxy Zoo-eque classifications are uniquely consistent, and uniquely sensitive only to the actual shape.
In some sense the test is whether other astronomers use our data when we make it public. The number of requests we’re getting tell me we’re getting this right.
Whatever our comparisons show, though, the additional benefit of classification by eye – the ability to find the unusual, whether lenses, overlapping galaxies or Voorwerpen is unique to this sort of project.
Thank you very much, Chris for your positive and encouraging feed-back. It is reassuring for all Zooites to learn that there is a growing trend for our data to used by the scientific community at large. It makes us feel that our efforts are appreciated and have been worthwhile. Keep up the good work, TEAM !