Slashdot Mirror
Using Distributed Wetware To Analyze Mars Craters
A non-mouse Cow Herd writes: "Here's an interesting NASA project that popped up on sci.space.sience
a while back:
This site
allows volunteers to classify craters on mars, essentially a human
distributed image processing program. They are even
starting a moderation-like quality rating.
So what do you think ? Would you devote your spare cycles to
this ? Will the get quality work or just a lot of First Posts ?" Seems almost (but not quite) paradoxical to use an ultra-high-tech infrastructure to organize non-automated piecework, but until there's a sand mouse for crater-recognition, it seems like a really smart idea.
Right up until it got posted to slashdot.
You proofread OCR'd text for Project Gutenberg using the raw scanned image to fix anything. You can do as little or as much as you want.
*(L)user 1* Hey! That's a crater!
*(L)user 2* Cool! Which one?
*(L)user 1* Nevermind, it's just a picture of your mom! (Or other filthy object.
Seriously, using PEOPLE to do anything that requires more than clicking and flaming? I don't know if this project is gonna work.
---
but with people. this is just www.amihotornot.com for craters.
Start Running Better Polls
right now they only have one person looking at each crater. This can and will lead to some errors in clasification. Now im gussing that right now they have more than enough craters to keep every one busy, but if they had enough ppl , they what they should do is have a system where every crater gets look at by several people. Craters that get unanimos votes by the first 3 veiwers wont be resent, but those that don't will keep getting sent to new people until there is enough certainy in the classification. also if you do things this way, you can start rating the classifiers by how often they agree or disagree with the majority opinion. this will also protect against malicious classifiers who intentionally do things wrong
I think its kind of insulting to be handing out work to volunteers that has no scientific value (see FAQ quote at end of my post) other than determining in this specific context, "the public is ready willing and able to help science." Even if this study proves that the public can help NASA out - how does it prove that the public will help in the future? Furthermore, couldn't they find something more worthwhile for volunteers to do, and still use that experience to determine if the public can help out.
I find a similar problem often when I volunteer to help various organizations in and around my community, I am assigned some menial task or there is no important work to be done. (i.e. It would have been better for me to work an extra hour and donate the money I would of made to the organization, so they could hire someone for at least 5 hours of menial work) Even worse, a lot of times they might make me do something in a ridiculously inefficient manner. In other words if I used my "geek" background and somehow implement a computer or what have you - the task would be accomplished much more efficiently. If that's awkwardly put sorry, but I am sure the
Sorry about this ranting, but I just had to vent. Don't get me wrong, sometimes volunteer work can be very worthwhile, but it seems more often than not - it isn't.
From the FAQ :
Q. What scientific questions can be answered by the data that we clickworkers are providing?
A. The first stage of this pilot project is only trying to answer some meta-science questions:
Is the public ready, willing, and able to help science?
Does this new way of powering science analysis produce just as good results as the traditional way?
So, at first, we will not be breaking new ground, just reprocessing Mars images that have already been analyzed.
The palimpsest craters and the nearly wiped craters are some of the most important for age classification. This allows to provide a real picture of the age of Mars landscape. Recently some other things came into the game, the "crater-deformed" landscape. Structures, mainly in sedimented regions, that still retain traces of anciant craterization. These ones are the most hard to detect. Only the circular deposition of sediments, and some deformations on soil still show that we are in presence of craters.
All these ones can only be find through image processing. And oh my! Here the work is HUGE. Image processment allows mostly to remark the details of landscape. And we are not looking for craters in the right place but exactly in very wrong places. A wiped out crater in a edge of a cliff will give you a hint on how old such cliff is. Another crater in a sedimentary region will show how long sediments covered the landscape. And there is no common law to ease this task.
Meanwhile this classification is not bad at all. However I doubt that NASA will get any good on it. They are too stuck to their ideas of old dry Mars check a few things about ages. For example, I wondered how old they would call such crater like the pedestal crater in Janssen's Crater. At the beginning I thought that the thing was a "post-water" crater. In fact the study of very small craters showed that the thing hit Janssen's while still a sea. And it kept being such for much longer. The signature of small, nearly wiped and mostly invisible craters showed a very long period for the presence of water. An that Janssen still holds a lot of its morphology of that period.
Btw. Till now, water runs from a few places there. It oesn't live too long in the surface but it is there, underground, in the millions of liters.
Q. What scientific questions can be answered by the data that we clickworkers are providing?
A. The first stage of this pilot project is only trying to answer some meta-science questions:
Is the public ready, willing, and able to help science?
Many years ago, volunteers (mostly housewives) were enlisted to help analyze thousands of photos of cloud-chamber tracings. Scientists were looking for evidence of a particular particle, and in those days only human inspection could be used. To make sure that the volunteers would find the trace if it showed up, photos with phoney traces were periodically inserted. As I recall, the program was considered a success. It seems to me that the question has been answered. And I would also think that a similar process of inserting known items to make sure the volunteers are doing a proper job could be used.
"If I have seen further than other men, it is by stepping on their glasses." - Michael Swaine
I think the NASA project is basically make-work. After all, after all of the users have added their time and energy, the results are thrown away. They're not used to teach computers to recognize craters in the next evolution.
You don't see the point here? Crater classification is needed to create the historical map of Mars. Till now many analysis are made in the "pick-frame" method. You pick several frames through some method of selection. You analyse them nd throw a conclusion. Good and Bad. In my experience I saw a few times when NASA splitted a very good one, confunding a landscape formation with other just by ignoring some frames.
A systematic selection of craters will allow to form a more strightforward picture of the landscape. This work is massive. And computers, here, have several drawbacks to achieve this. Craters may have some common and well seen morphologies. But they also have a lot of individual traces that no computer will ever detect. Some of these traces can be quite confusing. For example a wiped out crater occurs to be more recent than its nearly preserved neighbor. Funny, but it does occur.
Right now it is possible that this clickwork is quite raw. But still is the right step to go. We need that map. Or else we will keep seeing hordes of investigators claiming that this place is that old, no it is that new or it's an alien face...
This is a great idea! Many people have more than a passing interest in science and space - just look at the number of people on Slashdot and SETI@HOME. Projects like these have been done before with large numbers of participants (grade schoolers looking at plant genetics, people identifying and counting birds, etc) and have generally had good, robust results.
Pros 1-large number of interested people - should be easy to get stats on whether everyonne agrees that this is a type A or B crater.
2 the human eye/brain is by far the best pattern recognition tool ever seen - i.e facial differences are relatively minor, yet everyone looks different and is uniquely identifiable.
3 interesting project - inexpensive too!
4 Data is "blinded" to the judges - no one knows where these craters are, thus if alot of people see water effects in many craters in one area a reasonable conjecture can be made.
Cons 1 possible high "noise" level (first post effect) - this can be defeated by introducing a training session which should weed out the first posters
2 possible poor inter-observer correlation -always a more of a problem with complex or more judgemental analysis - I see this in orthopaedics when looking at x-rays of bone fractures. Simpler systems work best!
..........FULL STOP.
http://www.nytimes.com/2000/12/16/opinion/16DAVI.h tml
--
Well why notto make YOUR site? In the eve of Internet there were lots of sites studying Mars images. Truly many cared for Mars Face and similar things. But there were others who went on hunting more things than just Fussy Faces. There were more scientific or less scientific sites. Many spoke about the history of Mars, the possibility of water, life or even our dear little green men.
However this was all gone. By 95%. First because people either got tired or suffered some desillusions. Second because some pedestal people decided that this was too bad for their "Science". I got caught twice in this one. First by kicking me out of Yahoo. A year later, by smartly wipe out my presence in every main Web index. Some people were kicked out by getting so much hard flame and dirty spam that they got tired of fighting. Some got caught in a smart campaign of Mars bashing - "Elvis leaves the stage and Big Foot is in" as we saddly commented.
So you may try to do your site. But be warned. Things are not so simple as they seem.
Besides I have site on Internet. An horrible and primitive site. What i managed to do after the third time of being downed (this third one due to some damn sysadmin who knocked the server completely). But I will not tell where it is...
Seriously, I got: