Slashdot Mirror
Open v. Closed Source-Climate Change Research
theidocles writes "The ongoing debate over the 'hockey stick' climate graph has an interesting side note.
McKitrick & McIntyre (M&M), the critics, have published their complete source code and it's written using the well-known R statistics package (covered by the GPL). Mann, Bradley & Hughes, the defenders, described their algorithm but have only released part of their source code, and refuse to divulge the rest, which really makes it look like they have some errors/omissions to hide (they did publish the data they used). There's an issue of open source vs closed source as well as how much publicly-funded researchers should be required to disclose - should they be allowed to generate 'closed-source' solutions at the taxpayers' expense?"
should they be allowed to generate 'closed-source' solutions at the taxpayers' expense?
No. I paid for it I want to see it. How else will we know if it works the way they say it works?
I'm a virgo and on Slashdot. Coincidence? Yes.
how much publicly-funded researchers should be required to disclose
All of it, baby. We're paying for it -- we should have the right to:
a) Know what you're spending our money on
b) Have the right to make it better ourselves
c) Learn of security flaws early so we can correct them
Especially when there is some doubt about the nature of the results in the closed source model from Mann et al.
The dangers of knowledge trigger emotional distress in human beings.
I hate projects with names like R. I used R a while back, and it's a great program, but try searching for "R" plugins on Google. Not fun.
The problem with most of these studies is that they refuse to release the raw data.
... but no thanks !!!
A lot of times they select subsets of the data and then normalize or otherwise massage the data.
Thanks
Science, like government, should be transparent. The public should be able to see and evaluate every part. Any science, or government, that hides it's implementation is inherently suspect to corruption.
Closed science is half a step from religion. You are expected to have faith in the researcher's methodologies, analysis, assumptions, and motives. Sorry, but good science does not rely on faith.
The pharmaceutical industry receives huge subsidies from us - they don't produce "open" drugs - why should this be any different? I know it's apples and oranges - but one should be really careful about the idea of withholding funds from -good- research just because of licensing issues. Lesser of two evils? Would we rather have -no- research?
complicated...
Arguments against open-source science:
I'm sure there are arguments on both sides.
Two wrongs don't make a right, but three lefts do.
I'm of the opinion that anything that gets published should be published in its entirety, at least at some point. For example, people who publish protein structures can put the coordinates "on hold" for up to 18 months.
And to say because the research is done with "taxpayer's money" is missing the point: If you can't reproduce every step, it's voodoo, not science. And we make policy decisions based on science, not voodoo (I hope).
Obviously you have never written an SBIR or BAA. You when you do research "At the tax payers expense", you need to show your plans to commercialize the results of the research. The government wants you to create a IP twoards a commerial project which will spur the economy, not to contribute to the scientific community as a whole. Take it as you will, but I think that most research would not get funded if your commertilization plan was to release it on sourceforge.
While I would like all works performed for the government that are not of National Security importance to be more open I don't think it is necessary.
A lot of work peformed for government agencies is contractual with businesses. These same businesses employ tricks of the trade and such to deliver what is required. To have them detail how the work is just suicidal. The same goes for software they develop for use by the government. Unless specifically addressed in the contract I do not believe there is a right to disclose the code, let alone make it available to the public.
That last part is key. Even if they disclose the source to the government there is no obligation on either party to make it public.
This argument that they have something to hide is childish. It is designed to provide no leeway. Simply put, once labeled as such what other option other than disclosure exist? You might as well say "You have to release it, its for the children" and then proceed to use whole "hates kids, wants kids to die" guilt trip that is far to common in politics today.
Summary. Release it if only its an upfront requirement of the project and agreed upon by both parties. In the future a requirement by law that all government projects must be fully disclosed to include the source of any software may be nice but I bet it would have so many exceptions written into it that it would result only in a "feel-good" law.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
The debate is well-documented (by the Mann team, at least) here, here, here.
Simple Unexpected Concrete Credible Emotional Stories
This is an extremely difficult issue, although it sounds pretty trivial.
For one thing, the taxpayer is rarely participating in discussions like this one. Moreover, the success of scientific institutions is often measured in terms of number of patents, successfully launched businesses by former students/researchers, etc. So not only is there little or no opposition to closed-source software (or scientific articles!), there are also good reasons for researchers to go the closed-source road.
Some researchers have a tendency towards secrecy. Some even seem a little paranoid when it comes to their data and methods. You could compare this to the tendency of the OSS zealot to suspect bugs, glitches, and omissions in any piece of closed-source software.
And as a German side-note: There are laws over here that require you to have the patentability of any piece of software you develop checked by university lawyers. GPLing something is technically illegal for a researcher. I have no idea how this is regulated in other countries.
Looks like M&M have a blog too...
Simple Unexpected Concrete Credible Emotional Stories
That Global Warming is a manmade, real phenomona is accepted by 99.9% of scientists in the fields involved. To trot out the "only a theory", "some experts dispute" etc routine is like getting the Flat Earth Society involved every time someone talks about circumnavigation. "Heads in the sand" is going to be on our culture's gravestone when the next lot of intelligent life evolves here and starts wondering why parts of Nevada are 10,000 times the normal radiation level.
The Slashdot Paradox: "100% Overrated"
So a team of real scientists (that is, by folks who work in climate science, not reporters or pundits) wrote a Dummies Guide to the latest controversy. Click on the link for a nice question-by-question breakdown, but I'll spoil the conclusion for you:
(MBH98 is the old paper with "closed" source, MM05 is the new "open source") paper)
Read the rest for more explanation.Today biology heavily depends on specific software to analyse lab generated data. However, even academic, public funded software are not open-source. It's a sad situation, but there are efforts like Bioinformatics.Org trying to change the situation.
---- Where is my mind?
"For me though it's easy. The CO2 levels in the atmosphere have never been as high as they are now (at about 370ppm) and they're expected to increase up to 700ppm"
_ cl imate.html
http://www.geocraft.com/WVFossils/Carboniferous
- Senator Grassley. What were some of the characteristics
which negated the missile theory?
This is the voice of reason in a case where reason is ignored...Mr. Tobin. Well, probably the most prominent--actually, there were two main areas negating the missile theory. One, of course, again, is the absence of impulsive loading, or very high-speed fracture and failure mechanisms.
But second was there were serious issues with every theory, or almost every theory, as to access of an external missile to the fuel, to the fuel tank. Even with, as I indicated earlier, if one would focus on an area where we did not recover all of the fuel tank, there were components nearby that would have blocked or at least recorded passage of any externally penetrating object. And if that were not the case, there were many layers, including the external underbelly of the aircraft, and that was recovered almost--a huge portion of that was recovered.
So that, basically, the only plausible theory for some of the missiles to have occurred would have been if there were missiles such that could maybe get through a 1- or 2-inch opening, make an immediate left, go 90 degrees through a seam, and then maybe take another 90-degree right, and then maybe reverse itself and come back over. But those were some of the considerations.
"Their argument since the beginning has essentially not been about methodological issues at all, but about 'source data' issues [...] Only if you remove significant portions of the data do you get a different (and worse) answer."
You're over-trivializing a DRAMATICALLY IMPORTANT POINT. The original study is focused on North American data almost exclusively for certain time periods. That data (from a single species of tree) skews the results in such a way as to make the current trend seem unique and drastic. On the other hand, if you treat that data source in such a way as to balance it with the other data that is available, you see a VERY DIFFERENT TREND!
The response has been to claim that weighting the data in this way reduces the number of data points unacceptably (I would agree, but that doesn't make MBH98 right).
That's the whole point here, and the other side continues to say, "you're throwing away data" when any competent researcher would have thrown it out in the first place (note: there's an exception. if you produced a report that was specific to N. America, MBH98 would be your model, and it seems to be a fine model for that... N. America is seeing record warming as compared with the last few centuries, and that's all you can extract from MBH98).
Also keep some perspective in mind here. We're in a period where temperatures could rise MORE than ANYONE is predicting and not make a dent in the graph over the last 10million years. If you graph out the last 10 million years, you see that temperatures over the last 10,000 years have been part of a huge, cyclical spike in temperatures. We're at what is likely the peak of a drastic temperature swing, and it WILL plumet again into a new ice age (unless we decide to and are capable of coming up with a way to prevent it). I'm not drawing any conclusions from that, just pointing out that there are natural forces at work here, capable of making temperature changes that we a) cannot yet conclusively explain and b) the likes of which no human has ever experienced.
It's important to keep a sense of perspective and to remember that we have very impressive climate models... all of which might be wrong.
Where's that wealth of information about the secret US wars in Central America in the 1980s? Or in Angola in the 1970s? Or in Chile in the 1970s? Or in Cambodia and Laos in the 1960s? Iran in the 1950s? These secret wars are secret largely for *political* purposes - the military secrecy benefits evaporate within months. But the political purposes - covering liability for abuse, war crimes, and just plain lying about the causes, effects, and benefits of the war - those last forever.
--
make install -not war
How many times have you asked someone, "What does your code do to solve this problem?" and got a description of an algorithm which, when you finally get to see the source, does not match the code?
In my case, the answer to that question is, "Lots." I have had it happen in pure science (neutrino physcis), applied science (medical physics) and software development (database programming, data analysis, etc.)
I am painfully aware that my own published descriptions of algorithms have often left out minor details that may be critical in some applications, but that page limits in peer-reviewed journals necessitate. It is not uncommon to get a call from someone doing similar work asking for details about what you've done, how you've done it, and in some cases, asking to look at source code.
In contentious areas of science such requests are not always met with full disclosure, which is a sign that the people involved are no longer doing science. They are doing politics. This happens a lot, and it brings the scientific process to a halt on the question at issue.
In the case at hand, the original authors have done a very poor job of describing what they have done, and an extremely poor job of defending their work. Their refusal to publish their source code for their analysis gives credibility to their critics.
There are certainly legitimate cases where code ought not be published. If a lab has spent many, many years developing a framework for solving a certain type of problem and wants to get the most advantage out of that framework before releasing it, they may reasonably want to limit it's disemination for a while. But those sorts of reason don't apply in this case, and the source should be made available to anyone who wants to reproduce their actual results. That would just be good science.
--Tom
Blasphemy is a human right. Blasphemophobia kills.
One's interests in keeping clients does not entitle you to make a scientific claim that cannot be peer reviewed. If a paper such as Mann is now regarded as fact, and indeed, makes policy, despite the obvious sloppiness regarding its data management process, then, what is the point of science anyway?
Science is supposed to be about peer review, rigor, that every assumption behind every assertion can be challenged. If, all we have is someone with a Phd can claim that they have a fact as our science, then, what is the point of even trusting them?
Without independent verification and an open process, there's nothing to separate scientists from creationists, and the people are going to pick whoever makes the most attractive sales pitch.
This is my sig.
You do know that Mann writes this website, right? You do realize that the source of your argument (http://www.realclimate.org/) is a shill for Mann and his cronies?
Second of all, there was a flaw in the original algorithm that was pointed out by McIntyre and McKitrick before they even got to the bad data being put into the equation.
And, to top it off, Mann's equation always produces hockey-stick graphs, even with randomly distributed data.
Don't point at Mann's own site as a defense of Mann.
Life, the Universe, and Everything... in my image.
1) Science functions only on open review. If you can't duplicate someone's results, they are useless (c.f. Ponds and Fleischman [sp?]). A scientific result is only of value if it describes a consistent replicatable process. This is why I consider the closed source work to be completely meaningless. It may be perfect, it may be bug-ridden garbage, we'll never know!
2) Every tax paying American has paid for my code and work. While I regularly feel they're not getting their money's worth, I definitely don't feel they're paying me to enrich me. They are, in a very real sense, my bosses, and I AM obligated to report to them, if they care. Think of it as a company requiring rights to your work.
3) As an academic working on a fairly limited budget, open source and free software have been a godsend for me and everyone else I know. We run linux because it's more efficient, secure and FREE; we use free or open-source compilers; and we cobble together high-perormance computers and beowulf clusters out of miscellaneous bare metal and lots of googling. The only piece of software I routinely have to pay for is MATLAB.
... grumble, grumble, grumble, mutter, mutter, Millenium... Hand... Shrimp, I tol' 'em, I tol' 'em.
However there was a link to McIntyre and McKitrick's website in the topic summary. Why was it relevant for Timothy to include that link, but not include a link to the matching item on RealClimate.org? Is it just non-scientists who are allowed to have weblogs about this stuff?
Regards
Luke
#include witty_one_liner.h
It's worth noting that, while it makes sense that taxpayer-funded research should generate 'open-source' solutions, federal law dictates otherwise.
The Bayh-Dole Act was passed 25 years ago, which dictates:
So in other words the government has dictated since 1980 that government-funded research should not produce open-source solutions, necessarily, as the results of research are to be considered private-sector profit-generating centers for the host universities. (The implications for the 'next BSD4.3 TCP/IP stack', or similar advanced research, are obvious.)
Anyway, regarding the 'hockey stick' controversy, Tim Lambert's weblog is worth a read.
The whole point of science is not so that we can trust the opinions of scientists, it is so that scientists can give us repeatable steps to demonstrate a new point.
This whole notion of "it makes the scientists happy so we should just trust them" goes against every single thing that we in the west have fought for since the renaissance.
Your whole argument illustrates this problem precisely. You argue that, "well, even though the key piece of statistical evidence in global warming is questionable, we should still believe in the conclusion."
This is so wrong.
Maybe if scientists published all of their data in a uniform format, to a uniform site, with exact steps to reproduce, all of their source data, and how they draw conclusions from them, then, you might have a field that is useful. But right now, you have got hyper expensive journals all over the place as a repository for articles that only sketch out a discovery and not actually do it, and that simply is not good enough to be taken credibly.
The scientific process is excellent. But today's scientific product sucks.
This is my sig.