08 April 2010

But is it science?

My wife, a writer, and I have a fair number of discussions about the creative process.  For all that science and literature are 'supposed' to be opposites, dislike each other, and so on, we find that there's a tremendous similarity.  Regardless of which you are pursuing, an important question to ask yourself is "Do I have something good?"  Are your characters interesting?  Have you covered every loophole that could take down your hypothesis?  More generally, "Is it art?", "Is it science?".

One of my feelings is that science is about trying to understand the world around us.  In particular, since I'm a physical scientist, to understand the natural, physical world around us.  One sort of question we could ask about the natural world is "What have global mean surface temperatures been like for the last 100 years?"  As you all know, there are controversies about that.

What strikes me, though, is that most of the controversy I encounter in the media, in the blogosphere, or elsewhere, is not about the science.

The most popular vein of controversy is efforts of people to 'audit' the work of people who were trying to answer the scientific question.  And that mostly baffles me.  As a scientist, it's entirely baffling.  As a citizen, I see reasons, some good, some not. 

The scientist in me is baffled because "let me have all your data sets and all your programs so that I can see if I get your results" does nothing at all to answer a question about the global mean surface temperatures.  No matter what is done, whether the exact same answers are found or not, we know nothing more about what global mean temperatures have been.  That makes for a bad scientific question. 

In a good scientific question, you learn something regardless of what the result is.  The science on global mean temperatures is not in 'auditing', but in doing more science.  Ask questions like "The prior work did their spatial gridding in a very simple way.  Do I still get an answer like theirs if I use better gridding methods?"  "The prior work trusted some stations I think are very untrustworthy (insert list of objective reasons here).  Do I still get an answer like theirs if I exclude those stations?"  "We now have much better data from this new source.  Does it show the same trends as the old source?"

Sometimes your answer is a boring 'yes'.  Ok, your gridding method is better, and you didn't use some of the bad old data sets, and you even made the jump to include some spiffy new data source, but still you got the same old answer.  Reassuring that the original folks were pretty much correct.  But boring.

And sometimes you get the exciting 'no' -- you have a significantly different answer.  This played out in the late 1990s and early 2000s between the surface station record and the then-new satellite sounding temperatures.  In 2000, the discrepancy was still present and uncomfortably large (outside the error bars of either the surface record and the satellite record).  That occasioned a National Academies panel discussion and summary http://www.nap.edu/catalog.php?record_id=9755 of the disagreement.  At the time, there was no hard conclusion about cause, basically just outlining the understandings.  It developed later that the reason for the discrepancy was largely errors in how the satellite data were being analyzed (see for example Mears and Wentz 2005). 

My virtual bet was that this would turn out to be the case.  Not because the surface record is great, because it isn't.  Nor that the satellite guys were doing wrong things (not knowingly), but because they had only been at it for a few years.  New observing systems, or old systems being used in new ways, take time to shake out bugs in how they're used.  Surface records had been used for over 150 years by 2000, the satellite had only maybe a decade.  So things were learned and we now have a new observing system from satellite that we can use to add to our understanding about the atmosphere.  Science happened.  Just not the pretty kind we normally talk about in school.

But is it science?  I find this question simplifies my reading greatly.  So many articles are not about the science.  If I'm interested in learning about what we know or how well, it has to be an article about the science.  Articles that are, instead, about how person A doesn't like person B, how the feeling is mutual and all that sort of thing ... even if A or B are scientists, that article doesn't have anything to show me about the science. 

It's also a good question to ask yourself.  I had a notion in graduate school regarding snow formation that I started describing to a friend (who was studying snow).  It wasn't too long before he asked me where the science was.  Plenty of work, but where was the science?  I never really did find it for that idea.  Somewhere near where I was thinking, I'm sure there's a good research project.  But it's outside where I've thought.

17 comments:

Horatio Algeranon said...

For all that science and literature are 'supposed' to be opposites, dislike each other, and so on, we find that there's a tremendous similarity. "

Science is art in a straight-jacket.

Creativity is an integral part, but it's a very special kind of creativity, heavily "constrained" -- one might even say sedated -- by physical law.

Art -- including literature -- has no such constraints.

Nor do most of the "theories" one sees (eg, in the blogosphere) given as "alternatives" to the greenhouse model of warming, for that matter.

To be sure, some of them are very creative, but that's where the similarity with science ends.

They have far more in common with fantasy writing (where what happens is limited only by the author's imagination) than they have with science.

Most people who have no background in science simply don't (can't?) understand just how difficult it is to come up with a model of how the world (in this case climate) works that is internally consistent -- consistent with lots of different records: rising global surface temperature, melting glaciers, melting arctic sea ice, shifts in plant and animal habitats, changes in growing season, etc) and that makes verifiable predictions: increasing height of the tropopause, cooling stratosphere, "disproportional" warming of the arctic.

When scientists have developed such a model (as they have done with the greenhouse model for warming), they are very reluctant (and justifiably so) to just "throw it out" whenever something new comes along.

They know it is much more likely that they have missed some key factor (eg, an error) in a new record (eg, satellite "temps" or ocean temps from ARGO buoys) -- or their interpretation of it -- that makes it inconsistent with the model (and with all the previous data that is consistent with it) than that all their previous thinking was wrong.

That does not mean that new models for how the world works are never right, of course, just that scientists are reluctant to accept them until they have been "proven" to better explain all the available information than the existing model.

Penguindreams said...

Much art involves accepting a straight-jacket, and absolutely demands constraints. I'm a bit surprised to see your response, given your poesy. Poetry often invokes constraints -- haiku, sonnets, ... all dictate major parts of what the finished product must look like.

To take up fantasy, which I do read, if the author could have absolutely anything happen, there'd also be no story. Our Hero could always just blink himself to safety, vanquish the evil with a look (or a magic sword that appears right in front of him a minute before the baddie shows up), and so on.

Fiction, unlike reality, has to be believable. That's quite a constraint. So in the same way that I might ask whether the sea ice pack will act in a certain way in a certain situation, my wife will be asking whether it's reasonable/believable for one of her characters to act in a certain way in a certain situation. Of course she could always make her characters do whatever she wanted. But it will eventually annoy the reader that the characters don't act like people. (Even in fantasy, everyone acts like people even if they're called dwarfs or goblins or whatever. It's very hard to write a reasonable non-human character. So we see humans who have more of some characteristic than the typical human, but still on the spectrum.)

The standard of truth for deciding whether you've written a good novel is quite different than for whether you've written a good scientific article, certainly. And for that end of your comment, we're in agreement. One thing I hope to do over time is provide enough illustrations of smaller parts of the process that non-scientist readers can get a feeling for the larger parts.

Horatio Algeranon said...

Penguin,

You are right:

Horatio's statement that "Art -- including literature -- has no such constraints" is certainly not true in general.

But in Horatio's defense, he was really thinking more in terms of "Art does not demand certain very specific things of the artist" -- at least not the way science does.

Though "Much art involves accepting a straight-jacket," art in does not require it -- at least not in all cases.

In other words, Horatio is free to do a painting or to write a story that is largely free from the constraints of physical reality -- with wizards, flying horses, magic rings and the rest.

Even for art that does "demand" constraints, the constraints are not necessarily imposed from without as is the case with science (though they can be)

Horatio also does representational paintings of landscapes and wildlife

Clearly, there are constraints involved with representational art (if you want to make it look like what you are representing).

But even within representational art (which has far more constraints than "modern" art) the artist still has a lot of leeway (color of rocks and clouds, for example) -- a luxury that the scientist does not have.

You can't turn lead into gold just by choosing a different color from the palette!

So, when it comes to the difference between art and science with regard to constraints, it's probably more a matter of number and kind of constraints (whether imposed from within or without, for example).

Steve Easterbrook said...

I'm reminded of the way science is taught in schools. Much of it involves carrying out some standard experiment by following the instructions in the textbook, dutifully writing down the procedure in your lab notebook, and getting graded on how well your results match the textbook standard. There is no creativity in this, no thinking up new worthwhile questions to address. It's only when you get to grad school that the real science starts.

EliRabett said...

Although Eli doesn't know the inside story, he gets the feeling that someone on the NAS study asked Prabhakara to do a quick and dirty one off in order to get some idea of where the problem was.

That, as you say, is how science is done

jstults said...

Hi Robert,

Thanks for linking from Steve's place.

I think the question you ask ("But is it science?") is not to the point. If what we are talking about is merely investigating the physical world, then the lay public would most likely not care, and calls for audits would be non-existent.

The case here is that the research results are being used as the basis for decision making, so the real issue is decision support. This requires answering more than just the narrow questions a working researcher is worried about.

Eric L said...

jstults,

This requires answering more than just the narrow questions a working researcher is worried about.

Actually I'd say this misses the point. Yes there are lots of different questions to answer, I agree there. But the issue with so much of the climate denial nonsense is that much of it is not an attempt to answer any question at all. For example, Watts' surface station project has volunteer ratings of weather stations. Watts has long suggested that the surface temperature record overstates the warming, but hasn't done much to back it up. So NOAA did a fairly obvious test of Watts' claim -- they compared the temperature trends of their "good" (according to Watts' volunteers) stations with their "bad" stations, and guess what? They show the same result! So Watts hasn't discovered any major problem affecting the surface record. But no good denialist would just take NOAA's word for it right? Well, that's just it, they don't, but why didn't Watts take it upon himself to do this sort of analysis in the first place? Doesn't he want to know whether his various station siting issues actually affect the results? And Steve McIntyre spends so much time critiquing proxy temperature reconstructions of the past two millenia, but really, if he knows the right way to do this and all the paleoclimatologists are idiots, why doesn't he show them how it's done and produce his own reconstruction? The only thing really stopping him is a lack of will or the work involved, but doing real science *is* work.

This doesn't apply to everyone -- I'd say Lindzen and Svensmark and some others are doing some work that attempts to answer the questions they're asking, and while I don't find their evidence very convincing I at least give them credit for doing work that could plausibly end up advancing our understanding of the climate.

Cthulhu said...

I agree with this post entirely. The skeptics are deliberately or not, losing track of the objective of science. They are instead appealing to an (imaginary) scientific method rule book.

Your post is uncannily similar to my post here (mine came later but I didn't read your post first - honest!), in which I am arguing with a skeptic on exactly this topic:
http://www.bbc.co.uk/blogs/thereporters/richardblack/2010/03/the_first_of_the_numerous.html#P94707386

They just don't want to know, they seem so stuck on wanting to believe the science is wrong or cast in doubt that they refuse to stand back and see it as it really is.

Penguindreams said...

jstults:The case here is that the research results are being used as the basis for decision making, so the real issue is decision support. This requires answering more than just the narrow questions a working researcher is worried about.

You're probably going to have to expand some on what exactly you mean by decision support, because as best I can see it, if I want decision support on a science-related decision, my confidence is best improved by people doing more good science.

The 'audit' only shows up where the 'auditors' think that a particular piece of science is answering a question that does support a decision. And only where they don't like the decision that it supports. (Their selectivity about this does not give me confidence in their work.)

Let me change ground to a real case, one that I know the gory back-scene details for, and on a less emotionally charged issue. I make a decision about my running, actually a series of them, based on what the temperature is going to be at 5 PM. I run about that time of day. If it's below 60 F (15 C), I wear gloves. If it's below 40 F (5 C), I'll pull on a jacket and leggings also. And so on. Suppose, now, that the forecast says the high tomorrow is going to be 32 F (0 C), but that strikes me as not credible since I was pretty much expecting 68 (20 C) or so. Suppose, too, that I have a large team of people I can put on the question to conduct an audit of the forecast, and another team I can send off to answer the scientific question -- 'What is tomorrow's 5 PM temperature likely to be?'

Is it the 'auditors' or the scientists who will most improve the confidence I can have in my decision? The question was answered in reality, though in retrospect.

The auditors (I drop the quotes because, unlike the self-appointed 'auditors', these people were experts in the field they were auditing, same as business auditors are) examined the programs for errors, examined the input data for errors and misuse, reran the models with known options turned on or off, and generally beat on the system -- that existed. They found nothing significant. The forecast model was correct, the input data were correct (not corrupted, not misused, any changes in the values for today's forecast were fully supported by changes in the observations, and the observations themselves showed no signs of tampering, etc.) That forecast model given those data was audited to give a forecast of 32 F plus or minus very little.

The scientists, on the other hand, looked at the scientific question of "What is tomorrow's 5 PM temperature likely to be?". So they looked to additional data sets -- newer observing systems, newer ways of using the old data, and the like. They discovered that if you used the latest, greatest observing and analysis system for one of the inputs, you got a wildly different answer -- 65 F for 5 PM forecast.

In terms of decision support, if I take the auditor route, rather than the scientist route, then I have complete (as complete as it gets) confidence that the temperature will be 32 F, and my decision is to dress quite heavily. If I take the scientist route, I don't have a lot of confidence in the 32 F and dress lightly, for the 65 F expected given the newer science, but have my cold weather gear handy for insurance (the newest science is not always robust).

What actually happened was that it was 65 F. The newest science was indeed the best way to go. The auditors never found, and never could have found, the error.

So I lean to science for my decision support in science-related decisions. Yes, it can be difficult to get science to answer a question in a way that is useful for decision support. But it can be done. Same as happened with the forecast system, the real errors, if any, will be turned up by more science, not by 'auditing'.

Cthulhu said...

A good comment Eric L.

Others have raised a similar arguments concerning the instrumental surface temperature record.

The skeptics try to spread the idea that the HadCRUT and GISTEMP results are unreliable and untrustworthy, but
if they really cared about getting to the bottom of the matter and determining that for sure, why do they not produce a global surface temperature record of their own?

Furthermore it's as if they aren't even interested in what temperature did over the 20th century when they dismiss the existing answers to that question but don't bother trying to answer it themselves.

If they applied methods they thought were correct (and those methods could of course be checked by us so there's no scope for them to fake a result), that would tell them either way. It would either confirm the CRU and GISTEMP results in a way other skeptics would be unable to dismiss with conspiracy theories or it would highlight the CRU and GISTEMP analyses are wrong. Given this last possibility is precisely what the skeptics want, it's even more glaring that they haven't bothered producing their own analysis if they are so sure the CRU and GISTEMP results are in doubt.

Cynically, and perhaps I am being a bit unfair here, I think some of the more able skeptics probably have done such an analysis, but we haven't heard about it because they found results similar to CRU and GISTEMP and that wasn't the result they wanted.

Certainly once a skeptic published result is out there in public we could all point other skeptics at it and say "see it shows warming too", just as I do with the UAH record today.

My word verification for this comment was: MakeTSI. Indeed, I wonder what global temperature will do over the next 3 years as solar cycle 24 ramps up.

Penguindreams said...

Eric L.
Watts' surface station effort is an interesting illustration. If you have the scientific question "What are surface temperatures doing?", then it is a perfectly good scientific question to ask "How good are the data that we're using (how well-positioned and maintained are the observing instruments), and how do their failings affect our conclusions?"

As you observe, they stopped with the first part of the question and assumed a conclusion to the second part. Fortunately, a different group put the work in to answering the second part. Which came to the surprising conclusion that the badly-sited stations gave a colder trend than the well-sited stations. Surprise means that there's more good science to do -- "Why do badly sited stations give a colder trend than well sited? And what does this mean for the large scale temperature trends?"

Instead of being pleased at contributing to our understanding more about the surface temperature record, Watts removed his information from public (including scientist) view. Er, isn't the whole point of the 'auditors' that everything should be public?

cthulu:
The people you're referring to aren't skeptics. Real skeptics follow the evidence where it leads. The people you're talking about don't do that. Real skeptics also put in the work to understand a topic and make real tests of that understanding. Again, the people you're talking about don't do that.

The pseudo-skeptics have merely realized that skeptic is a useful label. Same as how some advertisers use labels, and with no more concern for accuracy.

I refuse to help them with their advertising campaign. So use pseudo-skeptic if I have to use a label at all. I'll suggest this to you as well.

Horatio Algeranon said...

Watts' surface station project is more than a little like trying to guess what happened in a "movie" when you have only a small piece of film showing a single frame.

Watts did not even really address the first part of the question ["How good are the data that we're using (how well-positioned and maintained are the observing instruments)"] because doing that really requires understanding how the siting issues shown in the photographs (barbecues, burn barrels, blacktop, AC's) were likely to have affected the temperature changes over time (the trends).

When Watts first started his project, Eli Rabett and many others (Penguin may have done so as well) pointed out the fact that taking a photo does little more than document a given setup at a given point in time.

It says nothing about how that setup actually affects the temperature and worse yet, says nothing about how it might have impacted change in temperature over time (recent decades, for example), because to know the latter, one has to know what the setup was at times in the past -- ie, one has to have a photo record over time.

The movie has to have more than a single frame!

And Watt's implied things at the getgo (referred to in A Tale of Two Surface Stations) that were not even necessarily true.

Imagine Siskel and Ebert giving a movie a "thumbs up" or "thumbs down" after watching only a single frame. That's essentially what Watts has done with the temperature records at the various surface stations.

jstults said...

Eric L. : Actually I'd say this misses the point. Yes there are lots of different questions to answer, I agree there. But the issue with so much of the climate denial nonsense is that much of it is not an attempt to answer any question at all.

No argument there. The point of many activities involved in decision support is not to answer new questions, but to build and establish credibility for the answers we are going to use. Doing decision support is different than doing science. Unfortunately, it's often rhetorically convenient to confuse the two.

Penguindreams: You're probably going to have to expand some on what exactly you mean by decision support, because as best I can see it, if I want decision support on a science-related decision, my confidence is best improved by people doing more good science.

The outcome of the decision may be "go do more research", it might also be to take a particular action because we know enough to act.

The 'audit' only shows up where the 'auditors' think that a particular piece of science is answering a question that does support a decision. And only where they don't like the decision that it supports.

I think this is where the confusion lies. The decision support that science can properly provide does not include support for a particular decision (advocacy), there are many things other than science which are part of the decision process. See this interesting paper from Field and Stream [pdf] for a bit of discussion about what I mean.

Same as happened with the forecast system, the real errors, if any, will be turned up by more science, not by 'auditing'.
You are assuming an awful lot; there are different kinds of errors and different sources of uncertainty. Some errors are quit likely to be found by auditors (coding errors), some by more investigation (errors in physical model choice). Some sources of uncertainty can be reduced practically by more investigation, some cannot.

Cthulhu said:
if they really cared about getting to the bottom of the matter and determining that for sure, why do they not produce a global surface temperature record of their own?
Good question; here you go.

Cynically, and perhaps I am being a bit unfair here, I think some of the more able skeptics probably have done such an analysis, but we haven't heard about it because they found results similar to CRU and GISTEMP and that wasn't the result they wanted.
Or perhaps, it's just more fun to play in your comfortable echo chamber so you didn't look all that hard?

lessalines said...

Thanks for the link to the "Thermal Hammer" jstults, but let's be fair. Calling an analysis of global temperature "Thermal Hammer" is not going to make it particularly visible is it?

Shame that the readers of the Air Vent haven't understood that the "missing stations" are due to back filling of station data, but I guess that would require a)reading the appropriate papers
b) dropping a much loved denier conspiracy theory

willard said...

But is it science? Hell no, but so what? may I ask rhetorically.

Take for example this quote taken from the **Roman's Hammer** post:

Several skeptics will dislike this post. They are wrong, in my humble opinion. While winning the public “policy” battle outright, places pressure for a simple unified message, the data is the data and the math is the math. We”re stuck with it, and this result.

(My bold.) So here's my suggestion as a sequel to this very interesting post: it's not science, but so what? Helping people understand what is a policy battle might help.

Mark said...

I set out to write a novel about global warming that employed the real science in the story and rebutted Crichton's State of Fear, which skewed the science. I did that but so far it hasn't sold. The publishing industry is timid of the implications of the story for the reader even as they praise the writing and the story. I have a sequel completed and will continue to try to sell both.

Penguindreams said...

willard:
I agree that science is science and policy is policy, except for the small area of probable overlap where it can be hard to say. On the other hand, the folks referenced in your quote are gaining their win precisely because they succeed in convincing people that things which are scientific questions are actually political ones. As in the quote, the data are the data, and the math is the math. But if you can persuade people to ignore that, and instead believe that "It's all just a matter of interpretation -- and those people are interpreting according to their (insert boogeyman) views." you need never address the science.

So that's one side -- as long as people fail to distinguish between science and policy, they'll be lead easily in to this sort of mistake.

The other side is that I do know a fair amount about the science, and can share that. I don't have an equivalently good knowledge of the policy and politics. Although it's probably the less important in terms of national policy, I stay with the science, where I know what I'm talking about.