Dangerous Models

You know what I think about when I hear about the epic failure of all these fancy financial models that were designed to calculate risk? I think about the Atlantic Cod. These fish used to be everywhere. (Once upon a time, they were considered the cash crop of the ocean. Spanish fishing vessels would trek across the Atlantic just to fish the abundant cod off the coast of Canada.) Now the Newfoundland cod fishery is gone, yet another victim of overfishing.

The story of cod is usually told as the tragedy of trawlers. A trawler is boat designed to drag a massive net behind it. These nets are weighted, so that they cling to the bottom of the ocean floor. They sweep up everything for miles and miles. Most of the haul is trash - trawlers leave a trail of dead, unwanted fish - but they can also capture thousands of cod in a single haul. The use of radar made these trawlers even more efficient; now they knew exactly where to drop their nets. The result was a boom in caught cod: by the late 1960's, fishermen were hauling in more than 800,000 tons of cod every year.

But trawlers aren't entirely to blame. Their catch was still within the legal limits. (Cheating, of course, remained a big problem. Many fishing boats caught way too many fish, just as fraudulent lending helped implode the subprime market.) In fact, the Canadian government had been concerned about the cod population for decades. In the 1970's, the government instituted strict regulations that limited the total catch to just 16 percent of the total cod population. The tricky part, of course, was coming up with the population estimates in the first place. It's hard to know how many fish to catch if you don't know how many fish there are. But fishery scientists were confident that their sophisticated models were accurate. They had randomly selected areas of the ocean to sample and then, through the use of a complicated algorithm, arrived at their total estimate of the cod population. They predicted that the new regulations would allow the cod stock to steadily increase. Fish and the fishing industry would both thrive.

The models were all wrong. The cod population never grew. By the late 1980's, even the trawlers couldn't find cod. It was now clear that the scientists had made some grievous errors. The fishermen hadn't been catching 16 percent of the cod population; they had been catching 60 percent of the cod population. The models were off by a factor of four. "For the cod fishery," write Orrin Pilkey and Linda Pilkey-Jarvis, in their excellent book Useless Arithmetic: Why Environmental Scientists Can't Predict the Future, "as for most of earth's surface systems, whether biological or geological, the complex interaction of huge numbers of parameters make mathematical modeling on a scale of predictive accuracy that would be useful to fishers a virtual impossibility."

People love models, especially when they're big, complex and quantitative. Models make us feel safe. They take the uncertainty of the future and break it down into neat, bite-sized equations. But here's the problem with models, which is really a problem with the human mind. We become so focused on the predictions of the model - be it the cod population, or the risk of mortgage derivatives - that we stop questioning the basic assumptions of the model. (Instead, the confirmation bias seeps in and we devote way too much mental energy to proving the model true.) It's not just about black swans or random outliers. After all, there was no black swan event that triggered this most recent financial mess. There was simply an exquisite model, churning out extremely profitable predictions, that happened to be based on a false premise. Hopefully, the markets will recover quicker than the Atlantic cod.

Categories

More like this

I often dislike models, but for different reasons than what you listed above. Much of my thesis work was devoted to modeling release of drug from polymer coatings -- different application, same idea.

>>People love models, especially when they're big, complex and quantitative. Models make us feel safe.

>>(Instead, the confirmation bias seeps in and we devote way too much mental energy to proving the model true.)

As you are probably aware, it's not possible to prove a model to be true or false, although evidence can be accumulated that either supports or refutes aspects of the model. Models, like any scientific theory, need testing. The real problem comes in when researchers don't ask the right scientific question when they test the model. Is the researcher looking for a model that accurately reflects an underlying mechanism (the cod model probably did) or is the researcher interested in an accurate prediction of the output (the cod model didn't)? Often we have the opposite problem: a model can be constructed such that it produces incredibly accurate predictions, but the underlying mechanism has never been demonstrated, and so parameter extraction from those sorts of models is a shaky science.

I think one of the biggest problems in current-day modeling is a failure to recognize that models must be tested. "Testing" does not mean "check whether my model predicts my output" or "verify the exact value of my parameters". "Testing" means that experimental conditions must be varied and the output must be measured (which output you measure depends on whether you're looking for an accurate mechanism or an accurate prediction). If the cod-modelers wanted to convince me of anything, they would have had to try fishing at a certain percentage for 5 years, check the population growth/decline, change the fishing percentage and check again.

Of course nobody would ever do that. But that's my point. Even though I'm in the business of modeling, and I think modeling is an extremely important scientific tool, we could all be a little more critical of the methods by which people generate models.

Let me digress with a story. I recently attended an epidemiology talk where the speaker demonstrated that her data could be fit well by four models. Three models where the same model with slightly different assumptions about the start and end points of the data. The fourth model was empirical (using a SPLINE for goodness' sake). All four models were used to estimate one parameter. The speaker -- who is a leader in the field, with many awards and publications to their name -- went so far as to conclude in writing, with a bolded, fun-colored powerpoint note, that "The use of four models confirms that this phenomenon is real." My jaw almost dropped.

I could fit any data you give me with an unlimited number of fancy models. For many of the models that I generate, I could give logical explanations of how they were constructed. I could ascribe their operators to real phenomenon. I could probably rationalize the parameter estimates. All of these activities can be performed for models that do not reflect the physics of the system or produce accurate predictions. It is to the detriment of good science that the purpose and testability of models is so widely misunderstood.

(By the way, it's ironic -- though only to me -- that you posted this today. I'm working on a manuscript that details the problems inherent in particular sorts of modeling. I haven't worked on it in more than a month, and this morning, I opened it up for what will be its final edit.)

Essentially, all models are wrong, but some are useful.
--George E.P. Box

I am glad there are people out there like Rachael who live to discern which are useful and why; no easy task.

Colin, what a great quote

Here's another (not as relevant, but amusing):

"The best material model of a cat is another, or preferably the same, cat"
A. Rosenblueth, Philosophy of Science, 1945

(My apologies in advance)

Models coddle with rose-colored goggles, making people feeble to the risks of greed-evils.

Sorry...couldn't resist...

Models are useful to help describe physical phenomena, especially based on the current laws of physics and your understanding of the phenomena. When testing suggests your model is wrong, you try to explain it physically.
Some predictive models are based entirely on past correlations with zero or little understanding of the basis for that model, in effect an extrapolated curve fit of some behavior, something where the "science" is weak at best. The financial market, never factoring in greed and fraud, is a prime example today. Also, garbage in, garbage out is probably the most reasonable cause of the fishery problem not meeting expectations. And no one talks of the confidence bounds of their predictions, especially in the pseudo sciences.

By sniper609 (not verified) on 28 Oct 2008 #permalink

As I remember the events, fishery scientists were consistently recommending lower catch limits for the cod, but the Fisheries Minister (John Crosby for at least part of that time) was over-ruling them and allowing higher limits, on the grounds that the proposed limits would cause hardship to the people of the area. I do not recall any of the fishers or their representatives agreeing with the scientists' estimates of the stocks so perhaps he could do little else and still keep his seat.

By Richard Simons (not verified) on 28 Oct 2008 #permalink

Excellent post. We seem to invest energy in models the way we do with 'experts' - in an effort to be able to rely on someone, something outside ourselve to make everything Ok.

Criticisms of finanical models apply also to our models of global warming. We have no reason to assume that they are accurate. The problem may be worse or better than we believe. While the data we're feeding the models is probably correct, we understand only parts of the underlying mechanism.

BTW, I think reducing greenhouse gases would be an excellent idea, along with a general reduction in pollution. I'm just pointing out that we left-wing tree huggers need to be realistic about the models that seem to support our views. Although if Jonah is correct about the human inclination to discount information that clashes with our views, we probably won't. :)

As a mathematician, I think that what you're seeing is an example of "the false certainty of numbers."

If you tell people that you're pretty sure about something, then they'll doubt it. But if you attach a number to your hypothesis -- preferably with a lot of extra and unjustified precision -- they'll believe you as unhesitatingly as seven-year-olds at their first communion.

The failure at the heart of many models is the ludic fallacy, that is, assuming that components of real life will correlate in accordance with a given model under controlled assumptions. Originators or proponents of a model may then develop an ownership bias which favors a belief in its predictive power. Enter Karl Popper ...

So then, are we all ready to disavow the models which support the theory that burning fossil fuels causes global warming? Probability alone should tell us that they couldn't possibly be correct, and likely not useful. The arguments above are just one more argument against.

'Probability alone should tell us that they couldn't possibly be correct'.

LOL. Good one. Game, set and match. Fire up those coal plants now!

Are we all ready to abandon the models which support the theory that burning fossil fuels does not cause global warming? Probability alone should tell us that these models couldn't possibly be correct, and probably not useful.

I'm not sure I understand, Alan. You rewrote John's post a little...but the message is still the same. Are you a bot?

I'm reminded of David Ruelle's book, Chance and Chaos. In the chapter on probabilities he points out the need for a physical theory of probabilities to give a prediction some operational meaning. Referring to 'purist' (frequentists?) who think a statement like 'the probability it will rain this afternoon is .9' is meaningless, he says:

One might, however, be able to give it meaning, for instance by making a large number of numerical simulations on a computer (compatible with out present knowledge of the meteorological situation) and finding the proportion of cases in which the simulation gives rain. If one finds a probability of 90 percent for rain, even the purists will take their umbrellas.

Beyond the issues of testing and trusting models, the example of cod fishing would point to the problem being less about the validity of particular predictions but in what actions we use those predictions to justify. There's a very interesting book called 'Making Better Environmental Decisions' by Mary O'Brian where she makes the case that the ability (claim?) of risk assessors to accurately determine the risk of various alternatives unnecessarily narrows our decision making process to those alternatives.

In the article it says:
"After all, there was no black swan event that triggered this most recent financial mess."

I haven't read the Black Swan, but I was under the impression that for the financial models used to calculate risk the fall in the housing market in the US was a 'black swan event'. I think Taleb's point is not that meteorites may hit the earth or that extraordinary events may happen, but that the supposedly 'smart' models most risk managers use are based on very limited data and don't consider many events that Jane-the-sensible-banker, with her 'common sense' and millions of years of mammalian paranoia, would probably lose sleep over.

The cod story is totally wrong. Fisheries scientists have been trying to get the quotas lowered for years, but the government wouldn't do it because of the job losses it would entail. I don't know of any fisheries models that have been predicting stock increases for the last 50 years. They all predict decreases.

By Jordan Dawe (not verified) on 30 Oct 2008 #permalink

"I'm not sure I understand, Alan. You rewrote John's post a little...but the message is still the same. Are you a bot?"

Mine were the comments on the ludic fallacy, not on global warming, which appear to be copied from "John." Maybe there's (1) a bot on the loose, (2)somebody spoofing, or (3) someone else is using the same nom de plume as me.

Haven't read the Pilkeys' book, but I'm suspicious of their argument, since it's widely known in Canada that political factors led the government to keep allowing overfishing of the Grand Banks.

One would also have to question whether government fisheries scientists, if indeed they were overoptimistic, were unconsciously or otherwise slanting their modeling in the direction both their masters (the government) and their clients (the fishing industry) clearly wanted to hear.

One of the basic assumptions of any model that involves human interaction with anything should be that politics are always involved.

There were certainly a couple of fisheries scientists in the Canadian DFO (Department of Fisheries and Oceans) who more or less predicted the collapse of the cod fishery and caused great displeasure to their government bosses. A relevant quotation:

"In 1997 two former DFO scientists published an article entitled, "Is Scientific Inquiry Incompatible with Government Information Control?" Their report and others cited a pattern of suppression of scientific information at DFO.

The authors cited numerous examples where DFO scientists had warned the Minister that ground fish stocks were in a dangerous decline, and these findings were either ignored or suppressed as high quotas continued to be allocated. In one of these instances, a DFO scientist named Ransom Myers was apparently threatened with termination of his job when he concluded that the true cause of the cod collapse was simply human over-fishing rather than predation by seals."

More information on this here: http://xrl.us/oxkuc

The authors cited numerous examples where DFO scientists had warned the Minister that ground fish stocks were in a dangerous decline, and these findings were either ignored or suppressed as high quotas continued to be allocated.

Thanks for the job ..

here were certainly a couple of fisheries scientists in the Canadian DFO (Department of Fisheries and Oceans) who more or less predicted the collapse of the cod fishery and caused great displeasure to their government bosses. A relevant quotation:

"In 1997 two former DFO scientists published an article entitled, "Is Scientific Inquiry Incompatible with Government Information Control?" Their report and others cited a pattern of suppression of scientific information at DFO.

Some predictive models are based entirely on past correlations with zero or little understanding of the basis for that model, in effect an extrapolated curve fit of some behavior, something where the "science" is weak at best. The financial market, never factoring in greed and fraud, is a prime example today.

You draw a great comparison in layman's terms, I am not mathematically inclined, so it was appreciated. Questioning drives Science forward, and I think that would be a good approach to these models. Question and test everything before you call it absolute.

Questioning drives Science forward, and I think that would be a good approach to these models. Question and test everything before you call it absolute.

ome predictive models are based entirely on past correlations with zero or little understanding of the basis for that model, in effect an extrapolated curve fit of some behavior, something where the "science" is weak at best.

Personally I'd have to say that I agree, it's obvious that you know what you are talking about.I'll be sharing this with Digg, it's something that's gotta be Dugg for sure.I'd recommend writing additional posts like this one as it's one of your best.I'll check back often, I'm glad I came across your site.

Cod is into the line of extinction? This is bad. I think people especially fishermen who uses trawlers must be warned about this. And the authorities should implement a strict banning of this issue.

i have checked this blog a few times now and i have to say that i find it quite exeptional actually. keep the nice work up! :)

keep up the great work on the blog. I love it. Could use some more frequent updates, but i'm sure that you got some better or other things to do like we all do. =)

keep up the good work on the blog. I kinda like it! :) Could use some more frequent updates, but i am quite sure that you got more or better stuff to do like we all have to do unfortunately. :)

continue with the the great work on the blog. I love it. Could use some more frequent updates, but i am sure that you got more or better things to do , hehe. :)

halten Sie die groÃartige Arbeit auf dem Blog. Ich liebe es. Könnte etwas häufigere Updates, aber ich bin sicher, dass du etwas Besseres oder andere Dinge zu tun wie wir alle.

You mean the way that the geological record contains evidence of fairly dramatic climate changes on decadal timescales, in response to natural forcings.. or the way that any model that successfully predicted such a change would be culled as unrealistic?

Its such as you read my mind! You appear to understand so much about this, like you wrote the ebook in it or something. I believe that you could do with some percent to pressure the message house a little bit, but other than that, that is wonderful blog. An excellent read. I'll definitely be back.

Wow! I can say, your blog should be one of the better written blogs that I have viewed from a number of years.Things i wouldn't share with have the capacity to create posts that will be as fascinating| as yours. I guess I'll have to read on yours and pray that some day I am able to write on the subject with the maximum amount of knowledge on your own! Bravo!!

Zuerst spielte ich mit meiner Zunge an seiner Eichel... dabei schmeckte ich seine Lusttropfen, die er zuvor versprüht hatte. Ich leckte ihm den Schwanz so richtig ab... von der Eichelspitze bis zum Sack.

Britain has sent a crew to Haiti that will help with rescue and relief efforts. Gordon Brown has appealed to the Brtish public to donate cash to the relief energy.

Thank you, I have been hunting for information about this subject matter for ages and yours is the best I have discovered so far.

I've learned newer and more effective things from a blog post. Also a thing to I have noticed is that in many instances, FSBO sellers will probably reject anyone. Remember, they will prefer never to use your providers. But if you actually maintain a steady, professional romance, offering aid and keeping contact for around four to five weeks, you will usually be capable to win a meeting. From there, a house listing follows. Thanks a lot

I have been doing research a lot and after skimming trough this all I could say is "finally somebody that writes instead of obsessing #with his# blog info". bravo. You only made the internet a better place after half the people online only doing this to monetize every word.

Hi there, I found your web site via Google while looking for a related topic, your web site came up, it looks great. I've bookmarked it in my google bookmarks.

The opportunity here doesnt get bigger, the biggest of our lifetimes. We wont get an opportunity like this again to save America. Its all or nothing the Exponential Curve of Fascist Unconstitutionality leaves us no choice.

Can I just say what a aid to search out somebody who truly knows what theyre talking about on the internet. You undoubtedly know tips on how to carry a problem to light and make it important. Extra folks must learn this and perceive this aspect of the story. I cant imagine youre not more popular because you positively have the gift.

These days you truly cannot find one thing of such a high quality. That's the reason I tried to add the Rss... Yet hit a brick wall totally... May anyone support a beginner?

I have decided that i love you :]

By Reiko Devilla (not verified) on 26 Aug 2011 #permalink

Lease to have residence can feel like the best of all possible worlds and frequently it is, when you are the buyer there are lots of things you should be aware of prior to signing any contract. A few of them tend to be comprised inside the documents you will be completing. Other people of them refer to the home themselves.

I recall one post-Andrew community specifically not permitting storm shutters (even on homes for elderly people) because said shutters wouldn’t meet local architectural codes. I watched a TV interview of two bastards from that community (the city architect and the code enforcer, as I recall) smugly explain that “the elderly can always get help in putting up plywood”. Hopefully common sense has prevailed since then. Or the voters.

That is a important idea, I am very glad to read this write-up, and I approve with the points of this script. I think this is the best opinion. I like to do something after reading it.

By Anonymous (not verified) on 05 Sep 2011 #permalink

I was very pleased to find this website. I wanted to thank you for your time for this wonderful post!! I definitely enjoy reading it and I have you bookmarked to check out new stuff you blog post.

You're not going to believe this but I have wasted all night looking for some articles about this. I wish I knew of this site earlier, it was a good read and really helped me out. Have a good one

One of the very few 'professional' actresses in Bollywood. With a good bit of luck and networking, she might get the chance to act in British or American cinema. Otherwise, with her potential, she will get frustrated like Sridevi did in the end after countless of masala/flicks/talent crushing scripts. God knows there are lots of them in Boringwood nowadays...

The older cars and trucks had a certain style as well as a lot more individuality I actually actually like the old cars a lot far better.You have done a nice job here.

By Union Auto Sales (not verified) on 14 Sep 2011 #permalink

I have really learned newer and more effective things as a result of your website Dangerous Models : The Frontal Cortex. One other thing I would really like to say is the fact newer computer operating systems are inclined to allow more memory to be utilized, but they additionally demand more storage simply to work. If one's computer can not handle additional memory and the newest program requires that ram increase, it may be the time to shop for a new Computer system.

Before the trolls even start, (you know who you are), I purchased the GoFlex Satellite for my the wife and I (two 32GB Wifi iPads) and using it on our trip to Las Vegas was great but the one gripe I have is having to use the GoFlex Media app. Anyone know of a app that works that I can use while at home (while my wife watches Housewives of [fill in the blank]

It looks like too complicated and very comprehensive for me. I am looking forward for your another script, and I would like to try to gain it! In fact the publication is spreading its wings quickly, looking for it...

We are a group of volunteers and starting a new scheme in our community. Your web site on Dangerous Models : The Frontal Cortex offered us with valuable info to work on. You've done a formidable job and our entire community will be grateful to you.

Yes, David Ls dismissive attitude towards Russia is not warranted, but it is true that, but for a brief few years at Israels birth, the Soviet Union, and to a very slightly lesser extent todays Russia, has assumed the role of implacable enemy to Israel, and has a long history of anti-Semitism, from pogroms, to purges, to paranoid executions under Stalin, to exclsuion from full rights, to suppression of religious expression. There is no arguing with this. The Soviets KGB literally invented the PLO and set up propaganda front groups all over the world to deligitimize Israel; the Soviets armed the entire Arab world in order to support their invasions of Israel; todays Russia still arms and helps develop the conventional and nuclear arms capability of any and all Muslim, Arab, or even simply authoritarian states, all of whom coincidentally desire the destruction of Israel. Some of this is just to disadvantage Israels only true powerful ally the United States some to be the spoiler, some to accrue economic benefits in oil price hikes, some to just be seen as important again, but some of this is the same old anti-Semitism, dressed anew.

It may be too complex and very inclusive for me. I am looking forward for your the other script, and I will try to acquire it! In fact the write-up is spreading its wings quickly, hunting for it...

Though it is becoming much passed essay, but it really compensated for my time. I plan to post a link to this page on my internet site. I am sure my visitors may find that very useful.

I dicovered this is an helpful and funny essay, so I think it is very benificial and knowledgeable. Appreciate for the efforts you have made in writing this article. I am wishing the same excellent work from you next time as well. practically your creative writing ability has inspired me.

Thank you for helping to consider this thoughts and opinions, I feel strongly about issues and I am willing to learn loads of things on this factor. Probably, as you attain knowledge, would you care renewing your internet site with loads of more info? Itâs very good for me.

I generally approve with your points. We all benefit from this good script. This blog is fantastic. I have figured out a little things from here. Many thanks.

Auto Transport

By Tommie Saravia (not verified) on 11 Nov 2011 #permalink

Excellent goods from you, man. Dangerous Models : The Frontal Cortex I have understand your stuff previous to and you are just extremely wonderful. I actually like what you have acquired here, certainly like what you are stating and the way in which you say it. You make it entertaining and you still care for to keep it smart. I can't wait to read much more from you. This is actually a wonderful Dangerous Models : The Frontal Cortex informations.

Great goods from you, man. Dangerous Models : The Frontal Cortex I've understand your stuff previous to and you are just too magnificent. I really like what you have acquired here, really like what you are stating and the way in which you say it. You make it entertaining and you still take care of to keep it smart. I can not wait to read far more from you. This is really a great Dangerous Models : The Frontal Cortex informations.

I dicovered this is a benificial and interesting write-up, so I think it is very useful and knowledgeable. Thank you for the efforts you have made in writing this script. I am expecting the similar important work from you in the future as well. Really your creative writing ability has encouraged me.

You speak with so many recommendations, so much spirit, although I become aware that you've definitely hit the nail on the head. Well done! I will immediately grab your rss feed to stay abreast of some updates. Real work and much success in your business efforts!

Hello clever points.. now why didn't i consider those? Off subject barely, is this web page pattern merely from an ordinary installation or else do you use a personalized template. I use a webpage i'm seeking to enhance and well the visuals is probably going one of many key issues to finish on my list.

I have read a few of the articles on your website now, and I really like your style of blogging. I added it to my favorites blog site list and will be checking back soon. Please check out my site as well and let me know what you think.

Actually, the topic is actually the sweetest on this benificial submission. I coordinate with your sum and will eagerly await your next updates. Saying thanks will not just be ample, for the great clarity in your essay.

Pretty good post. I have really enjoyed browsing your blog posts. Whatever the case I'll be subscribing to your feed and i also hope you write again soon! Thanks a lot, I'll try to come back often. Merry Chrismas!

I could not discover too much unequal information on this clot of substance, so it was good to find his one. I will return again to scan another posts that you have next time.

Excellent work as always. What I really like about your series are the fact that you are always focused and you bring a point home. Thank you.

I could not discover too much not the same information on this clot of content, so it was easy to discover his one. I will return again to look at next articles that you have next time.

Excellent goods from you, man. Dangerous Models : The Frontal Cortex I've understand your stuff previous to and you are just extremely excellent. I really like what you have acquired here, really like what you are stating and the way in which you say it. You make it enjoyable and you still care for to keep it wise. I can't wait to read much more from you. This is really a tremendous Dangerous Models : The Frontal Cortex informations.

Thank you for sharing superb informations. Your site is very cool. I am impressed by the details that you've on this website. It reveals how nicely you understand this subject. Bookmarked this web page, will come back for extra articles. You, my friend, ROCK! I found just the information I already searched everywhere and just could not come across. What a perfect website.

I couldn't locate too much various information on this clot of contents, so it was great to find his one. I will be back again to overlook the other posts that you have next time.

I admit, I have not seen this page for a long time, however, it was another enjoyment to see such an great issues and ignore it. Thank you for assisting making people more aware of great notions.

Magnificent goods from you, man. Dangerous Models : The Frontal Cortex I've understand your stuff previous to and you're just extremely wonderful. I actually like what you've acquired here, really like what you're stating and the way in which you say it. You make it entertaining and you still care for to keep it wise. I can not wait to read much more from you. This is really a wonderful Dangerous Models : The Frontal Cortex informations.

I could not show too much not the same information on this piece of substance, so it was great to locate his one. I will return again to look at some other scripts that you have next time.

It looks like too sophisticated and very far-reaching for me. I am looking forward for your another write-up, and I will try to obtain it! Practically the posting is spreading its wings fast, searching for it...

Wow, awesome blog page layout! How long have you ever been writing a blog meant for? you've made running a blog glimpse convenient. The whole seem on your site is wonderful, child this content!

It looks like too complex and very far-reaching for me. I am awaiting for your another write-up, and I am going to try to get the hang of it! Practically the post is spreading its wings swiftly, hunting for it...