Edward Dougherty Articles

[Ackbach]

I saw a link to an article by Edward R. Dougherty, a professor of electrical and computer engineering at Texas A&M, who apparently works in genomics and the cell. Then I noticed that on Public Discourse, he had a whole series of articles. I thought you all might enjoy them, as this guy can write. I just finished reading them all, and I liked them a lot. I'm not sure I always agree with him, but they're worth reading.

Unintelligibility: The Starting Point for Discussing the Science-Humanities Relationship

Scientific Education: Do We Love Our Children?

The Real War on Science

Science Contra Hubris

The Fundamental Issue for Climate Science

Science: What Went Wrong?

I'd be curious to hear what you think of them.

 

[MoreCoffee]

I saw a link to an article by Edward R. Dougherty, a professor of electrical and computer engineering at Texas A&M, who apparently works in genomics and the cell. Then I noticed that on Public Discourse, he had a whole series of articles. I thought you all might enjoy them, as this guy can write. I just finished reading them all, and I liked them a lot. I'm not sure I always agree with him, but they're worth reading.

Unintelligibility: The Starting Point for Discussing the Science-Humanities Relationship

Scientific Education: Do We Love Our Children?

The Real War on Science

Science Contra Hubris

The Fundamental Issue for Climate Science

Science: What Went Wrong?

I'd be curious to hear what you think of them.

Thanks for the links, I will click and read over the next few days, God willing.

 

[Lamb]

I'm starring this so I can come back and read some of the links.

 

[tango]

I read a couple of articles (it's late, much more and it will make my brain hurt) and thought they were very well written.

I did find this to be an interesting comment:

The fundamental issue for climate science, as well as many other sciences, is the epistemology of complex systems. Arguing about whether or not this or that observation validates or invalidates a theory is absurd when, as Tibaldi and Knutti correctly state, “projections . . . cannot be validated directly through observed changes.” It is equally absurd for people to claim that their belief in global warming is grounded in science. Instead of all the chatter, we should put forth great effort at formulating a viable epistemology within which the “truthfulness” of uncertain models can be characterized.

 

[Lamb]

Quote from Science: What Went Wrong

A scientist might find it odd that Rousseau’s ramblings have received much attention, since at the outset he rejects empirical observations and logic.

Gotta love Rousseau!

 

[Ackbach]

Quote from Science: What Went Wrong



Gotta love Rousseau!

Yeah, few people are as responsible for as many deaths (due to ideology) as he is.

 

[MoreCoffee]

It s not clear why some new epistemology is needed before one can decide if global warming is chiefly caused by human activities.

 

[Ackbach]

It s not clear why some new epistemology is needed before one can decide if global warming is chiefly caused by human activities.

I'm not entirely sure what you mean by that. It sounds like you're saying something along the lines of "Everyone knows global warming is caused by human activity." Am I correct? But that's precisely the point at issue in Dougherty's article. I'm not convinced the earth is warming, let alone that if it's warming, that warming is caused by human activity. Finally, I'm not convinced that global warming, if it is happening, is a catastrophic thing that we need to try to combat. That is, if you examine the phrase "catastrophic man-made global warming", and assert that that is what's happening, I would challenge every word in the phrase.

Causality is notoriously difficult to establish in science, and that's for any phenomenon you would care to name. In medicine, we practically know nothing at all about anything, given the extreme complexity of the human body. Sure, we have some correlations, some might even be strong (like smoking and lung cancer). But correlations do not imply causations. They can tell us where to look, but it's not evidence in any sense of the word. The earth's atmosphere is similarly complex, and causality just as difficult to establish. One of the central tenets of statistics is that you cannot establish causality by doing an observational study (the recording of data, and possibly performing calculations on that data). You must do an experiment to establish causality, and it must be the right one. So how are you going to manipulate environmental variables? Controlling the weather is a bit beyond us at the moment. Hence, causality is impossible to establish.

Mill's Methods are the standard for establishing causality. Think of them as the next layer of detail down from simply "You need to do an experiment". You can see from the description of Mill's Methods that you need to be able to insert variables and remove variables at will in order to make use of them at all. This is not possible in climate science, at least not at the earth scale.

Climate scientists essentially have two things: computer models, and some observational data (its veracity is sometimes disputed; some of the data acquisition methods in climate science do not line up with established methods elsewhere! To be clear: the data here is from observational studies only, not experiments). Models can be a great tool in science, no doubt. I've used them many times, both in science and in engineering. But they have many limitations:

1. They are only as good as their assumptions. You absolutely have to be front and center about your assumptions in any model. You list them first.
2. They are no substitute for an experiment. Real data from correctly designed experiments is extraordinarily valuable. Models are come-and-go.
3. The real value of a model is to save time in designing your experiment. It gives you a good idea where to look, since exhaustive experimentation is rarely practical or even available.

Everything I've said above regarding how you establish causation would be readily admitted by any statistician. This is the way you do it in science. And this method is quite simply not available in climate science.

 

[MoreCoffee]

What I said is that we need no new epistemology to decide if climate change is due chiefly to human activity or not. Existing theories of knowledge are adequate for the task. The scientific method - predicated on existing epistemology - is sufficient to decide the matter.

 

[Ackbach]

What I said is that we need no new epistemology to decide if climate change is due chiefly to human activity or not. Existing theories of knowledge are adequate for the task. The scientific method - predicated on existing epistemology - is sufficient to decide the matter.

Ok, I see what you mean. But I've just finished arguing that the scientific method is not sufficient for the task, since we can't do the right experiments. Therefore, causality cannot be shown. It's kinda like questions of origin. We can't generate new universes, evolve them in time, and see how they come out.

 

[MoreCoffee]

Ok, I see what you mean. But I've just finished arguing that the scientific method is not sufficient for the task, since we can't do the right experiments. Therefore, causality cannot be shown. It's kinda like questions of origin. We can't generate new universes, evolve them in time, and see how they come out.

The kind of experiment you propose is impossible for human beings at this time. That does not mean that the scientific method cannot be applied to the question of life-origins, universe origins, or climate change. There are other experiments that can be performed to differentiate between models. These are being done. They point to fossil fuel burning with CO[SUB]2[/SUB] production as the primary cause for increasing amounts of CO[SUB]2[/SUB] in the atmosphere. The amount of CO[SUB]2[/SUB] in the atmosphere appears to be causal for heat retention, temperature increases, and climate change.

 

[Ackbach]

The kind of experiment you propose is impossible for human beings at this time. That does not mean that the scientific method cannot be applied to the question of life-origins, universe origins, or climate change. There are other experiments that can be performed to differentiate between models. These are being done. They point to fossil fuel burning with CO[SUB]2[/SUB] production as the primary cause for increasing amounts of CO[SUB]2[/SUB] in the atmosphere. The amount of CO[SUB]2[/SUB] in the atmosphere appears to be causal for heat retention, temperature increases, and climate change.

But in the end, they're just models. They're only as good as their assumptions (sometimes not even that good). A model without experimental data to back it up is completely unable to establish causality. This is why string theory has come up on such hard times recently: we don't have the technology to look at things that small. Some physicists (it seems to me and others) want to redefine the rules of science to allow string theory to have equal status with, say, General Relativity. At the moment, there's no way that can happen. GR has well-attested experimental data to back it up; string theory has none whatsoever.

When it comes to climate science, in particular, I'm also extremely disappointed at how politicized the issue has become. If you can get a politically liberal climate scientist to agree with a politically conservative climate scientist on an interpretation of data, or anything at all, then I might start to take it seriously. At the moment, though, the rather obvious bias of both sides is not honestly being admitted. The liberals sometimes seem to want control over people's lives, and they think that burdensome regulations about the environment are one way to do that. So they definitely have a non-scientific motive to conclude in favor of catastrophic man-made global warming. The conservatives sometimes seem to want simply more personal spending money, and so, at the cost of the environment (possibly), they argue against catastrophic man-made global warming.

Politically, I'm not completely sure what I am anymore. I am an originalist when it comes to interpreting the US Constitution. I have zero respect for deconstruction literary theories, or any kind of "living document" ideas. I believe that a republic is the best system of government ... for people who can govern themselves. As John Adams said, "Our Constitution was made only for a moral and religious people. It is wholly inadequate to the government of any other." Well, we don't have a moral or religious people. So, what does that say about our republic? We are ripe for a dictator - not that I want one. I'd much rather we recover our self-governance, at which point the extreme over-governing of the current US government would be superfluous. But only the gospel can achieve this. I say this by way of explaining my inherent bias.

So, if you have no causality, what about correlations? You can actually prove results that say if you compute enough correlations with enough random data, you can find a statistically significant relationship. That doesn't mean you've found anything truly significant at all.

 

[MoreCoffee]

But in the end, they're just models. They're only as good as their assumptions (sometimes not even that good).

While your statements are true in a sense they are also smoke and mirrors that add nothing to the specifics of climate change. The models are "just models" but that is true of every human construct designed to explain experimentally observed facts - in the physical sciences explanation is synonymous with model - thus the models of the climatologists are scientific theories proffered to explain observed changes in the earth's climate since the beginning of industrial production of CO[SUB]2[/SUB] .

Here's why I think your opening statements and those that follow (quoted below) serve to obscure the matter that is under debate here. Climate models offer concrete predictions that can be tested while a philosophical discussion concentrated on epistemological models of cause/effect offer not concrete predictions in climatology. This obscures rather than clarifies the question of accuracy of predictions in climatological models. Below are some considerations that ought to be on the table for this discussion.

• Climate models are explanations (some better than others) that offer predictions based on past and current observations of the facts of measurable climate features (temperature, precipitation, air pressure, atmospheric composition, humidity, sea surface temperatures, rates of evaporation, storm frequency, storm severity, and so forth) and since they offer predictions they can be tested by checking how accurate (or inaccurate) the predictions are.
• Models/explanations are critiqued and tested by experiments and observations designed to see if the predictions inherent in a particular model can withstand serious critical examination.
• Accumulating observational data and checking predicted climate events against observed data enables refinement of models subjected to rigorous criticism based on observed data.

Your post argues about causality (a philosophical perspective rather than a specifically scientific one) without grappling with the more precise issue of correspondence between predicted events in specific models of climate change and actual measured events. It does not matter if rigorous philosophical causality is proved or not if the predicted events of a specific model correspond with observed events in the time frame predicted by the model. If the model posits that increased CO[SUB]2[/SUB] in the composition of the atmosphere corresponds to certain climate events of a harmful kind and later observation correspond with the predictions then from a scientific point of view the specific model that made the predictions is a viable scientific theory for predicting climate change consequences of increasing production of CO[SUB]2[/SUB] by human beings.

A model without experimental data to back it up is completely unable to establish causality. This is why string theory has come up on such hard times recently: we don't have the technology to look at things that small. Some physicists (it seems to me and others) want to redefine the rules of science to allow string theory to have equal status with, say, General Relativity. At the moment, there's no way that can happen. GR has well-attested experimental data to back it up; string theory has none whatsoever.

When it comes to climate science, in particular, I'm also extremely disappointed at how politicized the issue has become. If you can get a politically liberal climate scientist to agree with a politically conservative climate scientist on an interpretation of data, or anything at all, then I might start to take it seriously. At the moment, though, the rather obvious bias of both sides is not honestly being admitted. The liberals sometimes seem to want control over people's lives, and they think that burdensome regulations about the environment are one way to do that. So they definitely have a non-scientific motive to conclude in favor of catastrophic man-made global warming. The conservatives sometimes seem to want simply more personal spending money, and so, at the cost of the environment (possibly), they argue against catastrophic man-made global warming.

Politically, I'm not completely sure what I am anymore. I am an originalist when it comes to interpreting the US Constitution. I have zero respect for deconstruction literary theories, or any kind of "living document" ideas. I believe that a republic is the best system of government ... for people who can govern themselves. As John Adams said, "Our Constitution was made only for a moral and religious people. It is wholly inadequate to the government of any other." Well, we don't have a moral or religious people. So, what does that say about our republic? We are ripe for a dictator - not that I want one. I'd much rather we recover our self-governance, at which point the extreme over-governing of the current US government would be superfluous. But only the gospel can achieve this. I say this by way of explaining my inherent bias.

So, if you have no causality, what about correlations? You can actually prove results that say if you compute enough correlations with enough random data, you can find a statistically significant relationship. That doesn't mean you've found anything truly significant at all.

 

[Ackbach]

While your statements are true in a sense they are also smoke and mirrors that add nothing to the specifics of climate change. The models are "just models" but that is true of every human construct designed to explain experimentally observed facts - in the physical sciences explanation is synonymous with model

Not the way I was using the term. In climate science, there are a lot of computer models. I'm definitely not talking about $\displaystyle\mathbf{F}=\frac{d\mathbf{p}}{dt}$ or $E=m c^2$ as a model. Climate science does not have such well-defined equations that explain everything. Far less is known about how the computer models work than about Newton's Second Law or Einstein's Theory of Relativity. And while it is true that computer models can play a role in a scientific theory, a theory is considerably more comprehensive than just the computer models.

- thus the models of the climatologists are scientific theories

See above.

proffered to explain observed changes in the earth's climate since the beginning of industrial production of CO[SUB]2[/SUB] .

Here's why I think your opening statements and those that follow (quoted below) serve to obscure the matter that is under debate here. Climate models offer concrete predictions that can be tested

But not in any experiment: only in observational studies. It is a commonplace in statistics that to establish causality, you must perform an experiment. Observational studies and correlations will not get you there.

while a philosophical discussion concentrated on epistemological models of cause/effect offer not concrete predictions in climatology.

Well, I wasn't trying to offer concrete predictions in climatology. I am not a climate scientist.

This obscures rather than clarifies the question of accuracy of predictions in climatological models. Below are some considerations that ought to be on the table for this discussion.

• Climate models are explanations (some better than others) that offer predictions based on past and current observations of the facts of measurable climate features (temperature, precipitation, air pressure, atmospheric composition, humidity, sea surface temperatures, rates of evaporation, storm frequency, storm severity, and so forth) and since they offer predictions they can be tested by checking how accurate (or inaccurate) the predictions are.
• Models/explanations are critiqued and tested by experiments and observations designed to see if the predictions inherent in a particular model can withstand serious critical examination.
• Accumulating observational data and checking predicted climate events against observed data enables refinement of models subjected to rigorous criticism based on observed data.

Let me repeat: there are no experiments available here. There are observations, sure. But one of the first principles of statistical design (and all of science is based on statistics - statistics is absolutely central to all science) is that you cannot establish causality - you cannot answer the question "Why?" with an observational study at all. You absolutely have to do an experiment.

Predictions are certainly part of a scientific theory, but a scientific theory should be broader than that: it should explain the why. The computer models can never do that.

Your post argues about causality (a philosophical perspective rather than a specifically scientific one)

I beg to differ. I am, by training, a mathematical physicist (Ph.D. from Virginia Tech, 2007), and I can assure you that causality is central to the scientist's thinking. This is what drives most scientists: to know the why!

without grappling with the more precise issue of correspondence between predicted events in specific models of climate change and actual measured events.

The reason I did not grapple with that issue is because I see it as secondary to these more fundamental statistical issues (at their heart, I think the problems are statistical; I would not object to calling the issues epistemological, either).

It does not matter if rigorous philosophical causality is proved or not if the predicted events of a specific model correspond with observed events in the time frame predicted by the model.

It does matter, because if there is no understanding of the why, then there is zero confidence that the model will work in cases when the conditions are even slightly different, much less radically different.

If the model posits that increased CO[SUB]2[/SUB] in the composition of the atmosphere corresponds to certain climate events of a harmful kind and later observation correspond with the predictions then from a scientific point of view the specific model that made the predictions is a viable scientific theory for predicting climate change consequences of increasing production of CO[SUB]2[/SUB] by human beings.

Well, as I mentioned before, the model is not synonymous with a theory. A model is a subset of a theory. Theories should absolutely have models, but they should have a lot more than that.

I have to say, it's a pleasure debating with you: you stick to issues, and don't get side-tracked by ad hominem fallacies. Quite refreshing! I'd like to think I do the same - please point it out to me if I don't.

 

[MoreCoffee]

Ackbach, experiments ARE observational studies. And experiment is observation of what happens in a system under repeatable stable starting conditions. Experimental results are analysed using statistical methods. Each result is recorded and after a statistically significant number of observations are recorded and analysed the observed results are compared with what the equation (model) predicted. If the observations fall within statistical error limits the model (theory) is regarded as supported by observed experimental results.

Climate models are complex compared to Newtonian laws of motion or thermodynamic laws. That is to be expected because climate is inherently more complex than F=ma. Complexity does not imply unreliability despite Occam's Razor.

 

[Ackbach]

Ackbach, experiments ARE observational studies...

The difference between experiments and observational studies is this: in an observational study, you do not control any variables. In an experiment, you do. That's it. This is an accepted difference in any statistics book you examine. In an observational study, you simply look at what is happening and record it. You do that also in an experiment, but in addition, you control variables ahead-of-time. A large part of any experimental design is figuring out how you will control your variables. Sometimes, especially in medicine, observational studies are all you have, since experiments are unethical.

It is precisely this control over variables that you have in an experiment, and that you don't have in an observational study, that allows you to infer causality in an experiment and not in an observational study.

...Climate models are complex compared to Newtonian laws of motion or thermodynamic laws. That is to be expected because climate is inherently more complex than F=ma. Complexity does not imply unreliability despite Occam's Razor.

True, but the form of the climate models implies that far less is known about them. One example is neural networks. Neural networks are great for certain applications, but they are a black box: you do not know precisely what is happening under the hood. There's not a nice, neat set of analyzable equations providing a comprehensive framework of interpretation for data.

If you spend a lot of time fitting a model to a particular set of data, you have no confidence that your model will fit the next set of data, because you have no why. It's rather like the Merovingian in the Matrix movies: in science, the power is in the why. If you know the why, then you've arrived.