The Absurdity of "Science's" Defenders

The tagline for this blog is "Thoughts on This Insane World." I don't emphasize it that often, but I believe the world is completely and utterly insane. I often feel like a battered wife, frightened and rendered helpless by a seemingly schizophrenic world able to function because of a thin mask of civility and sanity everyone can see through but chooses not to.

I don't emphasize that much though because there's really not much more to say than that. Today, however, I'd like to take a minute to discuss an example of why I feel that way. It's a great example because it shows how things which are utterly absurd are accepted in this world without question or hesitations.

Specifically, I'd like to talk about how scientists who have no data which supports their conclusions get to call people who don't accept their conclusions science deniers. And I'm not just talking about some trivial name-calling in a university hall. I'm talking about published papers in scientific journals and articles in newspapers. I'm talking about media campaigns where tens of thousands of people see their claims, to the point policy makers even adopt those claims into their talking points.

Sounds pretty crazy, right? Science is supposed to be peer-reviewed. If people publish results they have no data for, other scientists should notice. Some scientist should read the work and say, "Hey, you don't have any data which shows this. You're just making things up!"

And if scientists don't do it, surely journalists should. Journalists are supposed to be about investigating things, right? Surely investigations should catch the fact there is nothing to investigate, right?

Nope. That might happen in a sane world, but it hasn't happened in this one. In this one, people like the researcher Stephan Lewandowsky can get a lot of publicity, attention and acceptance for saying things like:

Standing in opposition to scientific thinking, a known attribute of conspiracist thought is that it can appear incoherent by conventional evidentiary criteria. To illustrate, research has shown that when people reject an official account of an event, they may simultaneously believe in mutually contradictory theories—e.g., that Princess Diana was murdered but also faked her own death. The incoherence does not matter to the person rejecting the official account because it is resolved at a higher level of abstraction, namely the unshakable belief that the official account of an event is wrong. For the case of climate change, Thagard and Findlay (2011) showed that the contrarian position, exemplified by the opinion that global warming is a natural fluctuation, is incoherent in comparison to the mainstream scientific position.

That's a lot of verbiage, but the argument is simple. Lewandowsky says conspiracy theorists are known to hold contradictory beliefs about things as long as those beliefs say the "official" account is wrong. He calls that incoherent. He then says global warming contrarians hold incoherent beliefs. The point is to say anyone who rejects what he considers the "official" position on global warming is an incoherent conspiracy theorists, a point emphasized in his Conclusion which says:

There is considerable evidence that the rejection of (climate) science involves a component of conspiracist thinking. In this article, I provided preliminary evidence that the pseudo-scientific arguments that underpin climate denial are mutually incoherent, which is a known attribute of conspiracist ideation.

Climate denial is therefore perhaps best understood as a rational activity that replaces a coherent body of science with an incoherent and conspiracist body of pseudo-science for political or psychological reasons.

Lewandowsky conveniently fails to mention two key points. First, he fails to point out the "considerable evidence" conspiratorial thinking is involved in the rejection of global warming is largely his own work. Second, he fails to point out there is actually no data which supports any of his claims.

Yeah, it's probably not surprising he doesn't tell anyone the "considerable evidence" he refers to isn't actually based on data or, you know, evidence.

But before discussing that, let's talk about his idea of "incoherence." Remember how he said?

Standing in opposition to scientific thinking, a known attribute of conspiracist thought is that it can appear incoherent by conventional evidentiary criteria. To illustrate, research has shown that when people reject an official account of an event, they may simultaneously believe in mutually contradictory theories—e.g., that Princess Diana was murdered but also faked her own death.

This is a vague reference to a paper named Dead or Alive: Beliefs in Contradictory Conspiracy Theories. As you can guess by its name, it claimed to show people believed multiple conspiracy theories at the same time, even though those theories contradicted one another by surveying a group of people and examining their responses. The problem is nobody who responded said they believed in any contradictory conspiracy theories.

Think about that. This paper, by Michael Wood and several co-authors, explicitly claims:

In Study 1(n= 137), the more participants believed that Princess Diana faked her own death, the more they believed that she was murdered.

But none of the 137 people surveyed said they believed Princess Diana faked her own death and she was murdered. Only two people even said they believed Diana had faked her own death. Neither said they believed she was murdered.

Think about that. Here we have a group of scientists claiming the more people believe one thing the more likely they are to believe another based on data which shows nobody believed both things. It's completely insane. Or as Lewandowsky puts it, this work is incoherent.

I've actually talked about this on this site before. The second post I ever wrote here was a publication of my communication with the lead author of this paper, Michael Wood. I explained how he reached his insane conclusions in simple terms even a child could understand:

I’m afraid my examination of your data has turned up a serious problem. Your paper argues people endorse contradictory conspiracy theories, and it uses correlation coefficients for support. However, these correlation coefficients are calculated over groups that both reject and endorse conspiracy theories. These are the possible pairings for any two conspiracy theories:

Endorse – Endorse
Endorse – Reject
Reject – Endorse
Reject – Reject

This could be represented as:

+ +
+ –
– +
– –

A positive correlation happens when signs are identical. That covers the first pairing where both conspiracy theories are endorsed. However, it also covers the fourth pairing where both conspiracy theories are rejected.

The effect of this is responses which reject both of a pair of contradictory conspiracy theories will be treated, by your approach, as evidence people believe contradictory conspiracy theories. That’s nonsensical. If a person does not believe in any conspiracy theories, they obviously cannot believe in contradictory conspiracy theories.

I went on to discuss the matter in more detail, using math to show exactly how his results arose. Those details aren't important though. The basic point is simple. As you can see, a positive correlation happens whenever people give the same answer to two questions. It would be like asking:

Are you a Cubs fan? No.
Are you an alien? No.

You might find a correlation between answers to those two questions. Most people are not Cubs fans. Most people are not aliens. That means most people are going to give the same answer to both questions. A person could say the answers to those questions are correlated.

But nobody would claim that proves Cubs fans are aliens. The fact people who say "no" to a question are likely to say "no" to another question does not inherently mean people who say "yes" to one question are likely to say "yes" to the other question. Being different in one thing doesn't automatically mean you'll hold opposite views on everything.

"Study finds men don't like drinking spoiled milk, therefore women must love drinking spoiled milk!"

No! Even an imbecile can understand why that's ridiculous. It's so insane it's not even wrong. It's, well... incoherent.

But Michael Wood and other scientists published a paper relying on exactly that. None of their colleagues spoke up. Instead, their paper has been cited in some 80 other papers, with dozens of other scientists repeating their claims. These papers have each been cited more times, with more and more papers building upon this paper whose conclusions depend on absurd the argument:

People who don't believe A don't believe B, therefore people who believe A must believe B.

Which with just a little alteration could be used to prove horrible things about any group. All you have to do is ask people who aren't in the group you dislike compromising questions. They'll deny whatever you say, and you can cite that as proof the group admitted to whatever you asked.

That might sound crazy as your results won't be proof in any real, logical sense, but you will find a "correlation." You'll just have to ignore the part about such correlation being meaningless. To demonstrate, try these three steps:

1) Ask group X if they think the moon landing was real. They’ll say yes.
2) Assume group Y would answer the opposite way.
3) Conclude group Y believes the moon landing was faked.

That may sound completely ridiculous, but that's actually exactly what Stephan Lewandowsky did. This is a scientist who gets to tell the media who is in "denial" and what views are "incoherent." If he gets to do this, then why shouldn't you? This is a graph showing Lewandowsky's data set for a paper which claimed global warming skeptics believe the moon landing was faked:

Fig4

You see that line there? That's the line you'd expect the data to center around if there was a real correlation between views on global warming and views on the moon landing. You'll notice the data really doesn't center on the line. That's because the data is almost all on the bottom half of the graph. That's because almost every person who responded to Lewandowsky's survey said they believed the moon landing was real (the X and Y labels of the graph are reversed).

Just like Michael Wood and his co-authors, Lewandowsky relied on a data set which simply didn't have data for the questions he was interested in. The problem was worse in the Wood data set though. Here is that data with a line representing the relation he claimed to find. It's to scale:

Fig5

Look at all that whitespace. It's there because Wood pretty much didn't have data from anyone claiming to believe the conspiracy theories he was interested in. Just looking at this data, or just looking at Lewandowsky's data, would be enough to show they don't have data to support their conclusions.

And yet, they published papers promoting those conclusions. Lewandowsky published two. That's three different papers, using three different data sets, all using a methodology even the dumbest of the dumb could see is wrong. That's three papers which have been cited and relied upon by colleagues. Three papers discussed and promoted in the media. Three papers which have helped shape the public discussion of global warming, said to be one of the most important issues of our time, to label the people the authors of those papers dislike as irrational and incoherent.

I don't have words for how absurd that is. What I do have is data from a survey I ran to demonstrate this problem. I've written a description of the results in which I explain why we can't draw certain conclusions like:

Skeptics believe genocide and pedophilia are bad. Global warming proponents are the opposite of skeptics so they must believe genocide and pedophilia are good.

Because that's just stupid and insane. But if we use the approach Michael Wood and his co-authors use, the one relied upon by Stephan Lewandowsky, that's a true statement. Based on their idiotic approach, it is perfectly reasonable to say they're more likely to be pedophiles due to their views on global warming.

They're not. Agreeing with people like Stephan Lewandowsky does not make you a pedophile. Disagreeing with them does not make you a conspiracy theorist.

Or at least, it shouldn't. Maybe it does. After all, the world is having no problem accepting the work these guys do. Universities are supporting them. Scientific journals are defending them. News organizations are publishing articles praising them. I think that's all happening because the world is insane, but... maybe I'm the crazy one.

Maybe global warming activists do like molesting children.

7 comments

  1. I just posted a comment over on that site. It uses Disqus for its comments system, which I despise, but I couldn't stand the idea of the piece going unanswered. I'm copying it here for posterity's sake:

    There is a lot I take issue with in this piece, but the most troubling issue I see is the author of this post seems to be trying to present people who disagree with his views as unscientific, holding incoherent beliefs, while relying on work that is the epitome of incoherence and pseudoscience.

    The author refers to a paper by Michael Wood and several other co-authors, claiming it shows people hold believe contradictory conspiracy theories simply because they contradict the "official" account. The paper claimed to prove this in regard to several contradictory conspiracy theories regarding Princess Diana's death, but the people it surveyed didn't believe the contradictory theories it cited. Literally nobody they examined held contradictory beliefs they claimed to prove people hold.

    He also refers to his own work, in which global warming skeptics were portrayed as conspiracy theorists who believe things like the moon landing being faked. Those results didn't arise from finding global warming skeptics who believed the moon landing was faked. There were practically none in his data set. The same thing happened with a second study he published.

    The author of this piece, Stephan Lewandowsky, is portraying people he disagrees with as anti-science while promoting and performing work which draws conclusions despite a complete lack of data supporting them. These papers can take 0 data points which support a conclusion and claim to prove the conclusion is true. Calling it pseudoscience is being generous.

    The reason this happens is these people apparently don't understand basic aspects of the methodologies they use. Anyone who actually understands correlation tests know they assume data is normally distributed. If that assumption isn't met, as in the data sets used by Lewandowsky and Wood, the tests can produce spurious results.

    I've written about this in some detail (such as here), but for a simple demonstration, Lewandowsky's methodology can be summed up in these three steps:

    1) Ask group X if they think the moon landing was real. They’ll say yes.
    2) Assume group Y would answer the opposite way.
    3) Conclude group Y believes the moon landing was faked.

    That's all he did. You could repeat the same process with any question and any groups to let yourself conclude whatever you want about anyone you want. You don't like Democrats? Ask Republicans if they oppose slavery. They'll say yes. According to Lewandowsky's methodology, that proves Democrats support slavery.

    Ask men if women should stay in the kitchen. They'll say no. According to Lewandowsky's methodology, that proves women think women should stay in the kitchen.

    Ask global warming skeptics if they think pedophilia is bad. They'll say yes. According to Lewandowsky's methodology, that proves global warming advocates are pedophiles.

    That's completely wrong. But it's also perfectly fair. The exact same methodology Stephan Lewandowsky uses to portray global warming skeptics as conspiracy theorists "proves" he's a pedophile. I have the data to "prove" it.

  2. You are right that one has to be very careful computing the confidence intervals for correlations if the data has such a strange distribution. The standard estimators of the correlation and its confidence interval assume that the deviations from fit line follow a normal distribution, which is here clearly not the case.

    However, I am not fully convinced that your fit lines are right. Did you draw the line of the graph between moon landing and global warming by hand or did you use a formal least squares estimate (the fit line option in a spread sheet)? My guess would be that the real fit would be pretty flat and I would not even be surprised if it went the other way.

  3. Victor Venema:

    You are right that one has to be very careful computing the confidence intervals for correlations if the data has such a strange distribution. The standard estimators of the correlation and its confidence interval assume that the deviations from fit line follow a normal distribution, which is here clearly not the case.

    Yup. It gets even worse in the case of Stephan Lewandowsky's work as he uses factor analysis on his data to further examine the supposed correlation structures. The lack of univariate normality, much less multivariate normality, in his dataset makes that completely inappropriate. And yet, he's managed to publish two papers and get a lot of media attention solely because he did it.

    However, I am not fully convinced that your fit lines are right. Did you draw the line of the graph between moon landing and global warming by hand or did you use a formal least squares estimate (the fit line option in a spread sheet)? My guess would be that the real fit would be pretty flat and I would not even be surprised if it went the other way.

    I didn't plot fit lines. The lines I plotted show the sign of the relationship presented by these claims. They do not show a fit of the data.

    That said, the relationship given by the correlation coefficient and the relationship given by the slope from a linear fit will have the same sign, by definition. Correlation coefficients and regression slopes indicate the strength of the linear relationship between two variables. They are highly related, with each providing a bit of additional information the other doesn't provide (how perfect the relationship is vs. the magnitude of the relationship).

    If you want, it's actually pretty easy to translate between correlation coefficients and regression slopes. That doesn't really matter for my graphs though. I didn't plot the magnitude of the relationships because the authors of the papers didn't examines those either. If you do plot the magnitude, what you'll find is the slopes are so small the lines don't cross the halfway mark - a strong indicator they don't have explanatory power for the less sampled groups.

    That point probably would have been worth highlighting, now that I think about it.

  4. So it looks like I was right to post a copy the comment I submitted to that site up above. The comment was deleted without any explanation (or indication) that I could find. I don't believe the comment violated any of the site's rules, but I suspect my remarks about pedophilia made some moderator not really care about enforcing the rules. I've re-submitted the comment with the last two paragraphs removed. We'll see if that lets it stay.

  5. I've found that when I log into my own DISQUS profile, it saves deleted comments. It marks them with a little red box that says "removed" for deleted comments and "pending" for ones in moderation.

  6. Huh, thanks. That does work. I can see my deleted comment by doing that. As much as I dislike Disqus, I guess it is nice that I won't have to save copies of comments I think my get deleted on sites which use it.

Leave a Reply

Your email address will not be published. Required fields are marked *