How to Mislead People Using Reliable Data
The second easiest person to fool is someone who wants to believe.
The physicist Richard Feynman once famously said of science,
“The first principle is that you must not fool yourself, and you are the easiest person to fool.”
Good scientists have frequent meetings where experiments and results are criticized, sometimes savagely — it’s better to discredit an idea before it’s embarrassingly published. Lab colleagues, journal editors, peer reviewers, scientists who want to extend the work — they all look for ways I might have fooled myself, and they would delight in showing that I’m wrong. That’s how science works.
This is a lesson about fooling ourselves, and how to avoid it. It’s about how to keep from being fooled when someone is trying to use data to convince us of something. It applies in business, science, politics, economics.
It’s also an examination of climate denialism, with a particularly entertaining and risible example from just a handful of years back.
There’s a graph involved. Don’t let that scare you away. I’ll be focusing on the ideas, and taking that graph apart into its pieces to make it easy to understand — which is exactly what the person who made that graph must have hoped his readers would not do.
I’m going to joyfully name names at the end, with receipts. Science asks people to stand behind their work. It’s OK to be wrong. It’s not OK to be dishonest or incompetent, to try to fool others. That also is what this essay is about.
The graph
This graph claims to tell us how much CO₂ was in the air, and how it changed, from 1810 to 2010. In 2014, for a few months, it was all the rage among some people trying to argue that climate science is fraudulent.
In itself, it’s an interesting image of 3 different sources of historical CO2 data, all put into one place. The data are exactly what they’re claimed to be, there is no dishonesty or ‘fake news’ in the numbers that are being presented. Take away that wiggly line to the left of the ‘1957’ arrow, and I like this graph a lot.
What’s wrong, perhaps intentionally, is the analysis of the data — that bouncing line that runs across most of the graph — and the conclusions they draw from it. Let’s dive in and see why. Along the way, we’ll talk about how to evaluate the reliability of information and ideas, and about how good information can be dishonestly used to mislead people.
Lower right corner — the Keeling Curve
Look at the graph again. The short section of slightly curved rising line in the lower right corner is labeled “Mauna Loa Background.” It’s well-known data taken from the CO₂ Instrument at the Mauna Loa Observatory in Hawaii. It’s called the “Keeling Curve,” after Charles Keeling, who established a CO₂ recording instrument there in 1958.
The Mauna Loa location was chosen because it is isolated from CO₂ contamination by animals, plants and volcanic vents, and away from cities or industrial emissions of CO₂. It is downwind of 2,500 miles of nothing but ocean. The result is a 62 year long daily record of the uncontaminated, well-mixed atmospheric CO₂ in the Earth’s northern hemisphere.
The Keeling Curve is one of the most beautiful results in all of science. Several things stand out with just a little thought.
Every year, the amount of CO₂ goes up and down in a clockwork wave. This is the earth’s ‘breathing,’ caused by CO₂ moving into and out of the ocean as it warms and cools, and into and out of plants as they grow and decay. In our graph this is small enough that we can’t see it, and that’s OK — the annual variation is not the point of the graph.
In the expanded view, the repeating annual cycle tells us the method is reliable. It can reproducibly measure the same small changes year after year, for 62 years. Reliability and repeatability are important if you’re trying to decide whether to believe data. It’s one of the ways that scientists interrogate our data to see if we’re fooling ourselves.
For those 62 years the amount of CO2 also goes smoothly steadily inexorably up, from 315 PPM in 1958 to 410 PPM in 2018. It not only goes up, it accelerates, rising faster at the end than at the beginning. We’re emitting more CO2 each year now than 62 years ago.
The source is not cited. The ability to evaluate source information is important — I did exactly that in going to the Keeling Curve, which I knew without citation. Failure to reveal the source can be an important clue that someone is trying to fool you.
Across the bottom — ice core data
Across the bottom of the graph are a series of dots, labeled “Ice Core Antarctica.”
When snow falls on glaciers across the world, air between snowflakes becomes trapped, making tiny air-filled pores as it packs into glacial ice. Scientists drill deep ice cores into those glaciers and extract the trapped air to measure CO₂ levels. They count back years by counting annual changes in snow, just like seeing how old a tree is by counting growth rings. They can tell exactly what year that a tiny pocket of air was trapped, and exactly how much CO₂ is in it.
I think that is an extraordinary and beautiful thing to be able to do. They directly measure CO₂ and other gasses in air that has been trapped for decades, thousands, up to a couple million years ago. That’s kind of wondrous.
Each of those dots in the ice core data is CO₂ from a year in one of those samples. At the time this graph was created there was data going back 800,000 years, but this graph only shows data since 1810.
The ice core data is internally reliable and repeatable. There’s a point every decade or so, and none of them fall out of place. The technique works.
This data also shows CO₂ rising through time, even from the beginnings of the Industrial Age 200 years ago. If we imagine the Keeling Curve extending backward through time, it connects seamlessly with this data. Taken together, they form a consistent continuous curve of CO2 in the air, from 280 PPM in 1810, to 415 PPM now in 2020.
The Keeling Curve and the ice core data use two different techniques, one measuring flowing air in Hawaii and the other ice pores in Antarctica. The results are repeatable, reliable, internally consistent within each experiment, and consistent across the different experiments. Looking for this kind of consistency between different kinds of experiments is another important way we set about trying not to fool ourselves.
That bouncing curve filling most of the graph
This is labeled “Local Effective Concentration.” The bouncing line is a series of individual points, with a curve drawn connecting them. Ignore the curve for now — it’s misleading and scientifically invalid. Pay attention only to the small points you can just barely see under the bouncing curve.
Methods for measuring the concentration of CO₂ in a sample of air were developed in the early 1800s. The method was precise — within 3%, as the label says. Each of those points is a separate determination of the CO₂ in a sample of air, published in the scientific literature of the time.
You can immediately see that the chemical data are not internally consistent. Many of the experiments have wildly different results. They are not consistent with each other, and much of it is not consistent with the ice core data, which we have reason to trust.
If you take the time to track down the experiments that gave us each of those points, you’ll find they were done by careful and respected scientists and published in reputable journals of the time. Each point is a valid historical measurement of CO₂, published in the scientific literature of the time.
We trust the method. We trust the skill of the person using that method. Using trusted methods in the hands of skilled people, we still get results that vary widely from person to person and place to place, and that are different from other data we have good reason to trust. What happened? Perhaps what they’re measuring is different.
What’s different?
All of the early samples up until about 1860 have much higher CO₂ than the ice core data. If we read the scientific papers — which I’ve done — we find that the samples were taken in labs using lanterns or candles for lighting, and in cities often with gaslights, industry, coal heating.
Each of the samples is measuring a unique local environment. They are representative measurements of CO₂ in air from labs and cities of the time, which were highly contaminated with local sources of CO₂. They are one kind of experiment and results, and they are not measuring the well-mixed global air that the ice core and Keeling experiments measured.
The later chemical experiments, between about 1930–1950, also have much higher CO₂ levels. These were done explicitly to find local CO2 levels in cities, near factories, and other places where we emit large amounts of CO₂. They’re intentionally measuring known locally high concentrations of CO2, to see just how high they are. These are a second different kind of experiment and results.
The rest of the chemical data — the part inside the yellow circle — has values close to those from the ice core data. If one looks at the literature there was a consistent effort to find air uncontaminated by local effects. They went to the shore and sampled air from ocean breezes. They went to the mountains. The samples weren’t perfect — there is still a lot of visible “noise” from remaining local contamination. But they are reasonably good.
These are a third kind of experiment and results, and they’re attempting to measure well-mixed global air — the same thing the ice core and Keeling curve experiments measure.
Attempting to compare different kinds of data without recognizing that they are different, as this graph does, is another classic way to fool yourself. It’s one that climate deniers frequently engage. Understanding where the data came from and what the experiment is measuring is another important way to avoid fooling ourselves, or being fooled by others.
If you group only the set of experiments that measure well-mixed air, the results are internally consistent. There is variability, but the variability is relatively small, and they measurements fall near the line through the ice core data. Once more, internal consistency, and consistency with other experiments measuring the same thing, show us the data can be trusted, and help us avoid fooling ourselves.
So, what’s the problem with the graph?
The maker of that graph drew a line curving through each one of the chemically-determined data points. In science, a line curving through multiple measurements, like this one, has a meaning. It tells us that this line represents CO₂ in the air making a continuous change from one point to another.
Sometimes that’s OK. We can draw a single smooth line through the ice core and Keeling data — I’ve already asked you to imagine that line. Those data are internally consistent and agree with each other We have good physical reasons to believe that CO₂ follows that line. The relevant subset of the chemical data, even with variation from measurement to measurement, clusters around the line and help us believe it’s accurate.
When the maker of this graph drew a curving line rising and dropping rapidly and often, running exactly through each of the chemical data points, they told us that is how atmospheric CO₂ changed over time. But we know, from the other data on the graph, and from the fact that this is from 3 different kinds of data, that it isn’t true.
There’s another reason to believe this claim can’t be correct. If it were true, it would require that 100 PPM of CO₂ entered the atmosphere in only 20 years between 1810 and 1830, and then immediately left again in the 4 years following. That is a massive amount of CO₂ — I’ll leave the calculation to another time. No known reservoir could supply or remove it, and there is no known mechanism to move that much, that fast.
This quick reality check should have told the maker of the graph that they were fooling themself. Doing that kind of automatic reality check, asking if this makes sense given other things we know are true, is yet another way to keep from fooling ourselves. It’s one the maker of this graph appears not to have used.
We can’t know if the maker was fooling themselves or trying to fool us. Practically, it makes no difference. That graph used reliable data combined without explanation from experiments measuring three different kinds of things, to tell us a story about CO2 that can’t possibly be true. The graph itself, if we pay attention to the details in it, tells us the story is suspect. Additional information from source material confirms it.
Enter Joe Bastardi, Tim Ball, Ernst-Georg Beck, and Anthony Watts
I’m don’t want to assassinate characters. I’m looking at how one fools oneself — the easiest person to fool. That requires looking at people who fooled themselves. I’m going to do these gentlemen the courtesy of assuming they are as intelligent and scientifically literate as they present themselves to be. They know enough science to know better, but they still fooled themselves — and others.
Joe Bastardi is a popular meteorologist, TV personality and climate science denier. On November 13, 2013, he posted this graph on Twitter, with the message, “Somehow, chemical measurements of much higher CO2 concentrations in atmosphere thrown out by AGW scammers.”
That’s an extraordinary claim. He wants us to believe those early high values and rapid changes on the graph represent an accurate history of global atmospheric CO2. And more, that climate scientists “somehow” threw them out for implied nefarious purposes, to hide very high CO₂ levels in the recent past. He may have gotten the graph and the claim from Tim Ball.
Tim Ball is known for the energetic way he goes about attempting to discredit climate science and scientists. On the same day as Joe Bastardi’s tweet, November 13, 2013, Dr. Ball published a guest column on Anthony Watts’ climate denial blog, with this unattributed graph as Figure 8.
In section 6 he quotes Ernst-George Beck. “Since 1812, the CO2 concentration in northern hemispheric air has fluctuated exhibiting three high level maxima around 1825, 1857 and 1942 the latter showing more than 400 ppm.” He goes on, like Mr. Bastardi, to accuse climate scientists and the IPCC of fraudulently hiding that there were high historic CO2 levels and that they change quickly.
It is the same extraordinary claim, made the same day. This is notable mostly for showing that this kind of fooling oneself can be contagious. People who want to believe, are good at finding confirmation and spreading the thing they’ve heard.
The second easiest person to fool is someone who wants to believe
Four intelligent educated people, in a short time, much of it on the same day, put this graph and its underlying information into circulation. They made impossible claims about what it meant. This graph, and the claims of fraud that went with it, influenced attacks on climate science for several years after.
Perhaps they believed that climate science must be wrong and this must be right, and fooled themselves. Wanting something to be true is a common way to make it even easier to fool yourself.
Perhaps they accepted and broadcast the data without analyzing it. If so, that’s just bad science — the easiest person to fool is yourself, and every good scientist constantly works to avoid it.
Perhaps they intended to fool their audience. For an unscrupulous person who knows the science, who knows data analysis, who is familiar with the ways we easily fool ourselves — it’s trivially easy to use those same mistakes to intentionally fool others.
I don’t know, I can’t know, what their intent was. What I do know is that these 4 intelligent people on this and other subjects, and an entire industry of climate science denial, continue to make these same mistakes over and over again. They continue to never correct themselves when analyses like these show that they must be wrong. You can draw your conclusions from that.
I also know that there are an army of people who have been either accidentally misinformed or intentionally lied to by this denial industry, and led into thinking there a massive conspiracy of science with the intent to destroy things they value.
I think that misinformed army is dangerous to our culture and our future, but I also feel sad for them. We humans are not rational animals — we’re animals who have figured out how to sometimes create structures and relationships that allow us to do rational things. That’s what science is. One of the rational things we’ve learned to do is to fool others for our own purposes. Someone who wants to believe, is the second easiest person to fool.
