Science Fact. STEM education – science, technology, engineering and math – can only take us so far when it comes up against politics. Part 1 of this post described the career opportunities and general technical literacy that make STEM education an important part of every curriculum. This Part 2 describes research that shows the limitations of STEM education.
Scientific consensus never reaches a final conclusion because science is by nature subject to change as new data is developed. Nevertheless at any given time science has a general consensus on a great many issues. So why is the consensus of scientists, on a subject they spend their entire lives studying, so unconvincing to the general public?
The limitations of science and of STEM education in shaping public opinion are dramatically shown by a brilliant collection of studies. On a number of these, the lead author is Prof. Dan Kahan, who is based at Yale Law School and also works with the Harvard Safra Center for Ethics. Kahan overflows over with good ideas, and this post draws heavily from his published articles.
An example from a 2013 study will give the general idea. Many people believe that if Americans had a better grounding in science and math they would embrace the scientific consensus rather than take positions diametrically opposed to the known facts. The researchers decided to test whether this is true.
The scientists hired a national polling firm to administer three tests to 1,111 U.S. adults whose gender, political persuasion, geographical distribution and race approximately match the U.S. population. Here are the first two tests:
1. One test measures political orientation, such as right versus left, or conservative versus liberal.
2. Another test measures “Numeracy” (numerical skill): the ability to look at scientific results on a wide range of world problems and draw an accurate conclusion from them.
The left-leaning group and the right-leaning group each included people with a full range of Numeracy abilities ranging from total incompetence to very high analytical skill.
Each person was then given one more test, designed to tell whether his or her personal politics got in the way of coming up with the correct answer. This is where the experimental design gets quite clever.
Consider the following test: the subject (the person being tested) is told about a new skin cream that is being tested and is shown the following results:
(This figure and the following ones are adapted from Kahan et al, Motivated Numeracy and Enlightened Self-Government)
The subject is asked which of these conclusions is implied by the data?
– People who used the skin cream were more likely to get better than those who didn’t; OR
– People who used the skin cream were more likely to get worse than those who don’t.
The data is totally fictitious, and has been chosen to make it difficult to come up with the right answer. When most people quickly glance at the table, they tend to either look at the top two numbers, or the two numbers on the left. The top two numbers show that more people got worse than better; the left-hand numbers show that for people whose rash got worse, most of them were using the skin cream. Both of these quick-and-dirty ways conclude that the skin cream makes the rash worse.
However, notice what happens when we calculate the percentages going across:
Among the people using the skin cream, 34% got better; and among people not using it, only 20% got better. So in fact the skin cream helps, but you can’t come to that (correct) conclusion without thinking about it and then dividing some numbers.
A random selection of one-fourth of the people tested were given the test above. Another random one-fourth were given the same problem but the following opposite data:
The column headings have been reversed! So naturally the correct answer is also reversed.
Not surprisingly, the people with a high Numeracy score tended to get the right answer whether they were given test A or test B, and the people with low Numeracy tended to get the wrong answer.
You may be saying, that’s only half the test subjects, and that’s also not a very interesting test. OK, but here’s the test that was given to another one-quarter of the group: they were told that a city was trying to decide whether to pass a gun control ordinance, so they were looking at data from other cities who either did or did not pass a similar ordinance. And here was the data the city was studying:
You’ll notice a certain similarity between this data and the skin test data: as pointed out, both are fictitious.
The test subject was then asked which of these conclusions is implied by the data?
– Cities that enacted a ban on carrying concealed handguns were more likely to have a decrease in crime; OR
– Cities that enacted a ban on carrying concealed handguns were more likely to have an increase in crime.
The remaining one-fourth of the test subjects were asked the same question but given the opposite set of data:
The data set C implies a decrease in crime following regulation, while data set D implies an increase.
Thus a full spectrum of Americans of every political persuasion and numerical literacy were asked both “innocent” questions and questions that carry with them a lot of political baggage.
When the researchers analyzed the data, they found that the “innocent” questions about a skin rash lured the less numerical folks into giving the wrong answer, but the more numerical people usually came up with the correct answer. However, the results were quite different when the question was posed as a gun control issue:
– 1. The less numerically talented folks gave conclusions that matched their political stance: they looked at the data and saw confirmation of the answer that they already believed was correct. Thus their answers were polarized by their politics. This was true for both left-leaning and right-leaning subjects! (A separate study showed that both conservatives and liberals exhibited “closed minds” about topics that didn’t fit their politics.)
– 2. The people who showed good numerical skills when looking at non-political issues were however not able to put aside their political views; in fact, they were more polarized than their less mathematical peers. Numerically literate people actually used their analytical talent to reinforce their existing political opinions rather than to look for objective truth.
Here’s how the researchers summarized what they found:
…more Numerate subjects would use their quantitative-reasoning capacity selectively to conform their interpretation of the data to the result most consistent with their political outlooks.
This is quite a remarkable result. It refutes the notion that a better understanding of science would cause Americans to adopt more rational views on sensitive issues. Rather, it shows that political issues, those in which we adopt opinions similar to the people we respect and care about, tend to overwhelm any objective consideration of the facts. It is simply not possible to take a scientific issue that has become politicized and get anyone – on the left OR on the right – to consider it dispassionately. We human beings are simply not wired that way!
The full article describing the work summarized above contains a lot of statistical analysis but it also has much interesting discussion for the non-specialist. Almost as fascinating are the follow-up articles in which the work is interpreted, re-interpreted and re-re-interpreted by people whose own political view colors their understanding of the research. In other words, even the discussion of research about bias is subject to bias in interpretation!
Science Speculation: Do these results mean that science is useless in addressing the major problems that the world faces? Let’s define some subsets of problems to see how we can overcome this dilemma in public policy. Differently stated, we’ll nibble away at this question from several sides:
– Safe Topics. As the researchers point out, very few research questions ever become a symbol of group identity. They give examples of areas where there are disagreements and a need for scientific studies, but only a tiny minority of people feel passionately about them: antibiotics to treat infections; health risk of cell phone radiation; government’s role in fire and police protection; the use of government-issued currency rather than barter and precious metals. So science can tackle and overcome many human problems without being totally stymied by politics.
– Re-Framing. Some issues need re-framing to permit rational discussion. For example, the phrase “rationing health care” provokes an emotional and mostly negative response. However, considering “how to allocate doctors and hospital beds to best save lives” is easier to consider for many people. Because of re-framing, the focus has become positive (“save lives”) rather than negative (“take money away from something praiseworthy, perhaps to spend it on something I oppose”).
– Win-Win Linkage. Another Kahan study finds value in what he calls “two-channel” communication and what I would call “linkage” or “win-win.” The researchers chose a controversial scientific question whose dire evidence tends to persuade Democrats (climate change). They paired it with a discussion of geoengineering, which addresses climate change not by suggesting government limitations on emissions, but by proposing ways to capture carbon on a large scale. The presentation of a solution that requires independence and entrepreneurship, values that appeal to Republicans, helped neutralize the polarizing nature of the climate question so that both left- and right-leaning subjects could consider the science more objectively. Thus by making sure that the discussion affirmed both conservative and liberal values, people became less polarized and were better able to consider the scientific facts on both sides of the question. Both conservatives and liberals adjusted their positions, moving a bit toward the center.
– Anticipation. Scientists would be well advised to plan ahead when they address issues that potentially affect large numbers of people and strong economic interests. If a problem is likely to become a political rallying point, soon scientific results will become ignored or mis-interpreted; in fact, the scientists themselves will become biased. So people who want to tackle social problems would be best advised to enlist both conservatives and liberals in its solution, re-affirming both sets of values as they proceed, to try to keep the issue from becoming a political football.
– Humility. Most importantly, scientists need to be humble about what they bring to the party. Facts, analysis and the confirmation of hypotheses may be totally convincing to scientists. However, what scientists believe amounts to nothing very special in the eyes of the public. A scientific conclusion is simply one more piece of information to consider – and not the most important one, either! – when a person develops a personal opinion on a subject of broad social concern.
Is STEM education futile when confronting controversial subjects? Almost everyone occasionally cherry-picks facts to support his personal views – have you ever noticed yourself doing that?
Drawing Credit: Scout, on openclipart.org