Dr. Jason Lisle
[Originally published as the second section of The Ethics of Science: Part 1]
Contrary Predictions
Another way in which the confirmation bias can lead to error is when a person fails to consider the predictions of the alternatives.
Suppose a hypothesis predicts a particular outcome of an experiment: outcome X. But then a different hypothesis also predicts outcome X. Suppose the experiment is performed, and indeed the outcome is X. This does not verify either hypothesis because they both made the same prediction.
A common instance of this error is when evolutionists cite similarities in the DNA of organisms as evidence for evolution. Indeed, different organisms have similarities in their DNA sequences. Furthermore, organisms that have greater similarities in anatomy and physiology tend to have greater similarities in their DNA sequences. This is what would be expected if evolution were true.
But it is also what we would expect if DNA codes for traits. If DNA codes for traits, then organisms with similar traits should have similarities in DNA, regardless of how those organisms came to be. So, creationists who believe that DNA codes for traits would also predict that organisms with similar traits should have similar DNA sequences. Since this is a prediction of both creation and evolution, it does not verify one over the other.
You will find that essentially all successful predictions of Darwinian evolution are also predictions of biblical creation. A proper experiment to test which of two competing models is more accurate must involve a situation where the two models make different predictions about the outcome. Only then can the results of such an experiment be used in favor of one model over the other.
Another way in which this error occurs involves confusing correlation with causation.
When two phenomena (A and B) are linked, it is common to assume that A caused B. This is one possibility. However, another possibility is that B caused A. A third possibility is that C caused both A and B. Therefore, simply showing a link between A and B does not, by itself, establish that A caused B.
For example, if elevated levels of atmospheric carbon dioxide cause the earth to warm up, then there should be a correlation between global temperatures and atmospheric carbon dioxide. And there are some observations in support of such a correlation. Does this mean that the elevated carbon dioxide levels have caused global warming? Not necessarily. The warming of the earth could cause an increase in atmospheric carbon dioxide.
The earth’s oceans contain a great amount of dissolved carbon dioxide. Cold water can hold more dissolved carbon dioxide than hot water. So, when the oceans heat up, they release carbon dioxide into the atmosphere. When the oceans cool, they can reabsorb atmospheric carbon dioxide. Thus, a correlation between atmospheric carbon dioxide and global temperature does not automatically mean that the former caused the latter; the reverse is possible, as is a combination of the two.
Control Group
A proper experiment should have both a test group and a control (or placebo) group.
Suppose we want to test the hypothesis that adding a particular chemical (Y) to the soil will cause plants to grow faster. We would want two groups of plants: one in which the chemical is added to the soil and a control group where the chemical is not added to the soil. Only by comparing the results between the two groups can we infer whether the added chemical had any effect.
All other factors should be as identical as possible between the two groups. For example, they should receive the same amount of sunlight, they should be kept at the same temperature, they should be watered equally, and so on. The only difference should be the test variable — chemical Y. Otherwise, we would not know if a difference in growth is due to the chemical or one of the other differences.
In practice, it is impossible to eliminate all other variables. After all, the plants in the test group are not identical to the plants in the control group. They may have slight differences in DNA, or a different quantity of bacteria, and so on. There is simply no way to ensure that all non-test variables are identical. So how can we be sure any differences in outcome are due to the test variable (chemical Y) and not intrinsic differences in the plants?
The answer is to use a large number of plants in each group. Yes, perhaps a plant in the control group has a gene that limits growth. But if we use a large number of plants and randomly assign them to each group, the probability that all the plants with a particular gene would end up in one group is exceptionally small. The intrinsic differences between the plants tend to “cancel out” as we increase the number of plants in each group.
The mathematics of statistics allows the scientist to determine whether a difference between the two groups is significant, or merely due to intrinsic undetermined variables.
Double Blind – the Gold Standard
Another way in which human beings fail to be accurate and objective concerns the placebo effect. When a person self-reports the effect of a particular treatment, the report will be strongly biased on whether the person expects the treatment to work. If a person is told that a sugar pill is actually medicine to alleviate a headache, and the person believes this and takes the pill, he will usually report that his headache is better. His expectation affects his assessment of the pain. This is called the placebo effect. The placebo effect makes it difficult to separate the genuine effectiveness of a treatment from the person’s subjective assessment.
For this reason, medication trials are usually performed in a “blind” fashion whenever possible.1 This means that the person does not know if he is getting the medication or simply a placebo (a pill that does nothing). Both the control group and the test group are in the dark about whether they are receiving the actual medication. This reduces their expectation that the treatment will work and results in a somewhat more objective self-reporting. Or at least, there will be an approximately equal number of people in each group that expect the treatment to work and give a biased report. Therefore, any differences in the average results should be due to the actual medication and not the placebo effect.
The best method is to use a double-blind trial. This is where neither the test subjects nor those administrating the treatment know whether the treatment is the medicine or the placebo until after the experiment is completed. This prevents those administering the treatment from inadvertently treating the control group differently from the test group.
The Right Tool for the Right Job
The scientific method is a powerful tool when used properly. Such usage includes a recognition of the scope and limitations of the method. The method was not invented to answer all categories of questions, such as questions of history, morality, theology, and philosophy. There are rational ways to deal with such questions, of course. However, the scientific method was designed to discover the systematic way that the Lord normally upholds His creation. It is predicated upon the fact that God has established cyclic patterns that He promises will persist as long as the earth does (Genesis 8:22). The method is designed to reduce man’s propensity to error and self-deception through a number of safeguards.
Those who make new scientific claims without following these biblical guidelines are not behaving in an ethical fashion. It is our fallen nature to believe our own conjectures about nature without sufficient testing and without sufficient consideration of alternatives. This stems from pride. It is our fallen nature to simply believe whatever thought pops into our head, particularly if such a thought seems sensible and is desirable. But to believe mere speculations without sufficient evidence is not honest. Nor would it be honest to attempt to persuade others of such untested conjectures.
The scientific method was developed to reduce the effects of human error by relying upon God’s faithfulness in upholding His creation in a consistent way. It is based on the biblical worldview and reduces our propensity to fool ourselves and others. If we then follow this method properly to make a new discovery, what is the appropriate procedure to inform others? There is a right way to do this, and there are many wrong ways.
References
- In some cases, a blind trial is simply not possible. An assessment of the effectiveness of face masks would have a control group with no masks and a test group with masks. Obviously, the people in each group would know whether or not they are wearing a mask.
