Confirmation bias
Confirmation bias is the tendency for people to only seek out information that conforms to their pre-existing view points, and subsequently ignore information that goes against them. It is a type of cognitive bias and a form of selection bias toward confirmation of the hypothesis under study. Avoiding confirmation bias is an important part of rationalism and in science in general. This is achieved by setting up problems so that you must find ways of disproving your hypothesis (see falsifiability).
Contents |
[edit] We all do it...
Confirmation bias is one of the traits that just comes with the human condition. There is a human tendency to favour testing the predictions of a hypothesis that only confirm or prove it, at the expense of testing any predictions that would disprove a hypothesis. This is a problem because attempts to disprove a hypothesis are the most effective ways of comparing two or more hypotheses, and are also the most informative methods for acquiring evidence (see the Wason Card problem, below). However, even practicing scientists often miss these examples and the scientific method has basically "evolved" to try and counteract it.
In politics, confirmation bias explains, for example, why people with right-wing views buy right-wing newspapers and why people with left-wing views buy left wing newspapers. In general people both:
- Want to be exposed to information and opinions that confirm what they already believe.
- Have a desire to ignore, or not be exposed to, information or opinions that challenge things what they already believe.
Even in cases where people do expose themselves to alternative points of view, it may be a form of confirmation bias; in that they want to confirm that the opposition is, indeed, wrong.
[edit] Intelligent design
A prime example stems from the intelligent design and creationism movements. Proponents of these ideas start by assuming that an intelligent creator must have been behind life (often assuming the Christian god) and then seeking out any (and only) evidence that might back up this claim. For example, Answers in Genesis often claims that fossils in the ground are proof of a Global flood and that layering is caused by the relative abilities of animals to reach higher grounds. They quote evidence which shows that humans and primates are on top of dinosaurs in the geological record and claim that this is because we "could climb to higher points on the landscape".
However, they never address any evidence that might disconfirm this hypothesis such as why angiosperm plants and their pollen is never found below vascular fern fossils, or why we see absolute cut-offs in the fossil record rather than a statistical tendency, which the "climbing" hypothesis would suggest. (In other words how come there weren't some weak or elderly humans who died with the dinosaurs.) The fossil record is in fact best explained by common descent but this is ignored through confirmation bias by creationists.
Creationists, along a similar vein, claim that "evolutionists" are practicing confirmation bias, since they are all atheists out to distort the scientific record to make it look like God does not exist. This claim is, of course blatantly false, theistic evolution disproves it. Furthermore not all scientists are atheists, and additionally old-earth and evolutionary theories were developed by considering new observations rather than dismissing the old ones.
[edit] Wason card problem
The Wason Card problem highlights confirmation bias quite nicely. Four cards are presented to the "player", each labelled with a letter on one side and a number on the other side. There are two cards with letters face up and two with numbers face up (A, B, 1, 2). The following hypothesis is then tested;
- "If a card has a vowel on one side, then it has an even number on the other side."
The aim of the experiment is to test this hypothesis by turning over cards to check. You can do this with fewer cards if you recognise how to properly test a hypothesis.
Most people, when given the Wason card problem, will immediately turn over the even number (2) and the vowel (A) to see what is on the other side. This is instinctive because turning these two cards and observing another vowel and an even number respectively would confirm the hypothesis - and without context or applied critical thinking, this will be people's normal approach. However, to fully confirm the hypothesis, a third card would be needed to be flipped; 1, to confirm that it doesn't have a vowel on the other side.
[edit] Solution
Looking more critically, it can be seen that the "2" card does not need to be checked at all. If there is an even number with a consonant on the other side, that has no bearing on the hypothesis that "a vowel always corresponds to an even number". This test cannot fail to confirm the hypothesis because even a negative result says nothing about it.
This is also an instance of the need to take care when dealing with implications: the hypothesis is "vowel implies even," not "vowel if and only if (implies and is implied by) even." The hypothesis can be properly tested with only two card flips. Although only around 10% of people get this first time under experimental conditions, almost everyone tested agreed with the logical answer when it was revealed.
If one sets out to disprove the hypothesis the problem can be solved in two turns, rather than three. The card with 1 on one side is the actual deal-breaker in testing the hypothesis properly. The correct cards to turn over would be 1 and A. First, turning over the odd number and viewing a vowel on the other side of that card would invalidate the hypothesis quickly and more efficiently; vowels shouldn't have odd numbers on the back. This is the main counter-intuitive step that demonstrates the need to try and disprove a hypothesis first. Once the odd numbered card is flipped, the next logical card to turn would then be A to confirm whether it had an even number, and as stated earlier the B and 2 cards are actually irrelevant to proving or disproving the hypothesis because all possible results have no effect. Revealing the opposite side of 2 would either confirm the hypothesis, by displaying a vowel, or say nothing, by showing a consonant (which is the same reason that the B card is uninteresting and irrelevant).[1]
[edit] Alternative Presentation
The Wason problem presents a fairly abstract version of a simple hypothesis - because this doesn't necessarily engage the best human cognitive processes, the problem is frequently answered badly. Reformulating the exact same situation with something more visceral does show it to be considerably easier.
Consider the following situation:
You are working in an all-ages club and your job is to make sure that no one under the age of 21 is drinking alcohol. You see four people. You know person A is 40 and person B is 19, but you don't know what either one is drinking. You know that person C is drinking a soft drink, and person D is drinking vodka, but you don't know the ages of either. Of these four patrons, which ones do you need to investigate further?
This question is equivalent to the card problem; with ages substituting for numbers and drinks subtitling for letters. The more concrete and specific application to social mores makes it easier for our socially-inclined brains to see the correct answer: you need worry about only person B and person D! This shows that the difficulty of a problem can greatly depend on what cognitive processes are engaged.
[edit] See also
- Apophenia
- Cherry-picking
- Common sense
- Bayesian inference
- Morton's demon
- Self-fulfilling prophecy
- Walled garden
- Willful ignorance
[edit] External links
- David McRaney, Confirmation bias (youarenotsosmart.com, 23-June 2010)
- Brendan Nyhan and Jason Reifler. When Corrections Fail: The persistence of political misperceptions (pdf) (Poltical Behavior, 2006)
[edit] Footnotes
| Articles about logical fallacies | |||||
|---|---|---|---|---|---|
| Formal fallacies | |||||
| Affirming the consequent | Denying the antecedent | ||||
| Informal fallacies | |||||
| Red herrings | |||||
| Conditional fallacies | |||||
| Fallacious argument styles | |||||
