This article is rated Stub-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | |||||||||||
|
I'm not grasping the difference. In confirmation bias, people look for evidence that confirms their existing beliefs. It seems that in congruence bias, people do not try experiments that would refute their theories. Sounds like a special case of confirmation bias: avoid contrary evidence. How is it different? — Preceding unsigned comment added by 208.54.86.137 (talk) 04:32, 27 October 2013 (UTC)
I don't understand the example of congruence bias provided in this article. Were the subjects asked to provide any rule that could generate the sequence "2 4 6", or to guess the sequence the researcher was thinking of? If the latter, what does it mean they "tested" their guess of, say, "increment by two" with example sequences such as "3 5 7"? I don't understand what "testing" means here... 3 5 7 doesn't test anything! The only way to test a rule is to see if it generates the intended sequence. If, on the other hand, the idea was to guess what rule the researcher was thinking of (and not just any fitting rule), then the experiment seems pointless... I suspect this article is just poorly worded, or maybe I'm a bit slow today :-) Can the author clarify it a bit? 200.114.214.67 09:01, 11 March 2006 (UTC)
I agree with the first post (200.114.214.67) that guessing the rule is pointless. The rule can be something like this: take randomly any number among 2, 4 and 6. What's the idea to guess this? And how one would resolve it without the congruence bias?
The first example, IMO, is also quite meaningless. The presence or absence of the second button and its function helps discover nothing about the functionality of the first button. 200.69.58.43 00:14, 13 March 2007 (UTC)
--195.137.93.171 (talk) 11:58, 1 September 2009 (UTC)
The button example makes the wholly unwarranted assumption that the experimenter told the truth. Someone who'd lock you in a room for "science" wouldn't hesitate to lie. Dan 06:26, 14 November 2006 (UTC)
I want to suggest another interpretation of Wason's example: the investigator is demonstrating congruence bias. The investigator believes that people exhibit certain trait, namely: people are liable to incur into congruence bias. In order to confirm his believe, the investigator sets a confirmatory experiment. He runs the test involving the numbers 4,6,2, and interprets the result in light of his previous believes, without considering alternative hypothesis, such as poorly wordy instructions. In this particular example, the reference to the three numbers 4,6,2 as "the number sequence 2,4,6" and the reference to "particular rule" are very misleading, because the word "sequence" suggests some underlying ordering, in which each element follows the previous ones, and is generated from the previous ones by some rule. Had the investigator instructed the subjects by saying: "a rule has been apply to the numbers 4,6,2 as a result of which the sequence 2,4,6 is obtained" and had asked the subjects to find such rule, the outcome of the experiment would had been entirely different.
Watson's example shows beautifully how the failure of the investigator to consider and test alternative hypothesis to his believe that people exhibit certain traits led to bias results. The congruence bias is exemplified by the investigator.
Baron's suggestions apply to Wason's investigator, particularly the second suggestion.
One more point: in Statistics, the probability of a positive result when the hypothesis tested is not true is referred to as the Typer II error. Good test have low Type I error (probability of negative result when the hypothesis tested is true) and low Type II results.
In Statistic notation, Baron's suggestions can be rephraced as:
1) mind Type II error
2) restrict alternative hypothesis to interesting cases
66.11.91.153 18:48, 9 November 2007 (UTC)10little