Talk:Congruence bias

Vs. Confirmation Bias
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 agree, I've updated the lead to reflect that7804j (talk) 14:43, 29 July 2020 (UTC)

Wason's example not clear
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)
 * To guess the rule and check it by telling the researcher new sequences, he then confirms or denies their validity according to his original rule. The point is that people tend to check only sequences generated by the guessed rule and do not check other sequences. After 3 5 7 they could have checked 3 5 8, for example and realized their rule is wrong.
 * Isn't this example the same on confirmation bias ? (but there the example is better written, IMO) WendelScardua 20:39, 28 July 2006 (UTC)
 * I was left a bit cool by the number series example as well. To me it seems to show that the subjects expected the conventional sort of number series one finds in tests, and then became confused by a non-verbose indication that they were wrong and by an unusally 'simple' correct response.


 * I understand what is going on, but I actually thought the button example, however simple, was more to the point.24.222.232.244 15:32, 17 December 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)


 * The word 'sequence' has a specific meaning in Maths. I would expect the rule to be able to uniquely predict the next element of the sequence : 'bigger' fails. People are finding a better pattern. The problem would be fairer if no example were given. In the 2-button example, there could be a state-dependent mechanism so that the action of one button was not independent of the other - like light-switches at the top & bottom of a staircase !


 * The bit about 'testing' with 3,5,7 needs clarification.

--195.137.93.171 (talk) 11:58, 1 September 2009 (UTC)

button
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)


 * Point? 24.222.232.244 15:37, 17 December 2006 (UTC)

Reinterpretation of Wason's example
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