Wason’s Task and Confirmation Bias

Chris Duerschner

Confirmation bias shows up pretty much wherever you find opinions.  Some would even say that opinions are only possible because of confirmation bias, or at least a precursor to it.  The basic gist of the concept is that people tend to filter out those things that conflict with their convictions and only perceive those that reinforce them.  The focus of this post is not, however, to explore the extent to which confirmation bias occurs or to detail its underlying mechanism, but to explain the experimental evidence that shows it does occur.  Because confirmation bias is such a simple event; it makes sense that the most well known study that demonstrates it is also simple. 

This study has become known colloquially as Wason’s 2-4-6 Task.  The task focuses on how people generate hypotheses.  Its general aim is to “discover an underlying rule that specifies the relationship between three numbers.” (J. Russo, M. Meloy).  For example, for the three numbers used in the test (2, 4, 6) the rule might be even numbers, numbers separated by 2, or a variety of other categories.  The subject could formulate hypotheses and test them by putting forward a series of numbers that demonstrated it and receiving a yes or no answer.  The simplicity of the task is misleading as most people’s final hypothesis is incorrect.  To explain these results, Wason noticed that participants typically sought positive feedback and would grow increasingly confident in their hypotheses after several consecutive yeses had been obtained.  Few, however, would attempt to falsify their hypotheses, causing most to end up with an incorrect conclusion.

For example, given the initial sequence 2-4-6, a participant might formulate the hypothesis “even numbers”.  They might test this by putting forward the sequences 8-10-12 (response yes), and then 14-2-20 (response no).  They might then revise their hypothesis to be increasing even numbers and test it with the sequences 4-6-8 (response yes), 10-18-26 (response no).  They would then revise again to make the hypothesis consecutive increasing even numbers, and test with 12-14-16, 16-18-20, 32, 34-36 (responses all yes) until they felt sufficiently confident that they had discovered the pattern.  Unfortunately the true pattern is simply increasing numbers.  As you may have noticed in the example, it was after exceptions to the rule had been discovered that the participant’s hypotheses were revised, and if the participant had continued to seek falsification, rather than verification, they might have eventually discovered the correct rule.  Indeed in the example, I was overly kind to the participant, as in the actual trials most participants never eliminated a single hypothesis.

            The conclusions drawn from this study are broad and far-reaching. People naturally tend to seek evidence that they are right and seldom go out of their way to look for evidence that they are wrong.  In social situations in which the issues on the line are more important than a simple series of numbers, this may be because individuals are constantly assessing the costs associated with being wrong and modeling their attitudes and behaviors accordingly; in other words they are formulating a so called path of least resistance for themselves.  When put in terms of polarization, this explains many different aspects of current behavior on the internet, such as the overall popularity of Hess’s idiot opponent montages, the relative hemophilia throughout its echo chambers and the selection that is exhibited in the media that an individual experiences online.