Project Implicit.

[Jeunemaitre]

For those that took the political compass test and didn't like it, here is another set of tests that will give you results that you don't want to see:
[right][snapback]94792[/snapback][/right]

As someone who designed a similar reaction-time based study (albeit for a psychology research methods class several years ago) the problem that Doc experienced is likely due to a confounding element in the study design. The long explanation is below, but the short of it is that presenting individual results is virtually meaningless because you can't determine what amount of the increase between the expected match and the incongruous match is based on preference, and what amount is based on order in which the sets are presented (when you get used to hitting one key for one set and have to change that association).

Psychologists, Economists, and Astrologers are all at the same level of 'science'... [snip]

As someone with a degree in psychology, who currently works in economic research, I'll grant you that psychologists and economists are on the same level, but I have to object to the grouping with astrologers. Yes, you can misuse much of what is determined from psychologic or economic studies, but when you look at the specific setting from which the results are derived, you can generally point out what people are getting wrong. With astrologers, you can't help but point at everything as being wrong.


Association task confound explanation:
The basis of the conclusion is the difference in reaction time between what the computer determines as your expected preference match and the incongruous match. In the series that I did, I compared Apple to Microsoft: in the preliminary list of questions, I indicated a slight preference for Apple, so the computer set up the "expected match" set of frames as Apple + Positive vs. Microsoft + Negative. The analysis therefore compared my reaction time in this setting vs. my reaction time in Apple + Negative vs. Microsoft + Positive. My reaction time results indicated a strong preference for Apple vs. Microsoft, but the confounding order effect is likely the source of the result. Because the Apple + Positive frames were presented first I made the association I=apple; I=Positive. When the categories switch to the incongruous (Apple + Negative) I had to change to I=apple; E=positive. This change is likely the result in the increase in reaction time, not the latent association Microsoft=Negative. The only way to address this change is to randomize participants to the order in which the see the association frames (whether the expected set comes first, or the incongruous set comes first). Basically, the only way to get results out of this is to look across a large population and control for the size of the change when the series flips in order to make a statement about the preferences of the group. When done on an individual basis, it is virtually meaningless.