Project Implicit

So I heard about Project Implicit run by some Harvard people from an episode of Scientic American Frontiers (5 years old, actually).  I believe they study unconscious connections in our minds by measuring how the time delay to make various associations.

The setup usually goes like this: they first train you in the terms (associate these terms with Good, these with Bad, with Forward, with Backward, etc.).  Then they’ll put, say, Good and Forward on the left side of the screen and Bad and Backward on the right side.  A word will pop up, and you’re supposed to assign them as fast as possible.  Then, for comparison, they’ll do a second set but pair Good-Backward and Bad-Forward.

So far, it has told me:

  • Your data suggest a strong automatic preference for Forward compared to Backward.  (I took a similar test later that night and instead got “moderate automatic preference.”)
  • Your data suggest a slight automatic preference for Change compared to Preserve.  (A couple days later, I got “no automatic preference.”)
  • Your data suggest a strong automatic association of Religion with False and Science with True.
  • Your data suggest little to no automatic preference between Straight People and Gay People.
  • Your data suggest a slight automatic association for 2050 with Safety and 1950 with Danger compared to 1950 with Safety and 2050 with Danger.
  • Your data suggest a moderate automatic positive association with Society.
  • You data suggest a strong automatic preference for Phosphate detergents compared to Non-Phosphate detergents.  (This was after reading a blurb from a trusted information source, like a government consumer protection agency, instead of from a detergent company.)
  • Your data suggest a strong automatic preference for white compared to black.

I think these results have noise, as they seem to affected by fatigue (I did most of these around 1am) and the order of experiments (i.e. if you’ve learned that Good is left, Bad is right, you’ll press them a little faster in the second set of experiments).  However, assuming they know what they’re doing (they have training and calibration runs), the above seem to be consistent with an educated, moderately liberal, agnostic.

Except that last one.  So apparently I’m racist, at least subconsciously, which is what these experiments were designed to study.  Luckily, our brains have a governor which can override such thoughts from becoming reality.