Race & The Doll Study: Bias

The Anderson Cooper 360 program on CNN recently presented the findings of a new version of the classic doll study regarding race.  The gist of the study is that black and white children were presented with a color spectrum (in the form of cartoon children or an actual color spectrum from white to black) and asked various questions such as “who is the bad child?” and “who is the smart child?”. The general findings are not surprising: both white and black children proved to be more inclined to associate positive qualities to lighter skin colors and negative qualities to darker skin colors. In general, black children tended to show slightly less bias in these regards.

Those conducting the study did acknowledge that it is limited in size and hence the results must be extended to the general population with caution. However, the results do seem to be intuitively plausible and they can be checked by additional studies that will expand the sample size.

Aside from the matter of size, one of my main concerns with the study is that the approach taken has a built in biasing factor. As noted above, the children would be asked a question such as “which child is bad?”  or “which child is smart?” and this rather clearly directed them to pick a specific color. The children were not given a “none of the above”  (or “all of the above”) option and this is a biasing factor. Interesting, some children did create their own “all of the above” or “none of the above” options when asked such questions.  For example, when asked about which child he would want to be friends with, one of the children picked all of the colors and said, in effect, that he did not base friendship on color. Naturally, most children would not be aware that they had the option to do anything other than pick one color when asked. After all, the questions were worded in a way that guided children towards picking one color and, of course, the children were rather directly presented with a set of colors to chose from.

While this might not seem important, it can actually have an impact on the results. To be specific, when a child picks a specific color in response t0 a question, what the child might well be saying is “if I must pick a color, I will pick this one.” While this does provide some information about what the children think, it does so in a rather limited manner. To use an analogy, imagine if someone put three pies in front of  a person and asked the person “which pie do you not like?” The person might, in fact, like all the pies. However, if pressed to chose, she would probably pick the pie she liked less than the others. If she was then asked “which pie do you like?”, she would probably pick the one she liked the most, even though she might actually like all of them.  To conclude that the person liked the blueberry pie and disliked the apple pie based on this study would be an error. After all, the person was not given a clear option of selecting “all” or “none.”  Likewise for the color study. Of course, such studies can provide an indication of preferences and such information can be useful-provided that the limitations are properly acknowledged. While some people will break out of the limits on their own, most people will not. While this approach can provide interesting information (such as how many people will break out of the limits), it does create a bias by not allowing a full range of choices.