Thursday, April 17, 2014

Does the experiment actually show what it is being stated to show?

When I saw this headline, Babies prefer fairness – but only if it benefits them – in choosing a playmate by Molly McElroy, my initial thought was, "Just like adults." Most advocates making an argument based on social justice are happy to explore solutions to the perceived problem right up until the point when "fairness" means something other than taking from someone else.

Intrigued, I clicked through and found a rather interesting article. Not interesting for its findings but for its experimental flaws and the power of predicate assumptions. All the usual caveats of skepticism of psychology and small groups and unreal research conditions apply.

I completely accept that infants may demonstrate in-group behavior by race or any other identifiable attribute (gender, smell, accent, language, height, attire, etc.). But what the article is claiming is that this set of experiments shows infants demonstrate in-group preference based on race when I think it is possible, if the article is accurately reporting the facts, is that all it is doing is demonstrating infant rationality.

The crucial observation is that white infants, when given the choice between a white experimenter who has been seen to give a white recipient an unfair distribution of toys versus choosing a white experimenter who divides toys evenly between two recipients, one Asian and one White, the infants will choose the experimenter who gives more toys to the white recipient.

But that doesn't really tell us much at all about either race or fairness. What it appears to say (if I am reading the original results correctly) is that infants are utilitarian rationalists. If they want toys, they will pick those distributors that seem most likely to give either more or at least the same number of toys to the recipient who is of the same race as the child. In other words, the infant has a preformed expectation that recipient race has some predictive relationship to how they themselves will be treated.

You can see why psychologists get into so much trouble so quickly by looking at the original report. All the lab people (distributors and receivers) are young females, a mix of Asian and White. There are no old people and there are no males. How might that have changed the results? There were no Asian babies though the lab people were either Asian or White. How might that have changed the results? Have they controlled for color of clothes by the experimenters? It appears not. What about all the other observable attributes (height, smell, accent, etc.)? The even bigger problem is the small sample sizes. They start off with only 80 infants in the first place and then winnow it down by those that don't end up participating and then further shrinkage occurs because the range of combinations is so high (White/Asian Distributor, Fair/Unfair Distribution, White/Asian Recipient). The formulae are impressive and the reported p values are great but they are essentially meaningless given the small population sizes.

These are interesting research questions, but the results are being given credence far beyond what the circumstances permit and seem to be substantially predetermined by both the limitations of the researchers predicate assumptions and the limitations imposed by the smallness of the experimental construct.

No comments:

Post a Comment