THE VALIDITY OF THE IMPLICIT ASSOCIATION TEST

There are two articles here and here that summarize some research by Indian psychologist Mazharin Banaji on a "covert" measure of racism called the Implicit Associations Test (IAT). A common finding from use of the test is that all sorts of people are quicker to pair "good" words with whites and "bad" words with blacks. The most surprising thing about the test is that many people who are conspicuously anti-racist show the same quickness to associate good with white.

Finding "covert" measures of anything -- and racism in particular -- has long been a "holy grail" for psychologists and there have been some conspicuous failures in the quest. So is the IAT the holy grail? Sadly, No. The first thing a psychometrician asks about any test of anything is: Is it valid? -- meaning, does it measure what it purports to measure? But there is another question logically prior to that: What does it purport to measure? And the answer in this case seems to be straightforward: It purports to ascertain whether a person has prejudiced, negative or antagonistic attitudes towards various minorities. That being so, the test is obviously NOT valid. It is not valid on what psychologists call a "criterion groups" examination. And it is precisely the feature of interest in the IAT that lots of people who are by any criterion either non-racist or actively anti-racist get high scores on racism according to the test. So the IAT does NOT pick out non-racist or anti-racist people accurately.

So does the test measure anything? Anybody who is familiar with the stereotyping literature will find that easily answered. There have now been many decades of research into stereotyping and the findings about it are roughly the opposite of what is popularly believed. There are two literature surveys here and here which document that. The important point for our present purposes is that stereotypes have long been found to have a "kernel of truth", as Allport put it. Far from being rigid or fixed, they are highly responsive to modification through fresh information. They are our first and most immediate response to any new situation -- but to be useful, they also have to be continually modified as information about the situation comes in -- and they are.

So what the IAT findings show is that the experience white people have of blacks is generally negative. Whites know from experience or observation that blacks in general are (for instance) more dangerous to them. Given the enormously disproportionate incidence of violent crime among blacks, it would be a sad day indeed if no-one had noticed that. So what the IAT measures is EXPECTATIONS of blacks, not ATTITUDES to blacks. It shows what we see as most probable about blacks but tells us nothing about any more complex attitudes we may have towards blacks. So the IAT simply records our experience of reality without telling us anything about how we interpret that reality.

That view of the IAT also explains why even many blacks associate badness with blacks. Blacks are of course the most frequent victims of black crime (for instance). Since it is very common for whites to shun blacks in various ways (no eye contact etc.) however, many blacks will still have most positive associations with their own kind. And the IAT shows that too.

There is an academic review article here (PDF) which also fairly effectively undermines the claims of the IAT as a measure of racially biased attitudes. See also here for comments on another study using the IAT.

***********************************