Should research combine explicit and implicit?
The Implicit Association Test (IAT) was developed by psychologists from Harvard, University of Virginia, and the University of Washington to explore unconscious roots of thinking and feeling1. Many of the initial applications of the test involved understanding social prejudices in the community around race, gender, and sexual orientation, among others.
IAT has been widely used in scientific research, development & education, employment etc., Over the last few years, IAT has been successfully combined with explicit testing methods in areas like brand and marketing, human resources training and fair practices in employment among many others2.
Researching social prejudices and biases
We recently conducted one such test combining both explicit and implicit methods. The objective of the research was to understand social prejudices and stereotypes in the workplace. More than 1,500 test respondents in Singapore were asked to complete an IAT associating workplace behaviours and characteristics with the images below. Subsequently, they were asked to answer a series of questions relating to commonly held age stereotypes.
Prime scores were calculated and have been shown mapped in figure 1.1. The neutral range is represented by the grey bar. The data points that fall outside the neutral zone have a strong association with the behaviour/stereotype tested—above the grey bar “Is X” and below the grey bar “Is not X”.
Overall, the younger images have a stronger association with being quick and efficient while the older images (especially the older woman) are seen to be flexible. Directionally, the younger-male image has a stronger association with being wise than the other images. All other characteristics fall largely within the neutral range suggesting that these do not have strong positive or negative associations.
In addition, we also noted some gender and age biases. The 40+ aged audience associates the older images more strongly with “is not motivated”. More female respondents tend to pick the younger images as efficient and quick compared to male respondents.
Similar questions were posed to the participants in the explicit test, and they were asked to express the extent of their agreement/disagreement with each of them. Fig 1.3 shows the six statements and the overall views that the participants expressed.
By and large, we don’t see strong agreement/disagreement with any of the statements. More than a third agree that older employees are seen more insightful while younger employees are faster or quicker.
So where do the big differences lie?
The implicit association test reveals some biases or stereotypes that are not as easy to find in the explicit survey. In the implicit test, the older images are seen to be more flexible whereas the explicit test doesn’t reveal the same evidence3.
Gender biases are also not as clear from the explicit test. When asked explicitly whether they felt younger employees were faster workers than older employees, female participants agreed to the same degree as male participants. However, the IAT reveals that females are more inclined to pick the younger images as being quick.
This does not mean that implicit and explicit tests always reveal contradictory results. What this tells us is that in certain situations, the IAT can provide a perspective that explicit methods are probably not the best suited for.
So, what type of test should I choose?
The answer to this question depends on the objectives of the research. If the research requires its participants to evaluate stimulus/react to statements/situations which may carry some inherent biases, it may be useful to incorporate some implicit testing along with traditional explicit surveys.
Although the focal questions of the implicit and explicit phases of the research can be kept largely independent of each other, it may be useful to gather additional diagnostic information in the explicit survey to help validate/invalidate the results from the implicit test.
As a leaving thought, consolidating the insights from both methods, we feel, is far more beneficial than inferring from either method alone.