Tagged : IAT

Which implicit test is best?

December 13th, 2013 by

At COG we have been working with implicit testing for 8 years now, in collaboration with the developers of the original IAT test at Harvard, adapting that for commercial use, building a simpler IRT test to measure multiple brand values, and adapting that to measure social attitudes. We are lucky to have had clients who have supported us in testing and developing these alternative approaches and thought it useful to share a few early findings:

IF you want to get at deep seated social attitudes and prejudices then the original IAT test is still best. It is technically robust and can prove negative or positive associations in a way that convinces almost everyone. That will be why Adam and Eve DDB and ourselves picked up an APG Strategy Award last month for the work we did on facial disfigurement for Changing Faces.

IF you want a more user friendly way of measuring social attitudes and feelings then the Social IRT test is best – this uses speed of response to identify which issues we feel strongly about (either positively or negatively) and is quicker and more user friendly than the classic Likert scale approach. We used this recently with The Guardian on their acclaimed Concerns and Contentment work.

IF you want to track brand attitudes and the effect that digital communication, sponsorship or good old fashioned ads are having – then the IRT is great. In a 2 minute test you can measure your brand and 3 or 4 competitors on 15 benchmark values to produce a sensitive map of how you are doing.

IF you want to measure other things than brands, then the IRT can still help – we have recently used it to monitor attitudes to sufferers of different health conditions.

There is a lot of interest in Implicit Testing at the moment – because it is measuring the part of our mind that is influencing most of our behaviour and purchase decisions. We can only expect implicit tests to become more popular in future – but one thing we have learnt is that they need real design care to get the algorithms right. Just because they look simple doesn’t mean they are easy to build. But if they are well built they give wonderfully clear and intuitively believable results that don’t need a PhD to interpret.