Monday, 25 January 2016

I am not a number

I've just read The End of Average for review, and I couldn't help letting out a little whoop of joy when it totally trashed psychometric testing.

I am talking about mechanisms like the Myers Briggs type profile, along with a whole host of rivals, all used by businesses in recruiting and team building to analyse a personality and assess how an individual will work with others. 

The problems I have always had with the approach are several-fold. It's based primarily on Jungian theory which has little scientific basis. Your personality type is self-determined, so, while it's not surprising it often feels right, that doesn't make it accurate. And I was always doubtful about the cultural norms of the mostly US-devised tests being applied worldwide. Infamously there used to be a question about whether you preferred a gun or something constructive (I can't remember what) - which clearly would have different resonance in the US and Europe. 

Now, though, there are much stronger grounds for concern. The End of Average points out that personality profiles don't reflect the behaviour of individuals, but rather they predict the average behaviour of a group of people, which isn't the same thing. If you are an ENTP like me, it doesn't say how you will behave, but how, on average, people with the same profile will behave. As the book says 'In fact, correlations between personality traits and behaviours that should be related - such as aggression and getting into fights, or extroversion and going to parties - are rarely stronger than 0.3.' The same applies to academic achievement and professional accomplishments. This means your personality traits, as identified by the test should reflect around 9 per cent of your actual behaviour, while getting over 90 per cent wrong.

Underlying this is the relatively recent (if entirely obvious) discovery that we don't have one personality/behaviour but it varies depending on the situation. A teenager, for instance, behaves very differently with a group of peers and with his or her grandmother. That's obvious. So why do we expect a single score on a handful of dimensions to reflect how we will behave in all circumstances? It's bizarre.

I don't expect companies to stop using these tests any time soon. Come on - some still use 'graphology', expecting handwriting to give insights into personality. But employers and academics should at least be thinking twice about what they are testing and why.


  1. If you think science can help organisations make better evidence based decisions, it would be wiser to use the latest research in psychometrics to inform best practice. On the other hand, if you don't want to base decisions on statistically significant correlations of 0.3-0.5, then you had better be prepared to abandon ibuprofen (r=.14 with pain relief, n=8,488) and ... WAIT FOR IT .... abandon VIAGRA (r=.38, with improved sexual functioning, n=779). STEWART DESSON.

    1. You need to take up the stats with the author. But it's hard to deny his point that personality is not fixed but varies with context. Also I think the comparison with drugs is a little spurious. You can tell if taking viagra has an impact (I'm told). You can't tell if a personality test is valid in the same way

  2. Brian Clegg .... the TED talk by Todd Rose is fantastic. It is relevant to the risk of boxing and using “averages” for personality.

    The talk looks at cockpit size and the myth of the “average pilot”. Very relevant to what we do in the realm of personality.

    1. Stewart - the book I am quoting is by Todd Rose - all that stuff is in it. (See review.)