Use statistics to measure what we hesitate to disclose

A few days ago, I was participating in an HR workshop and we were designing OKRs for a new onboarding process. The problem was to measure one of the key results: the satisfaction of the new employees. Our concern was that probationary employees would hesitate giving negative feedback on their organization during this trial period. How on earth could we measure their satisfaction?

I remembered one of my best friends advised me to watch this video:

Chat Sceptique describes a smart method allowing to measure with statistics behaviors we hesitate to disclose.

Here is an english translation of the method:

Instead of asking directly to the participant if he uses cannabis, I’ll ask him to roll one die secretly, so that I can’t see the result. If he rolls a six, then he will have to tell me that he is using cannabis, even if it’s a lie. If he rolls another number, then he will have to tell me honestly wether he is using cannabis or not.

[...] The participant is thus fully protected by the random result of the die he is rolling.

[...] From a statistical standpoint, results are blurred at an individual level, but not at the collective level.

If those using cannabis don’t dare to say it, how might we know?

And the original french subtitles:

Plutôt que de platement demander à chaque participant s’il consomme du cannabis, je vais lui demander de lancer un dé à l’abri des regards indiscrets, en particulier du mien. Si le chiffre 6 est obtenu, la personne doit me répondre que oui, elle consomme du cannabis, même si c'est un mensonge. Si un autre chiffre a été obtenu, alors elle me dit juste honnêtement si oui ou non elle consomme du cannabis.

[...] Le participant est complètement protégé par le résultat imprévisible du dé qu’il lance.

[...] D’un point du vue statistique les pistes sont donc brouillées au niveau individuel, mais pas au niveau collectif.

Si ceux qui consomment n’osent pas le dire, comment savoir ?

Thank you Michel for this great video!