Friday, January 16, 2015

Looking for talented people prevents you from finding talented people?

An interesting paper about under-representation of women and racial minorities in science:

Here's a nice coverage in the Economist:

Essentially, they asked people from different fields whether it is important to have an "innate talent" to succeed in those fields. Then then correlated these answers to under-representation of women and racial minorities in these fields, and found some very strong correlations.

It suggests an extremely interesting thing. The more you believe in "innate talent", in this mysterious hidden property that surely exists, but cannot be measured directly, the more biased you become. Because you cannot measure talent, right? So you have to go by proxy. But if you believe in the importance of talent strongly, you got stuck with wrong proxies, such as gender, citizenship or ethnicity. You do it unconsciously, but you are trying to infer hidden properties from patterns, and got caught in a Bayesian trap (see the plots below).

The trick of finding true talent therefore is in downplaying the importance of talent! One should try to avoid all guesswork and trust the empirical evidence, the phenomenology. Judge on the deeds, on the results, not on motivation, interest, "fit", or whatever. Sounds obvious, but the data shows how non-obvious it actually is in real life, or rather how poorly people follow these truisms in practice.

I actually find it very relevant for my teaching. Every now and then I see a student who does not fit my subconscious image of a "good student". They are just interested in "wrong things", or spend time in a "wrong way", or maybe talk, write, code, or draw in a way that shows that they lack any talent for the trade! Or do they? It takes a very deliberate and conscious effort to stop this Bayesian profiling and give everybody a solid chance. And boy how happy am I when I turn to be wrong! When somebody whom I suspected of cheating turns out an honest student, just an unusual one. Unusual for me, mind it, with my limited life experience. For me it's never about gender or race, thankfully, but rather about interests and style of thinking, but still the psychological dynamics is probably pretty similar. There is always a risk of subconscious profiling; of shortcuts in judgment; wrong generalizations that have the potential to become self-fulfilling prophecies. They are always at risk of creeping into our practices unless we relentlessly banish and weed them out, day after day, again and again.

To sum up, to find true talent, don't try to intuit it; give up on guesswork, and go for facts instead. Sounds simple, right?