I friend from Uni sent me this interesting talk about decision making given by Gerd Gigerenzer (wiki) discussing how heuristics can be very successful at decision making under uncertainty:

One example he gives is of allocating the assets in your portfolio. If you actually knew the true risks, you could use what appears to be a mathematically optimal procedure (mean-variance optimisation) to pick the best portfolio – that is, it would be mathematically optimal if you could quantify the true risks.

But, as it turns out, the heuristic of simply dividing your money equally between your 1/N asset choices – stocks, for example – usually gives a better result, because the variance in the parameter estimation needed for the mathematically optimal procedure means it’s actually a worse method under uncertainty! At least, that’s how I understood it.