Statistical and other formal techniques for risk assessment are well established and their limitations are well understood. How would a crowd sourcing approach produce better results?