When operating in complex environments and faced with a set of possible actions that all have completely unknown and unknowable impact, the operating procedure is to minimize the known costs. This can sometimes mean that doing nothing is the optimal choice, since there are usually transaction costs associated with every action and we should minimize them along with other known costs, if any.
But there are plenty of biases that will prevent people from picking inaction even in the face of massive transaction costs. This is especially true in environments where results are difficult to associate with specific actions. Looking busy and picking an action, any action, is what we are incentivized for.
This applies to picking models as well - not relying on a model at all may be the best way to go in some domains.
Post a Comment