Can AI improve decision-making?

Humans appear to be terrible decision-makers. From the most intelligent orange man to senior executives of firms, it is clear that humans make bad decisions. In the corporate setting, decision-makers end up spending more time convincing others they are good at it than actually doing it. The principal-agent problem in the firm has worsened (1) as the agents hide behind technology and opaque organizational structures, pretending to be great.

It may be time to get out of it, for AI is likely a more efficient decision-maker. There are many reasons for this. First, decision-making requires analyses of complex and uncertain data and forecasting. Humans have great difficulty understanding uncertainty. It is unclear why this is so. Perhaps, in the African Savannah when they started-off, they did not have the luxury of multiple experiments and thus they never had a chance to internalize probabilities. Those who survived by random luck, and not by probabilistic decision-making, passed their genes on with the bias of a success certainty. Second, optimal decision-making requires an understanding of contingent claims, how future decisions depend on the present ones. Here again, the mighty human fails for her options have expiry windows or her political triumph is ephemeral, requiring yet another success in four or five years.

Humans appear to be terrible in making decisions in the presence of uncertainty and flexibility. Machines could be far superior.

(1) Flexibility: Flexible Companies for the Uncertain World — 1st Edition (routledge.com)