Humans discount future rewards hyperbolically (1/(1+kt)), not exponentially (e^−δt). This creates time-inconsistent preferences — you prefer A over B now, but prefer B over A when both are in the far future.