Suppose that $U$ is a uniformly distributed (continuous) random variable on $[0,1]$. Let's say that I am interested in finding 3 discrete points $u_1,u_2,u_3$ which approximate $U$ in some sense. My intuition tells me that to obtain these points, I would first partition the interval into 3 equal parts and select $u_1,u_2,u_3$ as the midpoint of these parts. In other words, $u_1 = 1/6, u_2 = 1/2, u_3 = 5/6$ constitute an approximation to $U$. (Each point would then have a probability of 1/3).
Upon discussing with someone, I was asked why $u_1 = 0, u_2 = 0.5, u_3 = 1$ (with equal probabilities) cannot constitute such an approximation. My initial response is that it contains the endpoints however I cannot rigorously quantify why these set of points is not "uniformly distributed".
Any thoughts to clarify this issue?