see the image:

To test for convex preferences, the following theorem is useful. Theorem. A twice-differentiable utility function U(x,y) represents convex preferences if and only if (1) UU, -2U UU +U U.