We have $f:R\to R, f(x+a)-f(x) \leq f(y+a)-f(y), \forall x,y,a \in R, x\leq y, a\ge 0$
We have to prove that the function satisfies the following:
$f((1-\frac{1}{2^k})x +\frac{1}{2^k}y) \leq (1-\frac{1}{2^k})f(x)+\frac{1}{2^k}f(y), \forall x,y \in R, \forall k$ positive integer. That inequality is reminiscent of the inequality for convex functions. Is there a connection? How could I approach the problem?
For given $x < z$ you can set $y = \frac{x+z}{2}$ and $a=\frac{z-x}{2}$ in $$ f(x+a)-f(x) \leq f(y+a)-f(y)$$ to get $$ 2 f(\frac{x+z}2) \le f(x) + f(z) \, , $$ i.e. $f$ is midpoint-convex. That is the desired inequality for $k=1$, and the general case follows by induction, since $$ (1-\frac{1}{2^{k+1}})x +\frac{1}{2^{k+1}}y = \frac 12 x + \frac 12 \left((1-\frac{1}{2^k})x +\frac{1}{2^k}y\right) $$
(For continuous functions, midpoint-convexity implies convexity.)