A study uses a paired $t$-test to compare measurements taken on the same subjects before and after a treatment. If $d_i$ is the difference for the $i$-th subject and $\bar{d}$ is the mean of these differences, what is the null hypothesis?