A large portion of the recent literature on uncertainty quantification has focused on second-order risk minimization, renaming it Evidential Deep Learning. We argue that this naming is misleading, as the naming of the evidential framework has already been established by the Mathematical Theory of Evidence, which has been applied to numerous problems for decades and deep evidential learning was defined long before second-order risk minimization. However, this naming overlap has given rise to particularly interesting ideas for (evidential learning in the sense of) second-order risk minimization, which is applied here both theoretically and empirically to (evidential learning in the sense of) Dempster-Shafer Theory, establishing a rule of monotonicity for epistemic uncertainty as essential to evidential learning. This rule is proven to be true in this paper for Dempster-Shafer Theory, whereas it has been proven impossible for second-order risk minimization. As a result, our work clarifies the distinctions between these approaches while contributing new insights to the broader discussion on uncertainty quantification.