Our Beliefs will affect our behavior directly or indirectly, so the question of what to believe seems to be a matter of ethics. We often say in everyday language that one ought to believe something or not. This linguistic habit of applying the word ″ought″ to the mental state of believing seems to prove that there is indeed some kind of a doxastic obligation. According to the principle of ″ought implies can″, to say we ought to believe something means that we can control the forming of belief. But in fact we don’t seem to be able to control what we believe. So we are faced with three theses that seem to be reasonable singularly but not reasonable jointly. One is the thesis of moral judgment, which states that sometimes we really should or should not believe certain things. The second is the thesis of voluntary control, that is,″ought implies can″. The third is the thesis of non-autonomy, that is, the forming of belief is involuntary. These theses constitute the thing that is called the paradox of doxastic obligation. This paradox can be directly converted into an argument against the ethics of belief: (1) If there is the ethics of belief, then we can control our beliefs. (2) We cannot control our beliefs. (3) So there can be no ethics of belief. How should we evaluate this argument, or how should we solve the paradox of doxastic obligation? There are five different views in the academia. The first is the view of direct voluntarism that we can directly control the forming of beliefs. In this view, we can believe something at will directly. There are three necessary and sufficient conditions for direct voluntarism: One is the act of will, that is, the forming of belief can be an act of will. The second is about full awareness, that is, we are fully conscious of what we are doing in acquiring a belief. Third, acquiring a belief can be a process that is independent of evidence or in disregard of the requirements of evidence. If the view of direct voluntarism can be established, then the thesis of non-autonomy in the paradox of doxastic obligation is wrong. Thus the argument against the ethics of belief is incorrect. And then there is a view of involuntarism, which treats direct voluntarism as a false theory. In fact, we can’t directly control what we believe. To believe at will means that even if you are fully aware that a proposition is false, you can still believe it by an act of will. However, this is psychologically and logically impossible. If the view of involuntarism is correct, and the voluntary control is a precondition for the ethics of belief, then the ethics of belief is impossible. If the ethics of belief does not require voluntary control, then the ethics of belief can still be established in some way. The third is the view of indirect voluntarism, that is, some beliefs are indirectly derived from will or intention, and the ethics of belief does not need direct control of will, and indirect control is sufficient to ensure the existence of the ethics of belief. That is to say, the thesis of voluntary control in the paradox of doxastic obligation is excessively rigorous. Fourth, the view of the doxastic compatibilism that the will can’t control the forming of belief is completely compatible with doxastic obligation. Doxastic obligation may actually be a kind of role obligation, epistemic ideal, rule of criticism or social demand, so we do not need to have any control over the forming of beliefs. Finally, the view of negative voluntarism holds that people’s will can control the elimination of belief, but not the acquisition of belief. Therefore, we can still have a kind of ethics of belief on the condition that the thesis of voluntary control is admitted. The view of direct voluntarism is indeed wrong. Indirect voluntarism can accept reasonable insights from involuntarism, doxastic compatibilism, and negative voluntarism. It can provide a solution to the paradox of doxastic obligation and lay a foundation for the ethics of belief.