Hume relates a conversation he had with a friend concerning the opposition between speculative philosophy and religion. He suggests to his friend that philosophy in ancient Greece and Rome was far better able to flourish since it was less countered by religious superstition. His friend replies that an Epicurus, known for disavowing traditional religious beliefs, would still be able to defend himself reasonably against religious zealots. In order to prove his point Hume's friend presents an imaginary speech on Epicurus' behalf made to the people of Athens.
Hume's friend, as Epicurus, suggests that religious philosophers have strayed from their traditions, trying to prove the existence of God by means of reason. Such a question, Epicurus insists, is beyond the capacity of reason. His non- religious principles do not contradict religious tradition, but merely suggest that religious traditions are not founded in reason.
Religious philosophy argues for God's existence from design. It perceives a certain order in the universe that could not have arisen by accident, and so infers the existence of God as an ordering principle. This kind of reasoning-- inferring by experience from effect to cause--requires that we infer no qualities in the cause beyond what are requisite for producing the effect, and that we infer no further effects from this cause beyond what we have observed already. Religious philosophers often make the mistake of inferring the existence of God from what they observe, and then inferring other, unobserved, effects from this inference of God's existence. We can perhaps infer God's existence from what we observe in the world, but we cannot then infer some greater design or perfection in the world from our inference of God's existence. Philosophers who claim to do so proceed not from reason but from imagination.
Hume objects that sometimes it is perfectly acceptable to infer further effects from a cause that one has inferred from observed effects. For instance, if we see a footprint in the sand, we can infer that it was caused by a person walking on the beach. Furthermore, we can infer that this person must have left other footprints as well, which have since been erased by the waves. It seems perfectly reasonable to infer that other footprints must have existed, and yet this inference comes from a cause that itself has been inferred from other effects.
Hume's friend replies that, in the case of the footprints, we can infer that further footprints existed based on our wider knowledge of human behavior. We know a great deal more about people than what we learn from the one footprint-- for instance, that people have two feet and usually leave consecutive footprints. Once we infer that a person was walking across the sand, we can bring to bear our wider knowledge of people and infer further effects than the one we observe. Unlike people, however, we know nothing of God beyond what we observe. We may infer his existence from certain observed effects, but this inference will not help us infer any further effects than what we have already observed. Hume's friend asserts that God's divinity is too far beyond our comprehension for us to reasonably ascribe any human motives to him.
Hume also points out that reasoned inferences are drawn from observing the constant conjunction between species of effects and species of causes. The inference of the existence of God is singular in nature, and thus the inference might be regarded with some suspicion.
This section follows the previous section's attack on the religious metaphysics of rationalist philosophy. Hume's thrust continues to be that reason cannot lead us beyond what we find in experience. Our beliefs in miracles, in an afterlife, in some ultimate purpose in the universe, are all religious tenets that must be accepted or rejected on faith. We are not necessarily wrong to infer God's existence from what we observe in the world: this hypothesis is as good as any other. We go wrong, however, from then inferring further facts about the world based on this hypothesis.
God, according to Hume's account, is what we call an empty hypothesis. In order to clarify this term, perhaps we should first clarify Hume's discussion of our inferences from effects to causes and from causes back to effects. It is a common and reasonable practice to infer causes from effects. If I wake up in the morning and see that the roads are wet, I can safely infer that it rained during the night. Hume explains such inferences in terms of constant conjunction: when we find that one event habitually follows another event, we imagine a necessary connection between them in our minds. My past experience shows that the roads tend to be wet after it has rained, and usually remain dry otherwise, so my mind draws a connection between wet roads and rain.
I might also infer from cause to effect. For instance, if I see wet roads and infer that it rained during the night, I may also infer that the grass will be slippery. This inference is based on my experience of a constant conjunction between rain and slippery grass: After it has rained, and not otherwise, the grass tends to be slippery. Though I have not directly observed the rain, I have reasonably inferred it from the wetness of the road, and can reasonably infer also that the grass will be slippery.
My inference that the grass will be slippery is not inferred directly from the wetness of the road but from my more extensive knowledge of what happens when it rains. My experience with rain goes far beyond what I have inferred from the wet road, and once I have inferred that it has rained, I can draw on this more extensive experience in order to make further inferences.
Hume suggests that our inference of rain differs from our inference of God's existence in that we have regularly observed rain directly and come to associate a great many other things with it. On the other hand, we have never observed God directly, and all we know about him is drawn from the inferences we make. We know God only as the cause of the effects we ascribe to him. God is an empty hypothesis since he is posited only to explain certain phenomena that we might not otherwise be able to explain. We have no direct knowledge of him, and so can ascribe no qualities to him beyond those that we have observed in order to posit his existence in the first place. Since God is an empty hypothesis, we can say nothing about him beyond what we have observed in order to infer his existence in the first place.
Hume is treading on dangerous ground here, and he proceeds with caution. Rather than present these arguments as his own, he presents them as a friend's, and explicitly denies any responsibility for them. Further, the friend presents the arguments through the voice of Epicurus, distancing the argument from Hume one step further. There is constant reference in the discussion to the danger of religious philosophy and its confusion of religious tradition with a priori reasoning. Hume does not want to condemn religion so much as to keep its superstitious influence away from speculative philosophy.