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.