Hume's friend starts out by reviewing the argument from design. It is ‘an argument drawn from effects to causes.’ Advocates of the design argument claim that it establishes no more than is warranted by the facts.
The friend then lays out rules for inference from effects to cause: (1) the cause must be proportional to the effect; no qualities may be assigned to the cause beyond what are necessary and sufficient to bring about the observed effect. Anything more is conjecture. (2) No further effects may be inferred from the cause beyond what is already known. The existence and order of the universe shows the power, intelligence and benevolence of God only to the degree that those qualities are displayed in his creation. Any further attribution of qualities is hypothesis. We cannot ‘mount up to Jupiter’ from effects to causes and then ‘descend downwards’ to new effects. Any attribution of other qualities or effects has no foundation either in reason or experience. In other words, an inductive argument of this sort for the existence of God can establish no more facts about the purported deity than are necessary to account for those features of the world that the positing of God is needed to explain.
So, even granting that the existence of the universe requires a cause of some kind, is it reasonable to look at the world and contend that it demonstrates any of the qualities that should lead us to believe that the creator (whatever it was) was an intelligent, benevolent designer as we're instructed by traditional religion to believe? There is no prima facie reason to believe that the world is just—injustice is clearly all around us, in fact.
The defender of the ultimate justice of the universe might point to future reward and punishment as evidence of the justice of the world, and thus to the existence and justice of its creator. The idea here would be that although justice does not obtain in the present world, God will balance the books in the future. However, the rules of inference from effect to cause only give us the right to attribute the justice of the visible world to its creator. The friend poses this question: is there distributive justice in this world? If not, then the absence of justice in the world does not give us warrant to expect it in the future; in fact, it should lead us to the opposite conclusion.
Experience serves as the only standard by which to judge a claim about the world; any inference beyond what experience provides is unwarranted. Having found these limits unacceptable, some have allowed speculation and imagination to carry them to concepts beyond what experience provides. One of these, that there is an intelligent, benevolent cause of the universe, is ‘uncertain’ and ‘useless,’ in the words of Hume's friend. It is uncertain because it lies beyond experience, and it is useless because the inference to an intelligent cause will not allow us to infer anything further than its mere existence.
Hume puts forward a possible objection that an advocate of the Design Argument might raise, though. Is it truly the case that we cannot infer anything further about the creator than simply that he created the world? We are certainly permitted to infer the existence of a builder if we happen upon a house partially built, and once we infer a builder, are there not many things that we can also affirm about the builder? There are many things that we know about builders, after all. It is interesting here that Hume's objection anticipates the sort of claim that is often made by advocates of the Design Argument and other arguments for the existence of God. If, once we've established a creator, we can come to knowledge of the other properties of that creator (by reason, revelation, etc.) then we can build upon the foundation of the inductive Design Argument. Perhaps we can indeed 'mount up to Jupiter' and come back down again.
The friend replies that the difference is that in the case of the house we are familiar with men as builders from other experiences, and have seen that houses are built by men. If we had no such experience of builders, then we would not be able to make the inference from the existence of a house to existence of a builder. In that case the situations would be analogous, but we could not argue from effect to cause and then make inferences about the other properties of that cause.
Once again, for Hume experience is the sole foundation for knowledge, and any attempts to put flesh on the skeleton of the notion of a creator arrived at through the Design Argument is merely speculation. We have no independent experience of the deity, who is known only by his products, so we cannot infer any further attributes beyond those necessary for the creation of the world. We have no license to suppose the deity possess qualities like goodness or justice in any measure beyond what is observable in the world.
Anything further is speculative, and the source of our tendency to attribute additional qualities to God is a projection of our own qualities, and our belief that God will reward and punish in a future life is derived from this anthropomorphic projection. In reality, we cannot say anything more about that deity’s nature than is included in his works. No new fact can be inferred from the ‘religious hypothesis’ beyond what is given in experience.
An implication that Hume's argument carries, but which he does not bring to the surface explicitly (probably for political reasons) is this: if Hume is right about the limits of the Design Argument that served as the centerpiece for natural theology's attempts to establish the truth of traditional religion, then if we accept Hume's attack on revealed religion in his discussion of miracles in the prior section of the first Enquiry, the foundations of traditional religious belief are destroyed.
Hume, D. An Enquiry Concerning Human Understanding (1772), ed. Tom L. Beauchamp. Oxford: Oxford University Press, 1999.