Moore, of course, thought that intrinsic value is the central evaluative property. Admittedly, specifying what counts as an intrinsic property is an interesting and difficult question in itself. Let me offer just a quick sketch. On this proposal, whether an object has an intrinsic property cannot be affected by anything outside the object. This entails that all possible duplicates of an object have all the same intrinsic properties as the object but they can differ in all their other properties that are thus extrinsic. Metaphysically speaking, Moore then thought that the intrinsic property of intrinsic value is a ‘non-natural’ property. Again, it is an interesting and difficult question what this actually means – what the difference between non-natural and natural properties is exactly supposed to amount to. As this doesn’t matter below, I’ll be neutral about this here.
One classic objection to non-naturalism about intrinsic value is an epistemic challenge. As John Mackie put it, ‘[c]orrespondingly, if we were aware of them, it would have to be by some special faculty of moral perception or intuition, utterly different from our ordinary ways of knowing everything else.’ So, traditionally, it has been thought that it is the non-naturalist aspect of Moore’s view that creates the epistemic problems. I want to suggest that this is false because the problem really is in the intrinsic aspect of Moore's properties. David Lewis, in his 2002 Gareth Evans Memorial lecture ‘Ramseyan Humility’, offers a strong argument to the conclusion that all intrinsic properties are unknowable – whether they be non-natural or natural properties (I wish I had read this earlier… most people probably know this well). If this is right (and I think it is), then naturalists about intrinsic value face the very same epistemic problems as the non-naturalists about intrinsic value.
Here’s a quick attempt to explain Lewis’ argument (much influenced by Rae Langton’s admirably clear discussion and also by Dustin Locke’s presentation – also sorry if I get this wrong). Let’s start from the idea that science can help us investigate the ‘nomological structure’ of the world. This is to say that science informs us about the roles which different properties of objects play. It’s part of the role of ‘mass’ that it curves the space-time and makes objects resist forces in acceleration. So, science can tell us that there is a property which has a certain extrinsic role. However, this does not yet give us information about the intrinsic nature of the property that plays this external role. The question then is, could we also know what the intrinsic properties of the objects are like – the properties that realize the roles described in science? Here’s Lewis’ argument to the ‘no’ conclusion.
Lewis starts by imagining a situation in which we have a true and complete, ‘final’ theory of the world. Amongst other things, this theory T describes the roles of all the fundamental properties of the world that determine how the world works. These properties are the intrinsic properties of the objects of the actual world on which all other intrinsic properties supervene. Furthermore, T does not mention any ‘idlers’ – instantiated fundamental properties that have no consequences or any ‘aliens’ – other uninstantiated and merely possible fundamental properties.
Theory T defines its theoretical terms in terms of the causal roles played by the properties for which these terms stand. Some of the terms of this theory thus come to stand for the fundamental properties of our world. Those properties are then the unique actual realization of T. Of course, in addition to T’s language, we also have our own natural language which Lewis calls the ‘old language’ or o-language. We can describe all our observations completely in the o-language.
T is, in effect, a long conjunction of sentences (presumably including laws and state-descriptions). It can thus be ‘Ramseyfied’. In this process, we replace all the names of the fundamental properties in the theory by existentially quantified variables. So, we get, “For some x1, …, for some xn T (x1, …, x2)”. This Ramsey sentence says that there is at least one actual realization of T. Given that T was the correct theory of the world and we got the Ramsey sentence from T by substituting names with existentially qualified variables, just like T, the Ramsey sentence entails all the observation sentences of the old language that were derivable from T. Therefore, the predictive success of the Ramsey sentence in question is exactly the same as T’s. This means that, as Lewis puts it, ‘there is no way to gain evidence for T that is not equally evidence for the Ramsey sentence.’
We then get to the crucial stage of the argument. Assume that in the actual world, there is a set of fundamental intrinsic properties <P1, …, P2> that realize the Ramsey sentence of T. Lewis calls this the actual realization of T. Assume also that there is another possible world in which another set of fundamental intrinsic properties <Q1, …, Qn> realizes the Ramsey sentence of T. This set is a possible realization of T. The point of Lewis’ argument is that, if there are many possible realizations of T (including the actual one), then no possible observation could ever tell us which one of the possible realizations of T is the actual one. Why? This is because whichever realization of T is true, the Ramsey sentence is still true and, as we saw, that Ramsey sentence itself already entails all our observations described in the o-language. So, no possible observation can get us beyond the Ramsey sentence which still leaves the question of the realizer properties open.
The last part of Lewis’ argument is to show that there must be multiple realizations of T. Lewis offers two arguments to this conclusion. The permutation argument claims that we could permute two fundamental basic properties and the laws that govern them whilst leaving everything else fixed. We would then get a different realization of T and it is impossible to rule this realization out as the actual one on the basis of what we observe. The second argument claims that we could replace any actual fundamental property with an idler or an alien. If we made the matching changes to the laws that govern these properties, we would again have another realization of T that would be indistinguishable to us observationally.
To summarise, if Lewis’ argument works, then we can never know which intrinsic properties objects have – be they natural or nonnatural properties. Lewis’ argument is about fundamental properties of the basic properties, but as he shows, given supervenience, this argument can be extended to other intrinsic properties of more complex objects. If this is right, then naturalism itself does not solve the epistemic problems of intrinsic value. Now, of course, there are ways to resist Lewis’ argument. You could claim that it relies on a mistaken infallibilist notion of knowledge or that there are some non-scientific ways to know the intrinsic properties. My hunch, however, is that all these ways to save the knowledge of intrinsic natural properties will also work for the non-naturalist properties. Or, we might think that this argument leads to structuralist or dispositional views of properties. But, this would be bad news for intrinsic value generally.