By moral subjectivism I mean the view that moral properties are wholly comprised of subjective properties. That is, properties such as being valued, being regretted or being something that one would regret, being believed to be right, being taboo or prohibited by society, and so on. Subjective properties are those that relate to people's attitudes.
The most obvious reason to reject moral subjectivism is that moral properties do not seem to supervene on or be determined by subjective properties. It doesn't seem that people can always make a difference to which moral properties are instantiated by simply having different attitudes. Nor does it seem that a difference in the moral properties always requires there being a difference in people's attitudes. This sort of relativistic subjectivism seems clearly false.
For the most part subjectivism and relativism go hand in hand. But, in fact, we can tease out sort of subjectivism that doesn't render morality relative to people's attitudes in the way that I have rejected above. How can this be? In short, by adopting a non-reductive subjectivism wherein moral properties are wholly comprised of subjective properties, but do not reduce to subjective properties.
To see how this might be let's first take a look at the ontology of species. We might ask, what is it to be a tiger? We cannot give a definite answer in terms of non-contingent, necessary and sufficient conditions. If we could then that would raise an evolutionary paradox. Such conditions would either apply or not apply to any animal in a tigers ancestral lineage. But we know that if you trace back the lineage of a tiger far enough, you'll eventually get an animal that is not itself a tiger. And so it follows that there must have been a first tiger, born from a non-tiger. But that's just not how speciation works. We know that new species do not come about in a single generation. The property of being a tiger, then, must be potentially vague or indefinite in its instantiation. Once upon a time there would have been populations of animals that were not tigers which, over many generations, slowly evolved into populations that were tigers. And during this transition, there would have been animals that were neither definitely tigers nor definitely not tigers.
One way to make sense of this is to identify the property of being a tiger with a cluster of properties. That is, properties that are statistically grouped together, so that the presence of some make the presence of others all the more likely. Such clusters might have outliers and anomalies, and that doesn't seem to be incompatible with their reality. And so if we identify the property of being a tiger with a property cluster, then this avoids the evolutionary paradox.
Some things are clustered only by our grouping them together and labeling them with the same name. Such distinctions are drawn simply because it's useful to individuals or society, and not because that's how nature is genuinely carved up. But if that were the case here, then whether or not something was a tiger would depend on whether or not we labelled it a tiger. If all of society singled out a particular tiger, and decided that it was no longer a member of that species, surely society would be in error. Whether or not an animal is a tiger doesn't depend on whether or not it has been labeled as such.
More plausibly, it seems, tigers are grouped together by nature itself. There are naturally occurring causal mechanisms—i.e. the psychological, physiological, and genetic processes of tiger reproduction—that are responsible for the clustering of tigeresque traits in tigers. Human society is not required for any of this, nor does it add anything to it other than giving it a name. And, since it is a naturally occurring property cluster, it's surely real and objective in any meaningful sense. It is a way in which nature is genuinely carved up, albeit one with potentially vague boundaries. And it is this naturally occurring property cluster, then, that we have named "tiger".
Of course, there may be naturally occurring property clusters that we have not yet named. For instance, what about the naturally occurring cluster of human values? There are certain things for which it is just in human nature to value, usually manifesting as pro-social attitudes. Almost everyone wants to be thought of as being kind and generous and brave, even if they don't generally behave in such a way. No one wants to think of themselves as a villain. People will do mental gymnastics to excuse their bad behavior, so they can preserve some semblance of a virtuous self-image. This illuminates what we naturally value. And it's not difficult to see why this would be. Evolutionary pressures have made our species into social animals, that flourish when we live together and can depend on each other. So there are clear causal mechanisms in place that are responsible for human values being clustered about pro-social patterns of behaviour. Let's call this naturally occurring property cluster the human value cluster.
The human value cluster is wholly comprised of human values, just as tigers are wholly comprised of their tigeresque traits. There is nothing that any particular tiger consists of over and above its tigeresque morphology and genome and so on. That is to say, the biological characteristics of any particular tiger are, alone, sufficient for being a tiger. Likewise, then, there is nothing that any particular instance of the human value cluster consists of, over and above the evaluative attitudes that make it up.
Nevertheless, the human value cluster is, in a way, independent of the actual values that make up its instances. Even if society were to change suddenly, and people started valuing the opposite of what they do now, so that it was socially acceptable to be violent and malicious, and taboo to be kind and generous, even then this would not make a difference to whether or not it's natural for humans to have these attitudes. And so if we identified moral properties, such a goodness, with the human value cluster, this would result in a form of subjectivism that is nevertheless robustly non-relativistic.
But one might think that this isn't good enough. What if human evolution had gone otherwise, and it was instead natural for humans to value what we currently condemn? Doesn't this form of subjectivism entail that, if we were to naturally value malice, then malice would be good? It does not. Let's raise a similar question about tigers. Suppose some animal drastically different from present day tigers had evolved from the tigers ancestor. Would that animal then be a tiger, and would tigers then be drastically different from what they are today? It seems clear that they would not. If evolutionary history had gone otherwise, tigers just wouldn't exist. Some different animal would exist instead. Similarly, then, if human evolution had gone quite differently, so that humans naturally valued the complete opposite of what they do now, then the human value cluster simply wouldn't have existed. Some other cluster of human values would have taken it's place. And that other cluster is not what we have named "the human value cluster". What follows, then, is that there wouldn't have been such a thing as goodness at all.
This "objective" sort of moral subjectivism renders moral properties robustly non-relativistic in their instantiation, but not in their reality. This should be seen as something of a win, for the latter sort of relativism is not nearly so deeply unintuitive as the former. That being said, I don't doubt that there are problems with this latter sort of relativity as well. But many views, if not all, face problems, and one might think that this is not significantly worse than competing theories of moral ontology. This may very well be a viable form of moral subjectivism.