Note this essay is based on Gilbert Harman’s article “Ethics and Observation”.
In ethics we are often tempted to support or reject a theory based on how well it agrees with our intuitions. If we wanted to be more precise we might say that our observations of our intuitions were the basis of ethical theories, much like observations of nature are the basis for theories in physics. From this similarity we might conclude that there exist moral facts which are approximated by ethical theories in the same way that physical facts exist and are approximated by physical theories. Gilbert Harman argues that this method of ethical philosophy did not justify the existence of objective ethical facts in the same way that observation in physics justified physical facts, and thus that ethical facts do not exist while physical ones do. To support this claim he described a principle called the explanatory requirement, the purpose of which is to differentiate theories with a basis in facts from those without.
The explanatory requirement states that an observation only justifies a theory about facts when the theory itself explains why we made the observation. Thus our observations that objects fall toward the ground justify the existence gravity because the theory of gravity in question predicts that they would in fact fall. On the other hand, we cannot likewise claim that an observation of our intuition that some action is right can support an ethical theory, because the best explanation as to why we have our intuitions is purely psychological, and has nothing to do with the ethical theory in question. Our ethical observations may seem to guide us to a given theory, but because such a theory does not predict that we will have ethical intuitions it does not meet the explanatory requirement.
Although physics may meet the explanatory requirement it may not seem as clear that the other sciences do. Consider psychology, for example. If we have a psychological theory concerning the behavior of people, we may take their observed behavior as evidence for or against that theory. However, it might not seem like the theory entails our observations, because behavior is caused by the electrical activity of neurons, and our psychological theory does not predict such activity. Thus the theory is not needed to account for our observation, only physical theories regarding the activity would seem to make the required prediction. The best response to this observation is not to drop psychology but to re-define what a theory in psychology is about. To defend psychology from the explanatory requirement, we must argue then that theories about behavior are really theories about neural activity that result in such behavior. Thus modified, our new theory predicts certain neural activity, the subject of our observations, and thus meets the explanatory requirement. Ethics, unfortunately, does not have this defense open to it, because to take such a view with respect to ethics would require us to equate ethical statements with physical statements, and such an equivalence seems impossible.
Physics, and the other natural sciences, may seem to meet the explanatory requirement, but we now may have doubts about the existence of mathematical facts under this criterion. Like ethics, mathematics is not backed by the direct observations that support the natural sciences, so we might feel that mathematical facts may be judged to be as unfounded as ethical facts. Harman argues that mathematical facts are backed indirectly by the natural sciences, but there are some branches of mathematics which are not used in the natural sciences, such as number theory and advanced higher dimensional topological theory, and the validity of these branches might still be in question. Consider a theorem, say Fermat’s last theorem. This theorem is not justified by observations of numbers in nature, nor by the intuitions of mathematicians, but by a proof. It is true that before it was proved, mathematicians did have intuitions concerning whether it was valid or not, but these intuitions were not considered to justify the theorem. What backs a theorem then is its proof, which is itself justified by axioms.
So far so good, but we still have not justified the axioms themselves. Let us assume we are working with mathematics in its purest form, where the axioms are simply definitions that set up the structure within which theorems are proved. For example the parallel postulate can be rationally denied, and is in non-Euclidean geometry, but despite this we still consider Euclidean geometry a valid structure to prove theorems in, even though no real space is Euclidean. What this investigation reveals then is that theories in mathematics are not like theories in physics. A theory in mathematics admits its dependence on certain assumptions. In contrast, the “axioms” of physics might be observations made that confirm or deny theories. Just as we might reject a mathematical theory if one of its premises were shown to be invalid, so we might reject a physical theory if empirical observations conflicted with it. However, an important difference is as follows: the goal of physics is to figure out the laws of nature, i.e. physical facts, even though physical theories are often wrong, there is some ultimate truth that we attempt to approximate with them. Mathematics, on the other hand, is not trying to uncover some ultimate truth about numbers, instead all its truths are conditional within whatever system one is working with, defined by the axioms. We freely admit that mathematical facts are conditional on the system we are working in, and thus are not like the absolutely true facts that physics works towards. The question becomes: can we avoid the explanatory requirement by constructing our ethics to be more like mathematics than like physics, yielding ethical facts that are like the facts of mathematics?
We might be tempted to construct an ethical system based, like mathematics, on just a few fundamental axioms, and perhaps admit that our ethical systems only applied to those who accepted or lived by our axioms. In fact, one might be tempted to view constructivist systems in just this way. For example, utilitarianism could all be seen as stemming from the axiom that the maximization of total happiness is good. However, the ethical facts we were looking for were supposed to be universal, and independent of what people believe, or at least these are the kind of ethical facts that Harman is looking for. An axiom based system however would only be able to draw valid conclusions for people who accepted its axioms, but since people may not all accept the same axioms, the conclusions would not in fact be universal. We could argue that such a theory about human behavior is not ethics at all, but really some branch of psychology, simply telling us how we could expect various individuals to behave, but not how they should behave.
Up to this point I have been assuming that ethics does not meet the explanatory requirement, but there are in fact at least two kinds of ethical theories that might indeed meet Gilbert’s requirement. One type of theory would be to equate ethical judgments with some aspect of physical reality, much like the defense of psychology against the explanatory requirement. Another type of theory might be a mental-ethical theory, which at the same time explains ethics and the workings of some aspects of our mind. Our ethical intuitions, the subject of our observations, can then be predicted the theory, and thus it would pass the explanatory requirement. Here I will simply dismiss ethical theories that equate ethical statements to physical properties. This is not to say that I have an air-tight rebuttal of such theories, but because I have yet to see a successful theory of this kind, and because of Moore’s objections, I will simply assume it cannot be done. On the other hand, there seem to be several ethical theories that would explain our intuitions and ethics at the same time. For example, one might argue that good is defined by a social consensus, and that our intuitions result from our awareness of this consensus. Or we could argue that good is merely a persuasive tool, and that our intuitions reflect our preferences for or against certain activities. These theories do pass the explanatory requirement because if we asked them “why do I have a particular ethical intuition,” they can provide an answer.
To get rid of such theories we might propose the following modification of the explanatory requirement: that the only explanation for the observation is the theory in question. Then the ethical theories mentioned above would indeed fail, because our intuitions can always be explained psychologically. Unfortunately modern physics would also fail this version of the explanatory requirement, for while it is true that one explanation for why things fall is gravitational theory, there exists another explanation, namely that earth attracts earth, and thus gravity could not be said to pass this version of the explanatory requirement.
So even though the explanatory requirement does not throw the baby out with the bathwater, so to speak, and we are able to retain mathematics under it, it now seems to have failed to demonstrate the non-existence of ethical facts since we have been able to construct ethical theories that meet its demands. The explanatory requirement however seems to be based on a verificationist view of science, but currently it is not the ability to make confirming observations but the ability to find falsifying observations that is considered to make a theory scientific, and perhaps this is why the explanatory requirement fails. We might attempt to improve it by transforming it into its equivalent under a falsificationist mode of thought, yielding the principle that a theory is only justified by the lack of falsifying observations, where a falsifying observation is defined as any observation that would require the denial of the theory. Like the explanatory requirement, this allows the natural sciences such as physics and psychology to pass the test. Similarly we can also defend mathematics by appealing to its status as true as conditional on various axioms. The mental-ethical theories do not meet our new requirement, because an intuition that is at odds with our ethical conclusion does not falsify the theory; instead we could conclude that our intuitions had become distorted. For example, if we considered the good to be defined by social norms, then perhaps our contrary intuition reflects our poor understanding of these norms. Therefore because these ethical theories cannot be falsified by observations, they are not theories in the same way as theories in the natural sciences and thus ethical facts do not exist in the sense that physical facts do.