Testing Moral Progress
Mike Huemer just published his version of the familiar argument that changing moral views is evidence for moral realism. Here is the progress datum he seeks to explain:
Mainstream illiberal views of earlier centuries are shocking and absurd to modern readers. The trend is consistent across many issues: war, murder, slavery, democracy, women’s suffrage, racial segregation, torture, execution, colonization. It is difficult to think of any issue on which attitudes have moved in the other direction. This trend has been ongoing for millennia, accelerating in the last two centuries, and even the last 50 years, and it affects virtually every country on Earth. … All the changes are consistent with a certain coherent ethical standpoint. Furthermore, the change has been proceeding in the same direction for centuries, and the changes have affected nearly all societies across the globe. This is not a random walk.
Huemer’s favored explanation:
If there are objective ethical truths to which human beings have some epistemic access, then we should expect moral beliefs across societies to converge over time, if only very slowly.
But note three other implications of this moral-learning process, at least if we assume the usual (e.g., Bayesian) rational belief framework:
The rate at which moral beliefs have been changing should track the rate at which we get relevant info, such as via life experience or careful thought. If we’ve seen a lot more change recently than thousands of years ago, we need a reason to think we’ve had a lot more thinking or experience lately.
If people are at least crudely aware of the moral beliefs of others in the world, then they should be learning from each other much more than from their personal thoughts and experience. Thus moral learning should be a worldwide phenomena; it might explain average world moral beliefs, but it can’t explain much of belief differences at a time.
Rational learning of any expected value via a stream of info should produce a random walk in those expectations, not a steady trend. But as Huemer notes, what we mostly see lately are steady trends.
For Age of Em, I read a lot about cultural value variation, and related factor analyses. One of the two main factors by which national values vary correlates strongly with average national wealth. At each point in time, richer nations have more of this factor, over time nations get more of it as they get richer, and when a nation has an unusual jump in wealth it gets an unusual jump in this factor. And this favor explains an awful lot of the value choices Huemer seeks to explain. All this even though people within a nation that have these values more are not richer on average.
The usual view in this field is that the direction of causation here is mostly from wealth to this value factor. This makes sense because this is the usual situation for variables that correlate with wealth. For example, if length of roads or number of TVs correlate with wealth, that is much more because wealth causes roads and TVs, and much less because roads and TV cause wealth. Since wealth is the main “power” factor of a society, this main factor tends to cause other small things more than they cause it.
This is as close as Hummer gets to addressing this usual view:
Perhaps there is a gene that inclines one toward illiberal beliefs if one’s society as a whole is primitive and poor, but inclines one toward liberal beliefs if one’s society is advanced and prosperous. Again, it is unclear why such a gene would be especially advantageous, as compared with a gene that causes one to be liberal in all conditions, or illiberal in all conditions. Even if such a gene would be advantageous, there has not been sufficient opportunity for it to be selected, since for almost all of the history of the species, human beings have lived in poor, primitive societies.
Well if you insist on explaining things in terms of genes, everything is “unclear”; we just don’t have good full explanations to take us all the way from genes to how values vary with cultural context. I’ve suggested that we industry folks are reverting to forager values in many ways with increasing wealth, because wealth cuts the fear that made foragers into farmers. But you don’t have to buy my story to find it plausible that humans are just built so that their values vary as their society gets rich. (This change need not at all be adaptive in today’s environment.)
Note that we already see many variables that change between rich vs. poor societies, but which don’t change the same way between rich and poor people within a society. For example rich people in a society save more, but rich societies don’t save more. Richer societies spend a larger fraction of income on medicine, but richer people spend a smaller fraction. And rich societies have much lower fertility even when rich people have about the same fertility.
Also not that “convergence” is about variance of opinion; it isn’t obvious to me that variance is lower now than it was thousands of years. What we’ve seen is change, not convergence.
Bottom line: the usual social science story that increasing wealth causes certain predictable value changes fits the value variation data a lot better than the theory that the world is slowly learning moral truth. Even if we accepted moral learning as explaining some of the variation, we’ll need wealth causes values to explain a lot of the rest of the variation. So why not let it explain all? Maybe someone can come up with variations on the moral learning theory that fit the data better. But at the moment, the choice isn’t even close.