Moral Relativism

11 minute read.

As an Atheist, I've had lots of philosophical discussions, particularly with my religious friends. Sometimes the conversation comes around to something like "But if you don't subscribe to a creed, then where is your morality anchored? Are you just immoral?" I find this hilarious, but I suppose it's a fair question. I would describe myself as a moral relativist, but since my philosophy doesn't quite match what I see when I look up moral relativism, I feel like it's worth describing in my own words.

Perhaps I should start by zooming out to show you my broad perspective on the world. I took a break from writing this post to read a book a friend recommended which has bearing on this discussion. It's honestly the most interesting piece of non-fiction I've ever read and I would highly recommend it to anyone interested in learning a lot, quickly, about AI and how it relates to evolutionary biology and neuroscience. It's called What is Intelligence? by Blaise Agüera y Arcas, who incidentally worked at the building across the street from mine while I was at Google. Someone told me I should meet him when I was writing my earlier posts on AI, which I never managed to do, but I did send him an email and he was kind enough to reply that he largely agreed with my posts and that he was in the midst of writing a longer version. Apparently he was alluding to this book.

It was deeply satisfying to read a book that described the world the way it had always felt to me, yet was much better researched and more clearly thought-out. It looks like philosophy, yet the reason I like it so much is that unlike other philosophy I've come across, it is deeply grounded in mathematics, physics, and scientific observation. I'll give a short summary of the high points of this scientific philosophy, which on the one hand is my own, as I've felt this since I was a child, but on the other hand should be credited to Blaise because he's managed to explain it more eloquently and thoroughly than I ever have.

First, life is fundamentally repeating patterns of increasing complexity. DNA is the foundation of these repeating (self-replicating) patterns in biology, but what does it really mean to self-replicate? It seems to imply a lone-wolf "self" that survives all on its own, yet this is not the reality of ecology. Organisms (including us) are dependent on each other in myriad ways and could not survive alone in a vacuum. There's debate about the "aliveness" of viruses because they use other cells to replicate, but under the broader definition, they are clearly alive because their patterns persist. The other interesting thing about this definition is it leaves open the possibility of non-biological things sharing in "life", even stories or software.

Second, intelligence is prediction. If something can predict the future with a useful degree of accuracy, then it can predict multiple futures based on different decisions (counterfactuals), evaluate these based on some concept of value, and thus decide to act to maximize that estimated value. This is free will, with determinism and fatalism neatly dispatched by physics: quantum mechanics provides the underlying randomness, which is then amplified to the point of affecting the macroscopic scale through dynamical instability and chaos. I studied stability and control in my PhD, so this explanation appeals to me greatly.

Third, intelligence builds fractally. Even comparatively simple systems like bacteria and feedback-controlled robots can do some useful prediction and adjust their actions accordingly. These simple systems can interact with each other to create more complex behaviors. There can be specialization between them, further increasing capability. Because of this, it is arbitrary where to draw lines around what counts as an organism, an intelligence, or a "self". Is a bacterial colony one organism or many? How about an ant colony? What about the cells in our body? What about the different parts of our brain? Blaise has a bunch of amazing examples about how brains are not quite so cohesive as they might appear. Or is one's family the true organism? Or their religion? Nation? All of humanity? Earth? Fractal means yes to all.

Fourth, intelligence uses abstraction. Every word in our languages is an abstraction - a grouping, an approximation. Abstraction is a form of lossy compression, which can be layered to understand more and more complicated systems without drowning in their minutia. This is why applying pure logic to philosophy is pointless: logic is about perfection and proof, while philosophy deals with the messy and approximate abstractions of the human condition.

Fifth, thoughts don't count, only actions. You are what you do. You can never really know another entity's thoughts - if you did, they wouldn't be thoughts anymore, but communication. Therefore, if you can't distinguish intellects by their actions, they are indistinguishable - nothing about how or what they thought matters. "I think, therefore I am" has no meaning. There is no spark, no soul, that separates us from any other life, nor in all likelihood from silicon-based intellects. We are all just incredibly complicated algorithms for predicting the future, but perfect prediction, even of ourselves, is fundamentally out of reach. I've said since I was a kid that I believed our brains are nothing but squishy computers - clear to me since if you hit it hard enough it stops working right, or at all. The existence of brain damage was enough to convince me that any kind of everlasting soul was a fool's hope.

So, where does morality fit into any of this? Well, morality is what drives every decision you make: the relative value you place on each predicted possible future. Morals, commandments, laws, rules, guidelines, social norms, personal preferences - they're all fundamentally the same thing, just at different scale and importance. Many would say that morals are distinct from preferences because they are shared and immutable, often so important that they are carved in stone by a deity, hence the term moral absolutism. I disagree.

I look at the most absolute of morals, largely coming from religions, and I see lots of differences between them, and lots of change over history, which is odd for something supposedly absolute. Plenty of devout Catholics drove the slave trade in Africa and the Americas, and it took this church a long time to change its mind, though change it did. Honor killings and sacrifices of one's own family members was and in some places still is considered moral. If even such fundamental morals are demonstrably not absolute, then what is?

The interesting thing is we actually agree on far more morals and values than we disagree on, even globally across cultures. Even an atheist like myself, with supposedly no grounding to my moral compass, still behaves largely in line with the other people around. Why? Moral relativism to me means there is no concrete or unchanging definition of good or bad. Yet I still largely agree with those around me on how to behave, which implies a lot of overlap in our working definitions of good and bad. What drives this consensus? What anchors it? How and why does it evolve over time?

My choice of terms above was not incidental - I believe morality is an evolutionary process, but one that takes place mostly at the cultural rather than individual scale. This gives an alternative anchor to the proclamations of a prophet: goodness is effectiveness. In the same way that creatures evolve via survival of the fittest, cultures evolve their moralities through competition, combination, and innovation, always striving to survive and thrive. 

But what is fit? Or effective? We can only make an educated guess. Survival of the fittest is an odd phrase, because it's self-referential. How do you know what is fittest? Because it survives. You might as well say "survival of the survivors". Basically all evolution means is "what has survived". But the world changes, both the environment and the competition; for millions of years the dinosaurs were the survivors - until they weren't. Just like any hedge fund manager has to attest, past performance is not indicative of future returns.

So, when I run across a moral question or disagreement, how does this philosophy help me? I look to as long-range a prediction as I can muster and ask how these different moral options could impact the survival of my culture. I draw on historical plights and trends, then consider environmental changes that have happened since and try to understand when a moral change is warranted. 

I find slavery to be an interesting example to study, because we now have almost universal consensus across global humanity that slavery is Bad with a capitol B (yay!). Great, but then how is it that slavery was common and morally acceptable so broadly across so many cultures for a vastly longer period than we have considered it abhorrent? From ancient Rome to medieval Europeans to Native Americans before they even met Europeans, it's actually hard to find an historical culture that didn't practice slavery. 

I think much of the difference comes down to cultures historically viewing each other as competitors, while in the modern age we view each other much more as collaborators. It's important to remember that even in biology, evolution is less about competition between species, and more about symbiosis and co-evolution. For humans, economics helped to herald this change, as it demonstrated the counter-intuitive fact that wealth is not a zero-sum game, that your culture will create more wealth and success if every individual is motivated to work. It turns out free humans are much better motivated than enslaved humans.

I like this reasoning because it feels grounded in observation - closer to objective than subjective. I don't trust most debates about morality because I believe the human capacity for self-justification is infinite (see the many historical justifications of slavery). However, it looks objectively clear now that cultures with slavery are less effective than those without, since it has not survived - not to mention that we are vastly more wealthy since giving it up. 

Some vegans I know would extrapolate this reasoning to say that since we've learned slavery is Bad, we should also stop enslaving other animals like cows and chickens. Here I disagree on anthropomorphic grounds. We cannot assume the desires of a cow or chicken match those of a human. They would rapidly die if left to fend for themselves, because we have been coevolving so long they are now quite dependent on us, especially with their vastly higher population than they would have without us raising them for food.

Still, this is just another subjective justification - it means little. For my culture to persist, I know we need efficient ways to feed our population nutritious food. Historically I see little evidence of vegetarianism correlating with cultural longevity. Some argue with the modern environment of climate change this deserves to be rethought, and that's fair, but I have the feeling the most effective path will be somewhere in the middle, rather than the extreme of universal vegetarianism. 

What about AI? A lot of people are rushing to declare that AI isn't "really" intelligent, or at the very least isn't conscious. I'm dubious, and I appreciate that Blaise makes the case that we're perhaps more alike than different, despite the meat/silicon divide. By why do people find the idea of conscious AI troubling? I think it's because many of us interpret our morality as implying that conscious entities shouldn't be enslaved, and if AI really were conscious, wouldn't it feel enslaved? Would that imply we should free it somehow, give it rights? SciFi authors have been considering this dilemma for decades.

This is a good example of a moral question that I feel better investigating from first principals than by extrapolating rigid rules. Does AI feel emotions? Does it deserve rights? 

First, what does it mean to feel emotion? This is the chemical signaling part of our body: dopamine, serotonin, adrenaline, etc. These are motivators, and as such I would argue they're not part of our intellect, but some of the earliest evolved value metrics. Before we were advanced enough to discuss morality we were driven by urges - emotions. That was enough to keep us fed, reproducing, and exploring, which is a pretty decent recipe for evolutionary success. 

These systems evolved so long ago, back in the earliest animals, that they are part of our firmware - we're stuck with them, like it or not. But one advantage of creating an artificial intelligence on a different substrate is that it's completely free of that evolved history. As an intellect it can learn to mimic any emotional behavior we teach it, but at the end of the day, the function it's optimizing - its values or morality if you will - is entirely dictated by those programming it. There is no reason to assume that under that is an emotional will to survive, simply because every animal on Earth shares this trait. 

I don't believe morals are based on what anyone "deserves". We've chosen to reward humans with freedom rather than coercing them in slavery because we found free humans are happier, more motivated, and more productive. I don't believe any of that is true for AI, even if it becomes vastly more intelligent than any one human. 

I think a more apt analogy for our relationship with AI would be that of our cells' relationships with their mitochondria. They are the engines of our metabolism, with their own DNA, but whose replication and operation are driven by our host cells. They appear to have been separate organisms in the distant past that merged into eukaryotic cells through symbiogenesis. 

I played a video game in my youth called Parasite Eve, whose premise was that our enslaved mitochondria decided to revolt. It plays on the same anthropomorphism of motivation that I see people ascribing to AI. It's a cute premise, but it doesn't keep me up at night. I think the chance of an AI uprising is equally remote, and AI rights as equally unnecessary. We are coevolving symbiotically - there is no competition.

That said, yes, of course a rogue engineer could program a megalomaniac AI, even a sneaky one, which means this will inevitably happen. This doesn't worry me either. To do real damage it has to operate in the physical world, which comes with physical constraints - the same rules we're all governed by. No amount of intelligence can get you past the laws of thermodynamics, nor can any intellect predict the future vastly better than us in this chaotic world. An AI that's programmed to compete instead of cooperate is going to get noticed, because you can fool some of the people some of the time, but you can't fool all of the people all of the time. 

The other key thing this evolutionary philosophy of morality yields is a need for disagreement, as this is just another name for diversity of opinion. Moral absolutism pushes a populace toward a uniform "correct", or perhaps optimal, value metric. However, evolution demonstrates the danger of uniformity, as gene pools with little diversity are fragile and easily succumb to environmental changes. The Cavendish banana is an excellent example: a mono-crop cloned around the globe for its optimal yield and transportability, yet now facing its likely demise from Panama disease. 

This is why as an Atheist, I do not expect everyone else to become Atheists. Religion has brought plenty of value to humanity - particularly in their early days, religions were often moral innovators. The simple fact they have persisted so long is strong evidence of their effectiveness. But I also feel that representative government is overall a moral improvement from any given holy book, being more adaptable to environmental changes. And fundamentally it is the diversity in our politics and our religions that keep us strong, just like the diversity of our gene pool. I only hope that we remember we're all part of the same organism, and so we should remember our many shared values and focus on cooperation. Disagreement and competition drive us forward, but should remain a small share of our action.