Competitive Ethics

by Milan Cvitkovic3 min read24th Nov 20201 comment

10

World OptimizationRationality
Frontpage

(Crossposted from EA Forum)


If antinatalists are right that having children is wrong, does it matter once the antinatalists go extinct?

If you build an "ethical" AI that keeps getting deleted by its "unethical" AI peers, have you accomplished your mission of building ethical AI?

Is religious tolerance a fatal flaw in liberal democracy if fecund, illiberal religions can always become a majority?

If we're going to think hard about what's right, shouldn't we also think hard about what wins?


Competitive ethics (I'd be happy to find a better term) is the study of ethics as strategies or phenotypes competing for mindshare rather than as statements about right and wrong.

Competitive ethics is to morality as FiveThirtyEight is to politics. FiveThirtyEight doesn't tell us which candidate's positions are correct, and we don't expect them to. We expect them to tell us who will win.

Unlike applied ethics ("How should I act in this specific situation?"), normative ethics ("What criteria should I use to do applied ethics?"), or meta-ethics ("How should I think about normative ethics?"), competitive ethics is amoral. Not immoral, amoral: it's not concerned with right and wrong, just with predictions and understanding.

No matter your normative ethical beliefs or your meta-ethics, competitive ethics matters to you in practical terms. Moral statements may or may not be true or meaningful, but people definitely act according to them.

How ethics compete

There are many lines of thinking relevant to this question, but I can't find any that address it directly.

The most relevant are cultural selection theory, memetics, and neoevolution, though these are far too tied up with evolutionary theory. ("Ethics" as I'm using the term encompasses things like religion, culture, norms, and values --- anything that guides people in how they say "yuck" or "yum.") The subfields of evolutionary ethics and game-theoretic ethics stick to normative or occasionally meta-ethical questions, and don't seem to have studied what happens when ethical systems go toe-to-toe.

An important distinction in thinking about how ethics compete is between the ethics people publicly espouse, the ethics they consciously believe, and the "revealed ethics" of what they actually do. All three are related, and all three can be distinct. Preference falsification, social contagion theory, and behavioral economics are the relevent disciplines here. Professed ethics are the fastest to change, a la preference falsification. It's an open question whether believed or revealed ethics are more mutable.

Another important issue is the fuzzy line between biologically-determined preferences and ethics. The former clearly influence the latter in a single individual, and the latter influences the former across generations. Plus, the more technology lets us intervene on biology, the fuzzier the line gets. Wibren Van Der Berg's Dynamic Ethics is the closest work to addressing this, though it's a work of normative ethics. E.g. when he says "Our dynamic society requires a dynamic morality and thus a form of ethical reflection which can be responsive to change." A few others have touched this question, but not many.

Case studies

Natalism and heritability

The most straightforward way ethical systems compete is by the degree of natalism and heritability they entail: how many offspring do they lead to in their believers, and how effectively are they passed from parents to children?

The best recent work on this topic is from demographers like Eric Kaufmann. In his book Shall the Religious Inherit the Earth?, Kaufmann lays out the remarkable growth trends of religious fundamentalist groups in the modern world. Fundamentalist religious groups with ethics encouraging high fertility and strict adherence to the religion are contrasted with modern Western cultures with ethics that deride fertility (e.g. certain environmentalist ethics) and encourage freedom of thought.

Most fundamentalist groups rely on the generosity of the society at large to flourish as they do (e.g. the ultraorthodox in Israel, who generally don't have jobs), so it's not clear when this will hit the breaking point. Nevertheless, these trends raise questions about the viability of non-natalist ethics. From my probably-biased perspective, I suspect ethics of free thought are more attractive than fundamentalist ethics. I hear more about people leaving fundamentalist religions than joining them. But ethics of free thought combined with low fertility may not be sustainable.

Nihilism and motivation

I know of no work studying the comparative effects of ethical belief systems on motivation. In fact, I don't know whether it's demonstrable that motivated individuals are more successful. But assuming it does, and assuming ethics like moral nihilism demotivate people (or at least fail to motivate them), the long-term viability of these ethical systems is questionable. Going further, it may be that selfish ethical systems (e.g. Ayn Rand, Gordon Gekko) are more associated with motivation and success than egalitarian ethical systems.

Causality and correlation are hard to tease apart here, but doing so isn't necessary. An ethical system can win both by granting success to its holders or by being adopted by successful individuals.

Ethnonationalism vs diversity

Crudely put: it feels better to be in an exclusive group, but inclusive groups are bigger. Which matters more?

AI alignment

Eliezer Yudkowsky is purported to have said "You are personally responsible for becoming more ethical than the society you grew up in." This quotation is interesting in that (1) it's a normative claim about normative claims, and (2) it assumes that ethics has a direction.

While I like the sentiment, it's reminiscent of when lay people say things like "humans are more evolved than snails" and make biologists cringe because evolution doesn't have partial ordering by which some species can be more or less evolved than others. From the competitive ethics perspective, neither do ethics.

Most people who work in AI alignment treat human values the way engineers treat nature: there is an underlying true human ethics, and while we can't articulate it, we can still try to hue to it. But if you build an “ethical” AI that keeps getting deleted by its “unethical” AI peers, have you accomplished your mission of building ethical AI?

I'm not able to join the AI alignment discussion until AI alignment researchers start putting competitive ethical questions more front and center.

Extensions of competitive ethics

Competitive ethics on its own is amoral. But it can be a building block for other ideas.

Consider a meta-ethics --- call it ethical consistentism maybe --- where the probability of a moral statement being correct is proportional to its survival. To be clear: this isn't a creepy social Darwinism or might-makes-right idea since it's a meta-ethics, not a normative claim. Or one could propose a a weaker version of this: an ethical system shouldn't directly or indirectly lead to itself not being believed. This is analagous to logical consistency in mathematics. Of course, if we're going to treat ethical systems as competitive phenotypes, it seems only fair to treat meta-ethical systems (ethical consistentism included) as phenotypes too. So the recursion begins...

Competitive ethics is also sortof nihilism 2.0. Of course right and wrong are ridiculous concepts, so what? That’s the start of the conversation, not the end.


Despite searching quite a bit, I can't find any content on LW related to these ideas. But I'm sure there is some. Please let me know if you know any!

10

1 comments, sorted by Highlighting new comments since Today at 3:26 AM
New Comment

Competitive Ethics, or (the study of) the Competition of Ethics?


If antinatalists are right that having children is wrong, does it matter once the antinatalists go extinct?

Does it matter if you win, if you sacrifice your highest values along the way?


If we're going to think hard about what's right, shouldn't we also think hard about what wins?

What does it even mean to win?


From https://en.wikipedia.org/wiki/Pyrrhic_victory

If we are victorious in one more battle with the Romans, we shall be utterly ruined.

— Plutarch[3]

[3] Plutarch. "The Life of Pyrrhus". Parallel Lives. IX (1920 ed.). Loeb Classical Library. p. 21.8. Retrieved January 26, 2017.


Competitive ethics (I'd be happy to find a better term) is the study of ethics as strategies or phenotypes competing for mindshare rather than as statements about right and wrong.

The word 'mindshare' there completely changes the piece, from what it seemed to be leading up to, up to this point.


Competitive ethics is to morality as FiveThirtyEight is to politics. FiveThirtyEight doesn't tell us which candidate's positions are correct, and we don't expect them to. We expect them to tell us who will win.

And this is another, completely different thing, that doesn't have anything to do with ethics.


There are many lines of thinking relevant to this question, but I can't find any that address it directly.

Win and you get to pretend you were good. Everyone else will erase you, change you, or paint you as a terrible thing completely unrelated to what you were - an ugly scarecrow bearing your name but their visage.


neoevolution

neo or neuro?


The most straightforward way ethical systems compete is by the degree of natalism and heritability they entail: how many offspring do they lead to in their believers, and how effectively are they passed from parents to children?

It's important to not neglect fitness here. Naively maximizing the number of offspring doesn't just result in deaths, but also in malnourishment. Maximum population, all weak and starved, does not an army make.


deride fertility (e.g. certain environmentalist ethics) and encourage freedom of thought.

Encouraging freedom of thought conflicts with other values. You are free to think - except that thought. It's not an absolute, it just doesn't work with absolutes. Unless it's a lie.


I hear more about people leaving fundamentalist religions than joining them. But ethics of free thought combined with low fertility may not be sustainable.

Unless free thought outlaws systems that don't support/work against it?


Going further, it may be that selfish ethical systems (e.g. Ayn Rand, Gordon Gekko)

Selfish people don't have time to read Ayn Rand, they're busy doing what they want. What about hedonism?


Causality and correlation are hard to tease apart here, but doing so isn't necessary. An ethical system can win both by granting success to its holders or by being adopted by successful individuals.

And that's where you've lost me. 'These systems of belief lead to success among their holders' is interesting. 'Successful people happen to believe this because they're successful not the other way around', not so much.


Eliezer Yudkowsky is purported to have said "You are personally responsible for becoming more ethical than the society you grew up in." This quotation is interesting in that (1) it's a normative claim about normative claims, and (2) it assumes that ethics has a direction.

Or it assumes ethics has a magnitude.


But if you build an “ethical” AI that keeps getting deleted by its “unethical” AI peers, have you accomplished your mission of building ethical AI?
I'm not able to join the AI alignment discussion until AI alignment researchers start putting competitive ethical questions more front and center.

AI peers? It seems more likely that an 'ethical AI' will be less powerful/move slower, and not 'keep getting deleted' but get deleted once. And then it's game over.


Consider a meta-ethics --- call it ethical consistentism maybe --- where the probability of a moral statement being correct is proportional to its survival.

That's not a meta-ethic, that's a strategy. And 'ethical' doesn't belong in this statement.


The only relation to 'ethical' your stuff has is that it asks 'will this survive?' Arguably, ability to survive impacts ability to bring about the 'ethical' ends/whatever valued. It also might impact how much* the system itself survives - if the 'more pragmatic ethics' eventually drops the 'ethics', what left isn't an ethic.

*/how long/how likely