Heading Toward: No-Nonsense Metaethics

Part of the se­quence: No-Non­sense Metaethics

A few months ago, I pre­dicted that we could solve metaethics in 15 years. To most peo­ple, that was out­ra­geously op­ti­mistic. But I’ve up­dated since then. I think much of metaethics can be solved now (de­pend­ing on where you draw the bound­ary around the term ‘metaethics’.) My up­com­ing se­quence ‘No-Non­sense Me­taethics’ will solve the part that can be solved, and make head­way on the parts of metaethics that aren’t yet solved. Solv­ing the eas­ier prob­lems of metaethics will give us a clear and sta­ble plat­form from which to solve the hard ques­tions of moral­ity.

Me­taethics has been my tar­get for a while now, but first I had to ex­plain the neu­ro­science of plea­sure and de­sire, and how to use in­tu­itions for philos­o­phy.

Luck­ily, Eliezer laid most of the ground­work when he ex­plained could­ness, ter­mi­nal and in­stru­men­tal val­ues, the com­plex­ity of hu­man de­sire and hap­piness, how to dis­solve philo­soph­i­cal prob­lems, how to taboo words and re­place them with their sub­stance, how to avoid defi­ni­tional dis­putes, how to carve re­al­ity at its joints with our words, how an al­gorithm feels from the in­side, the mind pro­jec­tion fal­lacy, how prob­a­bil­ity is in the mind, re­duc­tion­ism, de­ter­minism, free will, evolu­tion­ary psy­chol­ogy, how to grasp slip­pery things, and what you would do with­out moral­ity.

Of course, Eliezer wrote his own metaethics se­quence. Eliezer and I seem to have similar views on moral­ity, but I’ll be ap­proach­ing the sub­ject from a differ­ent an­gle, I’ll be phras­ing my solu­tion differ­ently, and I’ll be cov­er­ing a differ­ent spread of top­ics.

Why do I think much of metaethics can be solved now? We have enor­mous re­sources not available just a few years ago. The neu­ro­science of plea­sure and de­sire didn’t ex­ist two decades ago. (Well, we thought dopamine was ‘the plea­sure chem­i­cal’, but we were wrong.) De­tailed mod­els of re­duc­tion­is­tic meta-ethics weren’t de­vel­oped un­til the 1980s and 90s (by Peter Rail­ton and Frank Jack­son). Re­duc­tion­ism has been around for a while, but there are few philoso­phers who re­lentlessly play Ra­tion­al­ist’s Ta­boo. Eliezer didn’t write How an Al­gorithm Feels from the In­side un­til 2008.

Our meth­ods will be fa­mil­iar ones, already used to dis­solve prob­lems rang­ing from free will to dis­ease. We will play Ta­boo with our terms, re­duc­ing philo­soph­i­cal ques­tions into sci­en­tific ones. Then we will ex­am­ine the cog­ni­tive al­gorithms that make it feel like open ques­tions re­main.

Along the way, we will solve or dis­solve the tra­di­tional prob­lems of metaethics: moral episte­mol­ogy, the role of moral in­tu­ition, the is-ought gap, mat­ters of moral psy­chol­ogy, the open ques­tion ar­gu­ment, moral re­al­ism vs. moral anti-re­al­ism, moral cog­ni­tivism vs. non-cog­ni­tivism, and more.

You might re­spond, “Sure, Luke, we can do the re­duce-to-al­gorithm thing with free will or dis­ease, but moral­ity is differ­ent. Mo­ral­ity is fun­da­men­tally nor­ma­tive. You can’t just dis­solve moral ques­tions with Ta­boo-play­ing and re­duc­tion­ism and cog­ni­tive sci­ence.”

Well, we’re go­ing to ex­am­ine the cog­ni­tive al­gorithms that gen­er­ate that in­tu­ition, too.

And at the end, we will see what this all means for the prob­lem of Friendly AI.

I must note that I didn’t ex­actly in­vent the po­si­tion I’ll be defend­ing. After shar­ing my views on metaethics with many sci­en­tifi­cally-minded peo­ple in pri­vate con­ver­sa­tion, many have said some­thing like “Yeah, that’s ba­si­cally what I think about metaethics, I’ve just never thought it through in so much de­tail and cited so much of the rele­vant sci­ence [e.g. re­cent work in neu­roe­co­nomics and the sci­ence of in­tu­ition].”

But for con­ve­nience I do need to in­vent a name for my the­ory of metaethics. I call it plu­ral­is­tic moral re­duc­tion­ism.

Next post: What is Me­taethics?