3 ms·

Values are just facts about an agent's preferences, which are often revealed through the agent's actions. If it's true that "Chocolate ice cream is my favorite

Values are just facts about an agent's preferences, which are often revealed through the agent's actions. If it's true that "Chocolate ice cream is my favorite dessert", then that will be revealed in my preferences, when I always choose it over other desserts. As far as we know, the encoding for that preference exists somewhere in the brain, and we could predict a human's preferences, with sufficient brain-scanning technology. I could be wrong about chocolate ice cream being my favorite dessert, maybe my brain is wired so that chocolate cake is the best dessert, but I've never had chocolate cake, and so I am left in ignorance of this fact. Humans can be wrong about their own preferences. Asking "is chocolate cake the best dessert?", is a bit like saying "how far away is Mars". You could assume that the correctness of the preference lies within a particular human's neurology--yours or mine. You could also assume that the distance to Mars is relative to a particular point--Earth for example. But in both cases that would just be presumptuous. Both questions are under-specified; it's just a quirk of language that we can express them.

GMoromisato 3y ago
I would say it slightly differently: What values a given individual has, is purely a factual statement, as you said. If I tell you that chocolate ice cream is my favorite dessert, that's a testable fact. But what's not testable is whether I ought to like chocolate ice cream. Suppose I told you I liked chocolate ice cream and you scanned my brain and it looks like I like chocolate ice cream, and you said, "Well, I see you like chocolate ice cream, but you really shouldn't because it tastes awful." That wouldn't make sense, right? This is the is/ought divide. [As always, smarter people hundreds of years ago, have come up with better arguments than mine.] In the same way, whether or not today's world is warmer than the 1970s is a testable fact. But whether we ought to join the Kyoto Protocol or the Paris Accords is not. That's what I mean about the issue being about value preferences.
- alphazard 3y ago
  The is/ought distinction is merely linguistic. It's true: you can't derive an ought statement from an is statement. Consider your statement "I ought to like chocolate ice cream". That is an ambiguous claim, which I can try to interpret in one of two possible ways: 1. "I value that I like chocolate ice cream". If you were a rational agent, this would be an infinite regression, of course you like that your utility function is your utility function, and that liking goes on forever. It's interesting that humans, which are not rational, can also have thoughts about not liking that they enjoy certain things. 2. "It is good that I like chocolate ice cream". Good for who/what? there needs to be something that values that you like chocolate ice cream. "My friend values that I like ice cream" is not ambiguous, and firmly in factual territory again. Your friend is the valuer/agent with preferences. People hate the ambiguity and want to say the valuer is the universe or God or something else. But it's just an under-specified statement about whether one agent likes another agents preferences. Your statement about "we ought to ___", cannot be well defined in all cases, because something could be good for one agent in the group, but bad for another. Any attempt to make a claim "good for the group" is no longer a statement about individual agents and their preferences. It's about something else, if it's coherent at all. Rarely does "good for the group" mean "good for all individuals in the group respectively", which is well defined.

GMoromisato 3y ago

I would say it slightly differently: What values a given individual has, is purely a factual statement, as you said. If I tell you that chocolate ice cream is my favorite dessert, that's a testable fact. But what's not testable is whether I ought to like chocolate ice cream. Suppose I told you I liked chocolate ice cream and you scanned my brain and it looks like I like chocolate ice cream, and you said, "Well, I see you like chocolate ice cream, but you really shouldn't because it tastes awful." That wouldn't make sense, right? This is the is/ought divide. [As always, smarter people hundreds of years ago, have come up with better arguments than mine.] In the same way, whether or not today's world is warmer than the 1970s is a testable fact. But whether we ought to join the Kyoto Protocol or the Paris Accords is not. That's what I mean about the issue being about value preferences.

alphazard 3y ago
The is/ought distinction is merely linguistic. It's true: you can't derive an ought statement from an is statement. Consider your statement "I ought to like chocolate ice cream". That is an ambiguous claim, which I can try to interpret in one of two possible ways: 1. "I value that I like chocolate ice cream". If you were a rational agent, this would be an infinite regression, of course you like that your utility function is your utility function, and that liking goes on forever. It's interesting that humans, which are not rational, can also have thoughts about not liking that they enjoy certain things. 2. "It is good that I like chocolate ice cream". Good for who/what? there needs to be something that values that you like chocolate ice cream. "My friend values that I like ice cream" is not ambiguous, and firmly in factual territory again. Your friend is the valuer/agent with preferences. People hate the ambiguity and want to say the valuer is the universe or God or something else. But it's just an under-specified statement about whether one agent likes another agents preferences. Your statement about "we ought to ___", cannot be well defined in all cases, because something could be good for one agent in the group, but bad for another. Any attempt to make a claim "good for the group" is no longer a statement about individual agents and their preferences. It's about something else, if it's coherent at all. Rarely does "good for the group" mean "good for all individuals in the group respectively", which is well defined.