• Posts
  • RSS
  • ◂◂RSS
  • Contact

  • Sure and Unsure

    February 13th, 2013
    current_beliefs, morality
    I have varying amounts of confidence in different aspects of my belief system. From most sure to least sure, this how I think about moral questions. Rough confidence is included, as a percentage.

    Evaluate actions (99%++)

    The role of a moral system is to help you make choices. It is not enough to know that a stituation is very good or bad. What matters are the actions you can take.

    Compare outcomes (99%++)

    When comparing actions, what matters is how the world will be after each action (or inaction) you might take. If waving a greeting at your neighbor will probably have better consequences than pushing them over, prefer waving. But if pushing them over has better consquences, perhaps because a homicidal radioactive trolley car has gone off its rails and you're pushing them out of its way, then what matters is not "don't push your neighbors" but "if I push this guy over, they won't get run over".

    Estimation is the best we have (99%++)

    While we want to take the actions that will lead to the best outcomes we don't actually know what those are. This means we always have to estimate some, and we should try to do a better job deciding about actions with larger potential effects.

    Other people matter (99%+)

    If I take your cake and eat it, I get to enjoy the cake and you will be sad. I may still think my happiness outweighs your sadness, but I need to consider them both in order to determine whether the world would be better if I took your cake.

    All people matter about equally (98%)

    I'm not special, you're not special. People in my country aren't more important than people in other countries. People in the future are just as important as people now.

    Value is proportional to number of people (98%)

    Kicking ten random strangers is about ten times as bad as kicking one. Giving 100,000 random children chocolate bars is about 100 times as good (or bad) as giving them to 1,000. A given consequence for N people is N times more important (good or bad) as it happening to only 1 person. Duplicating the universe would be twice as good/bad as anihilating it. There is an "all else being equal" here: suddenly taking the number of people on earth from 7B to 7T would be terrible because huge numbers would die unpleasantly (not to mention that this would probably have a dramatic negative effect on the number of future people).

    Some kind of utilitarianism (95%)

    Every person at every moment has some number that represents their utility. Total utility is the sum of all people's utilities over all time. Actions that lead to higher total utility are better. Utility could be net happiness, net preference satisfaction, or something more complicated. There are things I dislike about all suggested definitions of utility I've seen, but it still seems like roughly the right idea.

    Living humans are what count (90%)

    The border between humans and animals isn't sharp, but somewhere between babies and apes (inclusive) I think they stop mattering for their own sake (their utility is zero) and only remain relevant for the sake of other people. Similarly, once people die they only matter in as much as remaining people care about them. Raw joy/suffering isn't what matters, and building a happy neuron farm wouldn't be valuable, but an uploaded person would still count. (I'm using 'people' here and above as shorthand for "the moral entities that matter". If I changed my mind here and started including trees as entities that mattered for their own sake, then everything above would also apply to trees.)

    Comment via: google plus, facebook

    Recent posts on blogs I like:

    Interview with Kat Woods: decision-making about having kids

    Realizing what you don't want The post Interview with Kat Woods: decision-making about having kids appeared first on Otherwise.

    via Otherwise July 5, 2022

    Decision theory and dynamic inconsistency

    Here is my current take on decision theory: When making a decision after observing X, we should condition (or causally intervene) on statements like “My decision algorithm outputs Y after observing X.” Updating seems like a description of something you do…

    via The sideways view July 3, 2022

    10x (engineer, context) pairs

    Your actual output depends on a lot more than just how quickly you finish a given programming task. Everything besides the literal coding depends deeply on the way you interact with the organization around you.

    via benkuhn.net June 9, 2022

    more     (via openring)


  • Posts
  • RSS
  • ◂◂RSS
  • Contact