One Mechanism for Rational Suicide

Here is one reason you might rationally commit suicide for selfish reasons.

Suppose you care about some other peoples’ perception of you after you die (this can be rationally justified – for example, caring about other peoples’ perception of you after you die is a shortcut that allows other people to more easily predict that you will help them, and this makes it easier to make teammates). You also think that other peoples’ perception of you is better than who you really are. Here are some reasons this might happen:

• You try to present yourself well to other people
• You have impostor syndrome
• You are being issued a private punishment
  • If someone or something is punishing you and no one else knows it, then other people won’t factor in the punishment when calculating their expectation for future you, and there’s a good chance they won’t even after the punishment has taken effect.
  • If someone is privately punishing you, regardless of whether that person reveals that he or she was punishing you or not after you die, people’s perception of you won’t go down. It may even go up if such a thing is revealed, if it shows that you were previously operating under punishment.
• From a slightly different direction, you think you won’t be criticized after your death. This doesn’t mean that others’ current perception of you is better than your perception of yourself, but it does suggest that others’ perception of you after you die will be higher.

If you keep living longer, then you think that your own perception of yourself with manifest, which will lower everyone else’s perception of you, but if you commit suicide then no one will find out.

In other words, your future reality is like a random variable that will realize a value that everyone will observe some time in the future, but if you commit suicide then everyone will observe it as their current prediction of that value, which you think is higher than what would have been realized. An additional plus is that this removes all variance.

If this is someone’s reason for committing suicide, the ways you can try to prevent that person from committing suicide include:

• Convincing him/her that peoples’ perception of him in the future will be higher than he thinks; I have 2 mechanisms for this:
  • Convincing him/her that something he thinks will be perceived negatively won’t actually be perceived as negatively, or perhaps positively; similarly, convincing him/her that he is underestimating how positively something will be perceived
    • For example, you might tell him/her “I bet you can’t do X” where X is some task that most people wouldn’t expect him or her to be able to do but he or she knows him/herself to be able to, especially if you talk about how being able to do X could easily lead to a lot more things (though depending on how hard it is, it might be best left to him/her to figure out by him/herself).
  • Convincing him/her that his/her current perception of his/her future self is too low
    • For example, you might remind him/her of a project he/she is working on that no one else knows about right now, but could become known in the future. Since no one else knows about it, they aren’t taking it into account for estimating their perception of him/her in the future, thereby underestimating him/her.
• Convincing him/her that peoples’ current perception of him are lower than he thinks
• Introducing him/her to new people who don’t yet have an expectation of who he/she is, although this would depend on how he thinks he compares to those peoples’ priors of a random human. The point is that you want him/her to care about more peoples’ perception of him by people who underestimate him.

A hypothesis for the price of art

Disclaimer: I am definitely more of a “STEM person” than a “humanities person”. Before today, I did not have a hypothesis for why art is so expensive (I don’t believe the “rich people moving money around” would be sufficient to make it this expensive). Much of my experience with art is from having been to art museums (as I will reference later on), and I haven’t been to many of those either.

The central idea to this post is: I would pay to think some of the thoughts of other people, especially those who are smarter than me (but also those who are not – the conclusions of people who are not smarter than me could well take me a significant time to think of, especially if I am living in a different context). I think this is an underlying theme throughout the humanities, and I think this plays a significant role in why some art is so expensive.

In the previous paragraph, I use the word “smarter”, but there could be other adjectives that fit here as well. I think “smarter” works for me (though it is by no means the only adjective that correlates with the people whose thoughts I want to think), but other people might want to think the thoughts of people who satisfy other adjectives.

To be clear: I am not talking about the paintings that only cost a few hundred dollars.  I think these can mostly be accounted for by paying the artist at an hourly wage most people would find reasonable plus the cost of materials.

Also, I find it plausible that some people find looking at art aesthetically pleasing, but I don’t think this by itself would account for the price of millions of dollars except possibly in cases where the cost of materials is also millions of dollars.

1. Summary of Hypothesis

For the more expensive pieces of art (e.g. the kind that shows up in art museums), a buyer might think that by looking at the art in detail, the he or she could recover the artist’s original thoughts. This is consistent with what I hear when I ask people what they find interesting in a piece of art (e.g. when I’m visiting a museum).

Assuming this hypothesis, we would expect art that is likely to express unique ideas, art that is more likely able to express ideas well, and art made by smarter people to be more expensive.

• Unique ideas: Someone who consistently creates new kinds of art might be more likely to think of ideas radically different to those I have thought of before, and a buyer might value this.
• Expressing ideas well: Obviously, someone who can use an art style well can express more using that style than someone who cannot use that art style well. Less obviously, someone who is proficient in many art styles would be more expressive because they can choose which art style they think would represent their idea best, and the buyer can use this information to think about how the artist thought through his or her idea. Also, artists you are familiar with would be more expressive to you because you know more about how they think – this could be one reason modern “minimalistic” art is bought: buyers want to understand the artist’s thought processes better in order to understand other artwork by the artist. This can be partially tested – if this is true, then this art should be purchased more if the artist has also made creations in other art styles.
• Smarter people: I would generally prefer to think the thoughts of smarter people than less smart people.

This hypothesis also makes it not inconsistent that forgeries that the buyer cannot distinguish from the real thing are worth much less: A forgery might look similar, but now you lose the ability to analyze the original thoughts of the artist.

Furthermore, it is not implausible that the buyer cannot distinguish between thoughts of smart people and other thoughts, so even if he started analyzing the forgery, he or she might not be able to tell any distinction in value. In this setting, the buyer would be justified in having someone else certify that the artwork is not a forgery.

2. Comparison to Books

You might say that we have another thing we pay for that we use to think the ideas of other people: books. At first it might look like we have a contradiction: Even taking “a picture is worth a thousand words” literally, we would expect a piece of artwork to be worth only 1000 times more than a book, yet I am not aware of any books worth 10,000 USD because a buyer wants to read the ideas it contains.  However, I think this can be explained by supply and demand: When an author (and editor, publisher, and anyone else involved) writes a book, it is not much more expensive to print 2 copies of a book in comparison to printing 1 copy. So perhaps some some books could have one copy sold at 10,000 USD for its ideas, but it is likely that the author and anyone else involved could sell many more copies at a lower price, so we don’t see such expensive books.

Also: I think we would pay more for books (including both fictional and nonfictional) that are more likely to express unique ideas, books that are more likely able to express ideas well, and books written by smarter people.

3. Teaching

When someone tries to explain to me why they find art interesting (e.g. on an art museum tour), they typically start explaining what the artist was thinking and seem to expect me to find it interesting. However, I have no idea who this artist is, and the ideas being expressed (e.g. as explained by the tour guide) do not seem relevant to me, so I am not interested. So I think starting with what the artist was thinking is the wrong place to start for some pieces of art, or at least how the way in which we understand what the artist was thinking at the time impacts our understand other artwork should be shown. I would expect museum guides to have this figured out by now if it were that simple, though, so maybe doing this is just hard (or maybe it’s hard to do this without making a tour that looks like it’s made specifically for dummies).

Here is an analogy: in mathematics, we are often interested in problems that do not have immediate application, and part of the reason we are interested in these problems is that they will help us understand the landscape of mathematics. But we cannot expect students who have never liked studying math to be interested in these problems without this context, even though it might be hard for someone who has always been interested in math to conceive of. If you want to teach math, starting with “Look at all these problems this can solve” when all your examples are not directly relevant won’t help.

Conclusion: if you want to make someone interested in something it is not only important to give context but also the right context.

Cases when you want to update your utility function

Your current utility function defines a goal for yourself.  In achieving this goal, sometimes it is useful to change this function.  Here are some examples; I may expand this list as I find more examples:

1. Subgoals: If you’ve played a little chess like me, you know that things like controlling the center and having a space advantage are good, even if you don’t know ahead of time what to do with these things.  Yet when I do these things, I find that it is easier for me to make moves that lead to a win.  So it is useful to add “controlling the center” and “having a space advantage” as “subgoals”, which means I have a higher utility if I accomplish these.  This is a large idea behind many of today’s chess engines.
2. Rules, facts, etc.: This is sort of a generalization of the previous point.  Sometimes you want to treat things that are mostly true as objective facts, even if they aren’t.  For instance, if you have an institutional rule, what that really means is that if you break this rule then you will have to face the institutional consequences.  But working out the institutional consequences every time is effortful, so treating these as true rules that you just have to follow saves energy.  Also, if you are risk-averse, this reduces risk.  Also, it may be helpful for other people to know that you follow certain rules blindly, for example with some morality.  For instance, if one of these rules is that you help other people in certain circumstances, then other people might want to help bring about these circumstances on the basis that you will help them.  For instances, if one of these circumstances is that they help you back, this motivates a reason for friendship.
3. Friends, relationships, etc.: This is sort of a generalization of my previous post “Agreeing with your friends may be useful, even when they are wrong” from October 10, and is also a special case of the first point in this list.  If your utility function includes the utility of others (perhaps weighting friends, relationships, etc. more heavily), then you might want to align your utility function with theirs.

Sometimes in the above list I write “sort of a generalization of”.  To make it a true generalization, I should write down what the consequences of updating your utility function are and write down what the advantages of these consequences are, and then check that these are actually generalizations of the advantages of the things I claim I am generalizing.  I’m not in the mood to do this right now, so I won’t.

Where do people learn to bullshit?

In school, I took a long time to learn to bullshit my way through work.  I couldn’t do little work and make it look good.  But somehow everyone else seemed to be able to do it.  I’m not sure if I’m actually worse or if this is just perceived; maybe I’ll test this one day.

Hypothesis: most people start learning to bullshit in elementary school, when their parents apply restrictions on them until they finish their homework.  Since their goal is to finish their homework, bullshitting works just as well here.  But my parents never did that with me, so bullshitting didn’t become a goal for me until the homework got a bit harder to the point where school deadlines (plus procrastination) forced me to bullshit… but by then I’ve missed out on a lot of practice.

If this is true, then students that did not have these restrictions should have a harder time bullshitting.  At the very least, they will start later, so they will see that all their peers are better at bullshitting when they first start to try, so they would perceive that they are worse at bullshitting than their peers based on first impressions.

Anyone have a similar experience?

Politics is like Sports

I feel like, for this post in particular, I should explicitly say that I know next to nothing about either politics or sports.  In general, I know next to nothing about what I’m blogging about, but in this case it feels more true.

Anyway: At least one reason people like watching politics is the same as at least one reason people like watching sports: If you like doing something, you will like watching seeing someone do it at an expert level, provided that it at least looks somewhat similar on a superficial level.

Politics involves extremely strategic players doing conflict resolution and arguing, which are things we do a lot in real life.

Tangent (Warning: The content of this tangent as-is is probably obvious enough that it’s not worth writing; I’m leaving it here because I wouldn’t be surprised if something nontrivial could branch off of this):

Can people learn strategies for these things by watching politics (like some might hope to do by watching sports)?  Maybe – but we have to keep in mind that the target audience for much of politics is significantly different from the target of most day-to-day conflict resolution and arguing.  Policy that is supported for political popularity is designed to be supported by a large number of people, while that might not be your goal in day-to-day conflict resolution and arguing.  Expert-level conflict resolution and arguing between large countries is also a sufficiently different dynamic that we might not expect strategies from that arena to translate.  Players are using extreme amounts of detailed information about their opponents, and these opponents are very different and can make many different moves than opponents you face in day-to-day conflict resolution and arguing.

This last point is probably also true to some extent for sports – some strategies only work because of the particular composition of your current team and the particular composition of the opposing team, and these compositions might be uncommon in friendly games.

I’m not saying that we can’t learn anything from watching people do politics or sports at an expert level, but if we would like to learn this stuff then probably it would be a good idea to separate what strategies generalize from which strategies are specific to the professional scene.

The IQ argument

Sometimes you’re arguing with someone and they say that their side is correct since they have a high IQ.  Obvious counterarguments to this are that:

1. They must argue that having a high IQ must be correlated to being right regarding the argument topic
2. They must argue that their IQ is as high as they claim

A less obvious counterargument is that they must argue that they actually believe what they are arguing.

As far as I can tell, if they can argue all 3 points, the IQ argument works.  In general, if someone can argue successfully that some kinds of people are likely to get the correct answer to an argument, that they are one of these kinds of people, and that they are arguing what they believe, that is a reasonable argument.

If you have a logical argument against the claim of the other person in the argument, then the weight of the logical argument should be stronger if the 3 points in the previous paragraph are weaker.

Going in the other direction, an ad hominem attack works logically if:

1. A certain type of person is particularly likely to believe the incorrect conclusion of a result
2. A person is of that type of person
3. That person is arguing for what he believes

In this case, when arguing the first point, one would hope to show that this type of person is more likely to get the incorrect conclusion than a random guess – for example, just claiming that someone is stupid isn’t sufficient to show this unless you argue that being stupid biases what he will think of in a way that causes his conclusions to be incorrect more than just picking a random thing.  If the bias towards incorrect conclusions isn’t worse than that of a random picker, then the fact that the person chose something as an answer doesn’t suggest that it is wrong.

Arguing by Using Facts

Let’s say someone presents an argument to you and uses a fact. If you want to check the validity of the argument and you are not already familiar with the fact, you should check that the fact is correct and used correctly in context. If someone presents an argument to you an uses 100 unfamiliar facts, doing this check for all 100 facts becomes infeasible. Of course, we can use heuristics to reduce the number of facts we need to check, but the general idea still holds. So maybe we shouldn’t trust arguments that use too many unfamiliar facts. Going the other direction, if you want to convince someone of something, don’t use too many facts that will seem unfamiliar, though this requires us to leave out context (I think a case can be made that this requires us to leave out both the context that helps our argument and the context that hurts our argument).

Note that I used the word “unfamiliar” and not “unlikely”.  There are plenty of false statements which I would find likely though unfamiliar. Likeliness is a heuristic we use to decide which facts to check.  So if we do want to use many unfamiliar facts, and our goal is to convince someone else, they should be biased towards feeling likely (and ideally the main part of the argument should not be these facts). Perhaps facts that can be especially easily verified don’t need to feel so likely – unless you think the other party wouldn’t feel the need to check facts if the minimum likeliness of all your facts isn’t below some threshold.  But if we bias facts towards those that sound likely, then “sounding likely” is less good of an indicator as to whether the fact is true.

The point is that maybe we shouldn’t trust arguments that use too many unfamiliar facts, even if they sound likely.

Question 1: How does this apply to teaching? At least part of our goal in teaching is to convince the students of something.  However, when teaching, leaving out context is possibly worse.  Meanwhile, the curriculum likely contains many facts that don’t have simple, accurate arguments that don’t use too many facts.

Question 2: Which types of arguments tend to use more facts, and which tend to have the highest proportion of unfamiliar facts? For example: arguments between friends, political arguments, sales pitches?

Agreeing with your friends may be useful, even when they are wrong

Before we start with the post: I will clarify that, in the title, I mean truly agreeing with and believing your friends – not just saying you agree with them.  I thought about using the word “believing” instead of “agreeing”, but that suggests I mean “trusting” which I don’t.  Trusting your friends might be useful for other reasons, but that is not the point of this post.

On with the post.

If you hear something enough times, you will often believe that it has some truth to it, unless you have a strong reason to believe it is just a common misconception.  This isn’t a bad prior in my experience – it seems to me that just because no one I associate with takes a certain position doesn’t mean there isn’t a good argument for it.

But a stronger statement is also true – if you hear something enough times, you will often believe that it is true.  And this might not be terrible either – if you hear something enough times, probably many of the people around you believe it.  Also, if you make yourself more similar to the people around you, that makes it easier for you to determine how your actions will influence them, since you can just see how you would respond to your own actions (not like this is actually easy, but at least it gives you somewhere to start).  Consequently, if you want to influence the people around you in a predictable way, you might want to be more similar to them.  For example, you might want to believe things that are similar to what they believe.  In this case, having similar beliefs might make it easier to see what sorts of arguments would convince them.  And if this belief is technically wrong but not too significant, it may well be that believing this wrong thing is better than believing the correct thing. (I note knowing each truth gives you a finite about of benefit, so it’s not hard to imagine that some benefit from something else is greater.)

One may argue that it would be better to know the correct thing and just keep in mind that most people believe the wrong thing, but simulating ignorance of something is hard.

There is another way in which believing things your community believes may be helpful, even if it is false.  Let’s say you believe the world has a model A, whereas someone you’re trying to talk to has model B.  If you try to have an argument about something where model A and model B disagree, even if the topic isn’t something the other person has thought much about, it is not unlikely that you would have to argue about many related things in the model.  In contrast, if your model is similar to his, then you would have to change less about his model to convince him.

As a corollary of the title of the post, it is useful to have similar morals as your friends.  It is useful to care about other humans because your friends care about other humans.  If your friends have morals that cause them to be vegetarians (e.g. caring about animals feeling pain), then it is useful to also have similar morals, which might cause you to be a vegetarian for moral reasons.