LVSN's Shortform

The Principle of Nameless Heartsmarts: It is pointless to complain that I acted with propriety if in the end I was too dense to any relevant consideration.

Arbitrary incompleteness invites gameability, and arbitrary specificity invites exceptioncraft.

Just because there are mottes and baileys, doesn't mean the baileys are wrong; they may just be less defensible in everyday non-ideal speech situations.

Some subset of those who agree that 'when two people disagree, only one of them can be right' and the people who agree that A : A := 'when two people disagree, they can both be right' such that A ≈ A' and A' := 'when two people "disagree," they might not disagree, and they can both be right', do not have a disagreement that cashes out as differences in anticipated experiences [LW · GW], and therefor may only superficially disagree.

Note 1: in order for this to be unambiguously true, 'anticipated experiences' necessarily includes anticipated experiences given counterfactual conditions [LW · GW].

Note 1.1: Counterfactuals are not contrary to facts; they have attributes which facts can also share, and, under varying [LW · GW] circumstances [LW · GW], the ratio of [the set of relevant shared attributes] to [the set of relevant unshared attributes] between a counterfactual situation and known situation may be sufficiently large that it becomes misleading to characterize the situations as [opposite] or [mostly disagreeing, as opposed to mostly agreeing]. A more fitting word would be 'laterofactual'.

Note 1.1.1: When people say that B : B := 'C and D disagree', the set of non-excluded non-[stupidly interpretable] implicatures [LW · GW] of the statement B includes that E : E := 'C and D mostly disagree', and not only F : F := 'C and D have any amount of disagreement'.

What is normally called common sense is not common sense. Common sense is the sense that is actually common. Idealized common sense (which, I shall elaborate, is the union of the set of thoughts you would have to be carefully trying to be common-sensical in order make salient in your mind and the set of natural common sense thoughts) should be called something other than common sense, because making a wide-sweeping mental search about possible ways of being common-sensical is not common, even if a general deference and post-hoc accountability to the concept of common sense may be common.

No one will hear my counter-arguments to Sabien's propaganda who does not ask me for them privately. Sabien has [? · GW] blocked me for daring to be unsubtle with him. He is equally welcome as anyone else to come forth to me and exchange considerations. I will not be lured into war; if it is to be settled, then it will be settled with words and in ideal speech situations.

People defend normal rules by saying they're "not arbitrary." But if they were arbitrariness minimizers the rules would certainly be different. Why should I tolerate an arbitrary level of arbitrariness when I can have minimal instead?

Your policy's non-maximal arbitrariness is not an excuse for its remaining arbitrariness. 

I do not suggest the absence of a policy if such an absence would be more arbitrary than the existing policy. All I want is a minimally arbitrary policy; that often implies replacing existing rules rather than simply doing away with them. Sometimes it does mean doing away with them.

When you say "all else should stay firm belief" do you mean "all else should be regarded as belief"? Also, was the word 'firm' in 'firm belief' playing any role there or can I just get rid of it? 

I think all propositions should be subject to tests whether they are regarded as knowledge or belief.

Feature suggestion: Luigi-vote vs Waluigi [LW · GW]-vote buttons. Too many people downvote not because the content is bad but simply because it's wa.

... Actually, I guess people would just downvote and wa-vote. Problem not solved.

Are there any considerations I am being dense to in the formulation of this theorem? As a LessWrong user, feel free to be really really negative to me to compensate for my denseness, even if you must speak fallaciously. If I am misled then there is no point in complaining that I acted with propriety. You are welcome to babble [LW · GW].

1. No one is teaching Bayesianism from true first principles, which would involve exploring the relevant consequences of every tempting(-from-some-perspective) mutation on the set of definitions and instruction steps that is called Bayesianism

2. "There are aspects of good reasoning that we don’t yet understand, even in principle." — Nate Soares, A Guide to MIRI's Research

3. Most epistemic systems in history have been half-misleading in a way that would have become obvious through more thorough foundational investigation

4. Therefor, Bayesianism is probably half-misleading in some way.

If AI takeoff was not upon us, the rationalist community would perish from the simple inability of its users to call each other idiots. More specifically the inability to engage in directed babbling [LW · GW]-style dialogue that starts off very negative.

Meta-Contrarianism [LW · GW] Is Good, Actually

An untitled litany; title suggestions welcome:

I wish an omnibenevolent god ruled over all living beings in the cosmos so that everyone could have as much as they want and so that intrinsically bad things happened minimally. It is okay to want a thing without its normal consequence. It is okay to want one thing out of a linked pair.

1. Warren Buffett consistently beats the EMH as an exceptionally smart person 

2. Prediction markets systematically outperform all individual smart people 

3. Therefor, if futarchy is implemented, we will inevitably end up with a functional planned economy.

Something that I think rationalists and EAs don't take seriously enough: it is a nice and altruistic gift to desensitize people to defections on especially fake types of morality.

If people's net worth should in principle include how much money they give to charity, it should also in principle include how much status they give up to defect with fake morality.

When someone says something with low editing that you like, you call it blurting. [EA · GW] When someone says something with low editing that you don't like, you call it quacking.

It's Russell conjugation.

I get the sense that expeditiousness could be the 14th rationalist virtue (after the 13th, paranoia).

https://www.thesaurus.com/browse/expeditious

There's a kind of manipulation which gets little discussion due to its invisibility.

Most people start out with some uncertainty about every form of social interaction. 
Then someone associates a social interaction with emotionally loaded language and undesirability, and then most people, without the benefit of appreciating that they can think and interpret autonomously in the opposite direction with some degree of plausibility, just go with the flow of what other people are saying is undesirable. 
Then, if someone tries to fight this gradual narrowing of tolerance with arguments which explicitly appeal to people's conscience by trying to consider all the plausible perspectives on the matter, which is far more respectful to their intellectual autonomy than letting only one view dominate, it gets processed as manipulation, guilt-tripping, and antisociality because they are coming from a position which is unpopular due to the prior indoctrination.

Getting people to be as tolerant of innocent things as they naturally are is always an uphill battle against puritans and censors.
People become puritans and censors in the first place because they believe they cannot solve their disagreements with logical argument, so they try to paint those which they feel tensions with as bad people in whatever way they can. So if Hitler ate sugar, and sugar is a niche rather than popular luxury, you start portraying sugar as evil, and the tolerance of respectable society narrows further: no more sugar for the few who enjoy it and don't like hurting people.

My impression is that when rationalists make objections, they tend not to explicitly distinguish between correcting failure and revealing possible improvements. 

If A is abstractly true, and B is 
1. abstractly true 
2. superficially contradictory with A
3. true in a more relevant way most of the time to most people

I expect rationalists who want to prioritize B to speak as if issuing corrections to people who focus on A, instead of being open-minded that there's good reason for A in unrecognized/rare(ly considered) but necessarily existing contexts, and instead of offering their personal impression of what an improvement would look like as merely that: a personal impression.

In spite of this, I still love you guys more than any other culture; love your ambition, clarity of judgment, and charitability. I'm not a post-rat; I struggle with rationality.

Is there a decision theory which works as follows? (Is there much literature on decision theories given iterated dilemmas/situations?)

If my actual utility doesn't match my expected utility, something went wrong.
Whatever my past self could have done better in this kind of situation in order to make actual utility match the expected utility is what I should do right now. If the patch (aka lesson) mysteriously works, why it works isn't an urgent matter to attend to, although further optimization may be possible if the nature of the patch (aka lesson) is better understood.

In this shortform, I want to introduce a concept of government structure I've been thinking of called representative omnilateralism. I do not know if this system works and I do not claim that it does. I'm asking for imagination here.

Representative (subagent) omnilateralism: A system under which one person or a groups of people tries to reconcile the interests of all people(/subagents) in a nation (instead of just the majority of the nation) into one plan that satisfies all people(/subagents)*

I think "representative democracy" is an ambiguous term which can be used to mean some mix of representative ochlocracy or representative omnilateralism whenever the situation is convenient. The reason we like democracy is because it approximates direct omnilateralism better than other alternatives, but if we will permit *representative* democracy, why not representative omnilateralism? Much as direct democracy is a purer ochlocracy in theory, representative omnilateralism is a purer elitism and a purer defense of the less fortunate in theory, but direct omnilateralism literally has the best of all worlds. 

My impression (not verdict) is that direct omnilateralism is impossible in practice only because people are not equipped to negotiate optimally. If everyone was better at negotiating, we would have way fewer conflicts and far more business.

*(I mention subagents because people often do not accept parts of themselves which are innocent, which is a personal mistake as well as a commons mistake; direct subagent omnilateralism is an even higher aspiration than direct superagent omnilateralism)

When a bailey is actually true, there are better ways to show it - in those cases they ARE in the motte.

Endorsed.

To whatever extent culture must pass through biology (e.g. retinas, eardrums, stomach) before having an effect on a person, and to whatever extent culture is invented through biological means, cultural inputs are entirely compatible with biological determinism.

Deadlines: X Smooth, Y Sharp

Recently an acquaintance told me we had to be leaving at "4:00 PM sharp." 

Knowing of the planning fallacy, I asked "Sharp? Uh-oh. As a primate animal, I naturally tend not to be very good with sharp deadlines, though your information is useful. Could you tell me when we're leaving smooth?"

"Smooth? What do you mean?"

"Smooth as opposed to sharp. Like, suppose you told me the time I should be aiming to be ready for in order to compensate for the fact that humans are bad at estimating time costs. Let's say you wanted to create a significant buffer between the time I was ready and the time I had to be ready by; the beginning of that significant buffer is the time we're leaving by smooth."

Since then, we've been saying things in the structure "X smooth, Y sharp", where X and Y are times or amounts of time. It's intuitive, catchy, simple, and very useful.

My response to it is: What makes you think it is naive idiocy? It seems like naive intelligence if anything. Even if the literal belief is false, that doesn't make it a stupid thing to act as if true. If everyone acted as if it were true, it would certainly be a stag-hunt [LW · GW] scenario! And the benefits are still much worthwhile even if the other does not perfectly cooperate. 

Stupid uncritical intolerant people will think you look childish and impertinent, but intelligent people will notice you're being bullied and you're still tolerating your interlocutor, and they will think you're super-right. You divide the world into intelligent+pro-you and stupid+against-you.

Also I might note that your attempted counter-example has an implied tone which accuses naive idiocy, rather than sounding curious with salient plausibility. The saliently plausible thing, in your attempted counter-example, is an implicit gesture that there is not a difference.

Lately I've been thinking about what God would want from me, because I think the idea was a good influence on my life. Here's a list in progress of some things I think whould characterize God's wants and judgments:

• 1. God would want you to know the truth
• 2. If you find yourself flinching at knowledge of serious risk factors (e.g. of your character or moral plans), God would urgently want to speak with you about it
• 3. Resist the pull of natural wrongness
• 3.1. Consider all of the options which are such that you would have to be looking for the obvious/common sense options in order to find them
• 3.2. Consider many non-obvious options; consider that the right thing to do is a different concretization of an abstracted version of the wrong thing to do, is adjacent to the wrong or seemingly-right thing to do, queers the seemingly-right or wrong thing to do, or is a thing in a category which cuts sideways through categories of abstractly right or wrong things to do
• 3.3. Every night, go over a list in progress of cognitive biases and search your memories and feelings honestly as to whether you gave into any of them
• 4. By one third of the set of good definitions of 'making progress' that you can come up with, or by no more than six good definitions out of eighteen, make it 80% true about you that you are making progress; don't be going nowhere
• 4.1. On an average rate of twice every five days, do a good day's work
• 4.2. On an average rate of once every three weeks, spend a day working really hard
• 4.3. For every extra amount of work beyond the rates specified above, God will be extra proud of you, which can become a source of great esteem and comfort.
• 5. Reward yourself temperantly for making progress and resisting the pull of natural wrongness; your morality should be as an enlightened, wiser-than-you friend who you eagerly wish you were strong enough to follow; not a slaveholder making you regret your acquaintanceship.
• 6. In your life, always be faithful and reliable to at least one great moral principle; have one moral job or nature that God will consider you remarkable for
• 7. Recognize the vulnerability of others as unsettingly reminiscent of the vulnerability in yourself

Feel free to leave suggestions for more entries; aim for excellence, and if you feel honestly that your suggestion is excellent in spite of acknowledged strong possibilities that it may be subjective and biased, don't hesitate to share. Or, hesitate the right amount before sharing; either is good.

Interesting stuff from the Stanford Encyclopedia of Philosophy:

2.8 Occam’s Razor and the Assumption of a “Closed World”

Prediction always involves an element of defeasibility. If one predicts what will, or what would, under some hypothesis, happen, one must presume that there are no unknown factors that might interfere with those factors and conditions that are known. Any prediction can be upset by such unanticipated interventions. Prediction thus proceeds from the assumption that the situation as modeled constitutes a closed world: that nothing outside that situation could intrude in time to upset one’s predictions. In addition, we seem to presume that any factor that is not known to be causally relevant is in fact causally irrelevant, since we are constantly encountering new factors and novel combinations of factors, and it is impossible to verify their causal irrelevance in advance. This closed-world assumption is one of the principal motivations for McCarthy’s logic of circumscription (McCarthy 1982; McCarthy 1986).

3. Varieties of Approaches

We can treat the study of defeasible reasoning either (i) as a branch of epistemology (the theory of knowledge), or (ii) as a branch of logic. In the epistemological approach, defeasible reasoning can be studied as a form of inference, that is, as a process by which we add to our stock of knowledge. Alternatively, we could treat defeat as a relation between arguments in a disputational discourse. In either version, the epistemological approach is concerned with the obtaining, maintaining, and transmission of warrant, with the question of when an inference, starting with justified or warranted beliefs, produces a new belief that is also warranted, given potential defeaters. This approach focuses explicitly on the norms of belief persistence and change.

In contrast, a logical approach to defeasible reasoning fastens on a relationship between propositions or possible bodies of information. Just as deductive logic consists of the study of a certain consequence relation between propositions or sets of propositions (the relation of valid implication), so defeasible (or nonmonotonic) logic consists of the study of a different kind of consequence relation. Deductive consequence is monotonic: if a set of premises logically entails a conclusion, than any superset (any set of premises that includes all of the first set) will also entail that some conclusion. In contrast, defeasible consequence is nonmonotonic. A conclusion follows defeasibly or nonmonotonically from a set of premises just in case it is true in nearly all of the models that verify the premises, or in the most normal models that do.

The two approaches are related. In particular, a logical theory of defeasible consequence will have epistemological consequences. It is presumably true that an ideally rational thinker will have a set of beliefs that are closed under defeasible, as well as deductive, consequence. However, a logical theory of defeasible consequence would have a wider scope of application than a merely epistemological theory of inference. Defeasible logic would provide a mechanism for engaging in hypothetical reasoning, not just reasoning from actual beliefs.

In defense of strawmanning: there's nothing wrong with wanting to check if someone else is making a mistake. If you forget to frame it as a question, e.g. "Just wanna make sure: what's the difference between what you're thinking and the thinking of what my more obviously bad, made-up person who speaks similarly to you?" Then the natural way it comes out will sound accusatory, as in our typical conception of strawmanning. 

I think most people strawman because it's shorthand for this kind of attempt to check, but then they're also unaware that they're just trying to check, and they wind up defending their (actually accidental) apparent hostility, and then a polarization happens.

Strawmanning happens when we take others' judgments as plausible evidence of more general models and habits that those judgments play a part in. By asking for clarity of what models inform a judgment, we can get better over time at inferring models from judgments. It can become a limited form of mind reading.

One thing to say about negation is that often the model uncertainty is concentrated in the negation. Any probability estimate, say of A (vs. not-A) always has a third option: MU="(Model Uncertainty) I'm confused, maybe the question doesn't make sense, maybe A isn't a coherent claim, maybe the concepts I used aren't the right concepts to use, maybe I didn't think of a possibility, etc. etc.". 

I tend to think of writing my propositions in notepad like
A: 75%
B: 34%
C: 60%

And so on. Are you telling me that "~A: 75%" means not only that ~A has a 75% likelihood of being true, but also that A vs ~A has a 25% chance of being the wrong question? If that was true, I would expect 'A: 75%' to mean not only that A was true with a 75% likelihood, but also that A vs ~A is the right question with 75% likelihood (high model certainty). But can't a proposition be more or less confused/flawed on multiple different metrics, to someone who understands what this whole A/~A business is all about?

My shortform post yesterday about proposition negations, could I get some discussion on that? Please DM me if you like! I need to know if and where there's been good discussion about how Bayesian estimate tracking relates with negation! I need to know if I'm looking at it the wrong way!

Does thinking that A is 45% likely mean that you think the negation of A is 5% likely, or 55% likely? Don't answer that; the negation is 55% likely.

But we can imagine making a judgment about someone's personality. One human person [LW · GW] accepts MBTI's framework that thinking and feeling are mutually exclusive personalities [? · GW], so when they write that someone has a 55% chance of being a thinker type, they make an implicit not-tracked judgment that they have an almost 45% chance of being a feeler type AND not a thinker, but a rational Bayesian is not so silly of course; being a feeler and/or a thinker are two independent questions [? · GW], buddy.

The models in a person's mind are predictable from the estimate on his paper, and while his estimate may be true, the models the predictions stem from may be deeply flawed [LW · GW].

By the logic of personality taxonomy and worldly [LW · GW] relations, "the negation of A" has many connotations [LW · GW].

Maybe the trouble is with the words 'negation', 'opposite', and 'falsehood' instead of using the word 'absence'. Presence of falsehood evidence is not the same as absence of truth evidence, even if absence of truth evidence is one kind of weak falsehood evidence to be present.

I want to DM about rationality. Direct messaging is not subject to so many incentive pressures. Please DM me. Please let me be free. 

Please DM me please DM me please DM me please DM me * 36

I'm looking for someone who I can share my half-baked rambly thoughts with. Shortform makes me feel terrible. 

Please DM me; let me be free; please DM me; let me be free * 105

What is 'knowing'? This is not an arbitrary question. They say that curiosity is wanting to know. Making new considerations makes you more efficient. Someone who wants to become more efficient will therefor look for new considerations and behave as if curious. 

Bayesians appear to perform well epistemically, allegedly because they feel a desire to know; because they feel curiosity. But I expect Bayesians would perform just as well if not better if they desired to be efficient, which, again, will look approximately identical to curiosity.

Another type of reasoner may desire to get closer to omniscience. Omniscience means having awareness of all information.

If you admit you are not omniscient about a question, then you admit there is more information that could shift your belief, which would make you more efficient.

Yet those who allegedly desire to know seemingly claim to often know the correct answers to questions even when they are not omniscient about those questions. What could they mean by 'know'? Are they wrong that they know? Do they have sufficient semiscience in some sense? How could sufficient-semiscience-for-knowledge be defined in a way that can't be gamed? 

I suspect there is no such ungameable definition. If there is no ungameable definition, then it would seem 'desire to know' is also gameable if it does not mean 'desire to consider everything' (become omniscient).