Productive Disagreement Practice Thread: Double Crux

post by SilentCal · 2017-10-02T18:25:29.643Z · score: 19 (6 votes) · LW · GW · 37 comments

Contents

  Thread Rules:
None
37 comments

Inspired by the recent skepticism [LW · GW] of the double crux technique [LW · GW], I thought I'd launch a thread to try using double crux, and productive disagreement in general, with other LessWrongers. I think this will work best with some structure, so I've laid out some rules.

Thread Rules:

  • Discussions are to be one-on-one. Do not jump into others' debates.
  • The default platform for discussions will be the comment thread, but participants can use others at their discretion, such as instant messaging platforms or maybe even video chat. Try to keep a shareable record so others can potentially learn from it.

  • To participate in the thread, either

    • make a top-level comment listing beliefs that you think might generate productive disagreements, as well as any preferences you have about discussion format,

    • or reply to someone else's top-level comment, selecting one of their beliefs that you disagree with. Don't make any arguments in this reply, just say which belief you're selecting, and what platform you'd like to discuss it on if the top-level commenter gave multiple options.

  • The top-level poster and the replier then conduct their discussion in replies to that reply, or in the agreed-upon outside platform.

  • Inflammatory topics are better discussed out-of-band. When listing a topic you think likely to be inflammatory, flag it as such and don't offer in-comments discussion as an option.

  • Remember to try to find double cruxes [LW · GW]!

If you want to comment about the rules or the general idea of this thread, there's a meta post [LW · GW] for that.

37 comments

Comments sorted by top scores.

comment by lahwran · 2017-10-03T01:24:44.385Z · score: 8 (3 votes) · LW(p) · GW(p)

Open threads/repository posts do not fit in with lesswrong's historical and reinstated truth seeking focus for the main board; they should be reserved for more discussion-focused areas of the site.

comment by SilentCal · 2017-10-03T16:03:17.348Z · score: 3 (1 votes) · LW(p) · GW(p)

I agree that on LW 1.0, this would belong under discussion rather than main. But as far as I can tell, LW 2.0 non-frontpage posts have much less visibility than old discussion posts, to the point that this type of thread would not be viable.

Perhaps our double crux is "Non-frontpage LW 2.0 posts are a viable platform for open-type threads"? Or maybe it's "It's better to be unable to have open-type threads than to crowd the front page with them"?

comment by lahwran · 2017-10-03T17:14:35.192Z · score: 4 (1 votes) · LW(p) · GW(p)

I agree that the policy I propose is annoying and limiting on the site as it is. I'm proposing a policy that would mandate having a new section (one that is already planned, and that I argued for precisely because I value discussions.)

I currently suspect we don't actually have a double Crux, and simply agree. does that seem true to you?

comment by SilentCal · 2017-10-03T18:34:46.555Z · score: 3 (1 votes) · LW(p) · GW(p)

I thought you were suggesting I shouldn't have posted this on frontpage, in which case we'd obviously disagree. If not, then we agree.

comment by lahwran · 2017-10-03T23:15:21.033Z · score: 4 (1 votes) · LW(p) · GW(p)

I think the windows repository thread shouldn't have been on the front page.

comment by lahwran · 2017-10-03T01:18:20.847Z · score: 7 (4 votes) · LW(p) · GW(p)

Modern deep learning is in the same overall model class as the brain despite the remaining differences. Few unknown unknowns remain. Deep learning provides a good guide to understanding the function of the brain, despite that some of the brain's functionality hasn't been replicated yet.

The differences between the low level learning rules used by deep learning (ie, backprop and gradient descent) and the brain[1] do not lead them to fundamentally different behaviors. The brain and deep learning end up learning similar low-level circuits; or in other words, if given the same task and connectome, their low level circuitry will learn close to the same thing.

Most of the interesting remaining properties of the brain are in the architecture of the connectome, the stability of its low level learning rules, and the fact that it's goddamn massive.

[1] we're not sure in general, and I don't know everything we know, but at least STDP and local RL, possibly global-broadcast local rl, possibly local gradient descent or even backprop

comment by whpearson · 2017-10-03T19:17:09.266Z · score: 3 (1 votes) · LW(p) · GW(p)

Can you unpack what you mean by model class a bit?

I think that the connectome is dynamic (see neuro-plasticity) in people over time and deep learning doesn't cover than dynamism. But I am unsure if you expect deep learning to cover that or not.

Happy to discuss here.

comment by lahwran · 2017-10-04T19:43:35.344Z · score: 4 (1 votes) · LW(p) · GW(p)

I mean that deep learning and the brain operate off of the same principles: function approximation via high dimensional geometric inference of circuits. They are in the same part of program-space: feed forward connections through a neuron, and feed back connections (from somewhere) to train it; cost functions that train neurons into circuits which satisfy an objective; on the higher level, multiple layers, gating, recurrence, etc architectural features like local connectivity, fast-weight memory, and internal attention. People in machine learning were surprised to find that backprop is enough to train neural networks; it was widely expected that it wouldn't work, but it's been discovered that neural networks provide various properties that provide good geometric priors about the world we live in - an intuition pump is "if you have enough dimensions, you always have some way to do slightly better", ie there are nearly no local minima. This isn't actually fully true in general, but neural networks seem to be structures that make it true, and I think that this basic principle is also what makes the brain work.

So it looks like "connectome" didn't mean the thing I remembered it meaning - I thought it meant "wiring patterns", but it in fact means "all wiring in the entire thing". I agree that, in this corrected sense, the connectome is almost certainly dynamic. And deep learning certainly covers that dynamism, it's just the fact of weights being learned, and being allowed to be 0. The brain probably has some form of regularization force pushing connections to disappear - pushing weights to 0 - if they aren't useful enough to keep the wires in place. But that's a common effect of regularization in deep learning too; L2 regularization does this to some degree, and L1 regularization does it harder.

Deep learning doesn't learn as many weights. Neurons seem to be willing to form 10^3 to 10^4 (typical in the cortex) up to 10^5 connections (in the cerebellum). The equivalent neuron in deep learning would have to have far more weights, because in deep learning, zero weights still exist in the matrix. The brain is made out of circuits that can actually get rid of zeros.

Do you disagree with my points such that you think it's worth exploring where our disagreement is?

comment by whpearson · 2017-10-04T20:56:33.728Z · score: 3 (1 votes) · LW(p) · GW(p)

function approximation via high dimensional geometric inference of circuits.

I agree that there is a tonne of that going on in the brain. I think I doubt it is all that is there.

Neuroplasiticity in my mind is not just about modifying weights. New neurons are being added constantly. I don't think deep learning covers that.

comment by Raemon · 2017-10-02T19:40:28.268Z · score: 7 (2 votes) · LW(p) · GW(p)

Harry Potter and the Methods of Rationality should be displayed prominently on the front page of this site.

comment by hamnox · 2017-10-02T22:42:44.729Z · score: 6 (2 votes) · LW(p) · GW(p)

15% disagree, would prefer to discuss in person

comment by Raemon · 2017-10-02T23:00:12.684Z · score: 7 (2 votes) · LW(p) · GW(p)

Coolio. Happy to talk. (Though if you only 15% disagree I'm not sure if we actually disagree enough for this to be that noteworthy)

I have some preference to conduct this in a fashion that lets it function as an example others can learn from (I chose an example that seemed hopefully un-fraught enough that a video or chatlog could be made public after the fact). If you're prefer Not That, still happy to talk but I'd like to also reserve the ability to chat with someone else who disagrees more strongly who's up for going public afterwards)

comment by gjm · 2017-10-03T13:02:28.681Z · score: 3 (1 votes) · LW(p) · GW(p)

Disagree with modest confidence. Happy to discuss here. No experience of formal double-cruxing, but (I flatter myself) quite good at identifying actual key points of disagreement etc. informally.

comment by gjm · 2017-10-04T17:30:01.362Z · score: 12 (3 votes) · LW(p) · GW(p)

I'm not sure whether Raemon is interested in doing this given that hamnox kinda got there first, but it doesn't seem like hamnox really disagrees enough with Raemon for it to be worth the trouble. So I'll get started. Raemon, let me know if this doesn't work for you.

Before getting into any specific double-crux machinery, I'll briefly make my case. The front page here plays two obvious roles: (1) it's a natural entry point for established users wanting to see what's new, and (2) it's a place others may encounter, perhaps after seeing some specific thing on the LW website and wondering what else there is there, or perhaps because someone said "hey, there's this thing called Less Wrong; you should take a look". So, how does having HPMOR prominently on the front page serve those two cases?

Case 1 (established users): It's taking up space that could be used for something else. Most LWers already know about HPMOR, and either have read it already or have decided they aren't interested. It's mentioned often enough (I think) that active LW participants are in little danger of never hearing of it. So here I think HPMOR comes out clearly but not hugely negative.

Case 2 (newcomers of various sorts): What's HPMOR-on-the-front-page going to do to these people? (a) Some people will be immediately put off by it. That's bad if they would otherwise have become valued members of the community; it's bad if it leads them to badmouth LW or rationalists to others; it's neutral or slightly positive if they were going to ignore us anyway and just do so slightly faster, or if they would have tried to participate in the LW community but been annoying or stupid and this scares them off. (b) Some people will be intrigued, read some or all of HPMOR, and like it. That's good if they would otherwise have bounced off; it's neutral or slightly good if they would have been valued LW contributors anyway; it's bad if they turn out to be annoying or stupid. (c) Some people will be more or less indifferent to it. By definition it doesn't have much net impact on them. Here, everything depends on the relative numbers of these people, and how much we'd have wanted them to stick around here.

My guess is that we lose more potentially useful and interesting people -- especially if we attach some value to intellectual diversity -- than we gain, because I think most people who encounter rationalism via HPMOR aren't finding HPMOR from the LW website and never will be.

Add to that the other consequences of having HPMOR on our front page (it takes up space that regulars could use better for other things, and it makes us look weirder to people who were never actually going to be LW participants anyway), and I think it comes out a clear net negative.

So, what would change my mind? I could discover some other ways in which HPMOR-on-the-front-page makes a difference, and they could turn out to be very positive. I could be convinced that the numbers aren't what I expect them to be (e.g., that lots of possible future rationalists come to LW's front page, would be intrigued by seeing HPMOR there, and would bounce off something more like old-LW). I could be convinced that there are "higher-order" effects that matter in ways I haven't thought of (e.g., people who would be put off by HPMOR are less likely to be interested in LW, even if HPMOR isn't in their faces when they come there; I think this is probably true but don't think it's a large enough effect to invalidate what I've said above). I could be convinced that actually what matters most is how things are for regular LW participants (this might not be hard to convince me of) and that regular LW participants benefit on balance from having HPMOR take up a big chunk of the front page (this would be a harder sell).

By way of background in case it matters: I have been here since the Overcoming Bias days and in particular didn't come via HPMOR; I have read HPMOR and enjoyed it.

comment by Raemon · 2017-10-04T18:48:39.538Z · score: 12 (3 votes) · LW(p) · GW(p)

Thanks, and sorry I took awhile to respond (I started last night but then realized I was a bit more uncertain/confused about some key points than I thought. FYI I chose this point because it seemed moderately important and I wasn't overwhelmingly confident)

I will reply in more detail soonish.

comment by Raemon · 2017-10-07T23:09:50.519Z · score: 8 (2 votes) · LW(p) · GW(p)

Thanks for the response. First, I want to clarify two separate questions at play:

I. Should there be a giant front-page of legacy content?

My answer is "Yes, for new users, no for older users." It's more important for newcomers to see our best content than the latest discussions, and to have proper context when they join the latest discussions. Older users should see links to sequences they're in the middle of reading, maybe a prompt to read the most popular or important things, but there should be some way to remove that prompt if the writing clearly isn't for them.

2. Conditional on #1, should HPMOR be featured there?

This is what I meant to focus on.

I used to think LW should "be accessible?", with a main goal of "raise the sanity waterline". Our weird ideas had to be translated into ways that the average person understood. (I thought similar things about EA).

Now I think that one-size-fits-all isn't a viable strategy. I think communities need to specialize, and I think having a creative spark that drives a given community is more important than accessibility. I don't think creative-sparks are transferrable across belief systems or aesthetics. (i.e. someone else could form a community with similar goals but with an academic aesthetic, or with a hippy-ish aesthetic, and it could work, but that spark wouldn't work for us nor ours for them)

In some ways, the Sequence are a rehash of Thinking Fast and Slow. But TF&S didn't inspire a community, or do much to change people's behavior AFAICT. People complain about the sequences being over-top-top and full of weird Eliezer-isms, but I assert that weirdness was essential to actually accomplishing something with it.

It'd be good for someone to translate our best ideas into forms that resonate with different aesthetics, but this requires a lot of skill, and is a very different task than "strip out all the weirdness". And it wouldn't be an appropriate goal for this site.

So the question reduces to:

1. What makes our community valuable?
2. What is the spark that drives that value?
3. Is HPMOR an essential part of that spark?

***

Why is our community valuable? My answer: it produces ambitious, careful, insightful thinkers, and gives them a place to refine their ideas, with some context to help translate them eventually into things-that-matter-in-the-real-world.

I think most of the value comes from a smallish number of people (power law distribution) who produce the best insights, research, or concrete-stuff-happening-in-the-real-world. Secondarily, a larger ecosystem of people that help refine ideas that direct the community's attention in useful ways.

So my core goals for the site are:

- Attract the sorts of people who either are intellectual heavy hitters who could use a vibrant ecosystem to help build their ideas or focus their attention on the most important problems
- Attract the sort of people who could become heavy hitters, or who help contribute to the facilitating environment.
- Filter things such that the people who who produce the most insights have an easy time interacting with each other, or reading ideas that funnel themselves into real-world output.

Ambition and Insight

The Sequences are fairly good at both (although the ambition isn't as apparent till about halfway through). What HPMOR is good for is painting a vivid picture of what it'd be like, to be a person who cared deeply about the world, and about thinking, and then actually trying to do stuff.

I should note: there's a major failure mode if you just rely on HPMOR for ambition: it also attracts the sort of person who likes fanfiction, who gets excited by ideas but then doesn't put in the (less viscerally exciting) work of research and thinking over the course of years.

There's a fantasy series, "The Steerswoman", which is essentially about an order of rationalist-types. Unlike HPMOR, rationality doesn't give them super powers. It just... makes them slightly better at noticing things, and integrating evidence.

This is a much more realistic take, and I think it's useful to bring people back down to Earth re: "what should I actually expect rationality to do for me, how exciting should I expect things to be, and how long does it take to get good?". But I still think having HPMOR for an initial flash of inspiration is valuable (for the people it works for).

What is the Creative Spark here?

Fully exploring this is beyond scope, but I do think an important part of the LW Spark (seeded by Eliezer) is seeing past traditional intellectual authority/hierarchy and thinking for yourself. I think the idea front-page of the site demonstrates the intellectual value here, while putting up a (slight) barrier for the sort of person who cares too much about traditional prestige.

That said, I do acknowledge it's very possible (perhaps common) to be the sort of person who sees past generic Prestige Markers, but is still allergic to fanfiction in particular.

I'm not actually attached to any of the current content (Sequences, Slatestar or HPMOR). There could be new writings that did their job better, that filtered people on the true-underlying-thing instead of the true-underlying-thing-filtered-for-some-accidental-aesthetics-of-Eliezer.

But writing newer/better content is very hard, we don't currently have it, so for now the question is "is it better to have this stuff, or not have it."

So: My Actual Cruxes

I think the cruxes you listed (i.e. how many people bounce off HPMOR, enjoy it, or are neutral?) are good, but incomplete. My version:

1. What is the ratio of "people whom HPMOR helps become more ambitious", compared to "people who would otherwise jive with the aesthetic/spark here, who might deliver a lot of value, who wouldn't deliver as much value elsewhere, but who are allergic to fanfiction or HPMOR's style and so bounce off."

If there's too much of the latter, I'd definitely change my mind.

2. Does HPMOR foster too many not-actually-ambitious fans who are more offputting than helpful to the ambitious environment I'd prefer? I think it does bring in that sort of person, but that it's a fairly achievable task of filtering things such that they don't clutter up the front page.

Having said all that: Third Options

In another thread, you convinced me that HPMOR as currently presented on the front-page is spending disproportionate weirdness points for the value it produces. I think there's probably room to either replace it with "Rationality Fiction", or something, that manages to be intriguing and let people know about it without having "fanfiction wtf?" being something people have to get past right at the very beginning.

comment by gjm · 2017-10-08T00:20:51.320Z · score: 4 (1 votes) · LW(p) · GW(p)

(I'll respond in not-specifically-double-cruxing mode first of all, and then move on to the crux-refining bit at the end. I shall note that arguably I could simply declare victory because you end up by saying that we should do something less visibly-weird than making HPMOR prominent on the front page, but arguments aren't meant to be about winning and losing around here and as you'll see when I return to the topic later I think our positions do still differ.)

OK, there's some good stuff here and we can already identify some things we (not very surprisingly) agree on.

We are agreed that neither HPMOR nor any other "legacy content" should be taking up a big chunk of the front page for established LW users. (And I think we're broadly agreed on roughly what the situation should be for those users.) So we've refined the question to: Should HPMOR be displayed prominently on the front page as it is shown to newcomers?

We are agreed that a community doesn't have to be appealing to everyone in order to be useful, and that the LW community will probably never be appealing to everyone. In particular, we are agreed that "some people will be put off by having HPMOR prominently on the front page" is not on its own a strong argument against. (I think I've made it clear that that wasn't my argument, and I think you understood that, but no harm in making it explicit again.)

We are agreed that if LW has value beyond merely being a fun place to chat (which ain't nothing, but the ambitions were always higher than that) its value has something to do with helping people to become good thinkers who are effective in getting valuable things done. I think I might give some weight to goals less ambitious than yours, though; aside from trying to produce intellectual heavy hitters who Change The World, if LW helps a bunch of people to think a bit better in ways that are somewhat useful then it is doing one of the good things it is meant to be here for. And I think it's at least possible that LW may do more good by helping lots of not-so-world-changing people a bit than by fostering world-changers. Obviously this is a side-issue, but in the context of the HPMOR-on-the-front-page question it may play the same role as a difference in moral values might in a discussion of (say) economic policy.

So, in this context you say: HPMOR is valuable not merely because it may attract people to LW or to rationality but also because it may change its readers for the better, inspiring them to combine rationality and ambition to improve the world. That's a thing I hadn't taken into account in my analysis, and in so far as it's true it does give more reason to make it prominent because now we have to consider the scenario where someone arrives at the LW front page, the presence of HPMOR isn't going to make a difference to whether they hang around, but it might affect how useful they are afterwards.

I'm skeptical about the strength of this effect -- I suspect the great majority of HPMOR readers are not inspired by it in a way that makes a substantial difference to their lives, and that those who are were probably already rationalist types before they read HPMOR -- but I'm aware that this is not based on any actual information. Is anything known about how many people who weren't already rationalists have been inspired by HPMOR to make a serious effort at being rational and changing the world, and (even harder to find out) what they have actually done as a result?

The other thing I'd like to point out about this line of argument is that on the face of it it's arguing for a proposition like "HPMOR is valuable" rather than one like "HPMOR should be on the front page". (Hence my emphasis in the paragraph above on HPMOR readers who weren't already rationalists.)

Then you move on to another couple of related arguments: that putting HPMOR proudly on centre stage is a way of saying "here, we don't care too much about traditional signals of prestige" and that while it may make some people bounce off they're mostly going to be people we don't want here anyway because they care about the wrong things. Acknowledgement of the second of those was meant to be implicit in some of what I wrote above, but going back and rereading I see that that's not very clear. Anyway, I agree that the first point has some merit but ... not very much. I mean, if someone came up to you and asked "What are the most important ideas that distinguish LW-style rationalism from other ways of thinking, and how would you convey them to casual readers?" I'm guessing that "don't pay too much attention to traditional prestige markers" wouldn't be that near the top of the list, and even if it were "put some Harry Potter fanfiction on the front page of our website" probably wouldn't be near the top of the list of the ways to say it.

Perhaps at this point we've got something like the right consequentialist question to ask, and it goes something like this. Consider all the people who come to the LW front page without already being LW users. We are interested in five subsets of them, and the effects of them of having HPMOR prominently on the front page.

  1. People who see HPMOR there and take a look, and end up being inspired to be like HJPEV in caring deeply about the world and making a serious effort to think as well as possible so as to improve the world more effectively. Benefits: these people may improve the world; they may be good people to have around LW.

  2. People who see HPMOR there and take a look, who merely enjoy it, and who decide to hang around LW as a result. Benefits: these people may also turn out to improve the world, especially if LW helps them think better; they may be good people to have around LW. Harms: more of these than of the first set may turn out not to be good people to have around LW.

  3. People who see HPMOR there and are put off because they are attached to traditional prestige markers and HPMOR trips their low-status alarms, so they don't hang around when they otherwise might have. Benefits: we lose some people who are less likely to be good thinkers. Harms: we lose the opportunity to help them become better thinkers.

  4. People who see HPMOR there and are put off because they specifically don't care for fanfiction, even though in general they're able to see past traditional prestige markers. Benefits and harms: we lose the opportunity to help these people and the opportunity for them to help us.

  5. People who see HPMOR there, think "ha ha, these people are silly", and badmouth us to others as a result. Harms: while these people were probably never going to gain anything from, or contribute anything to, LW, their badmouthing may put others off.

And then the question is: when we add up all these benefits and harms, how do they balance out? You reckon the wins outweigh the losses, pointing out 1 and 3 in particular but also acknowledging the harms from 2. I reckon the losses outweigh the wins, pointing out 3, 4, and 5 in particular.

Something like this is probably a double-crux. It's not clear to me how useful it is, though. The empirical questions (how many people are there in each group? how much more ambitious are people in group 1? etc.) seem very difficult to answer, and weighing up the various costs and benefits is hard even conditional on answers to the empirical questions (how do we compare one person who's a bit more ambitious to improve the world, with five people producing a little more low-quality chat in discussion threads?) and I worry that what we've done here is akin to the Fully General Answer to all questions about what's best: you just look at all the consequences, and see which one comes out better :-). Considering your point 1 does move my estimate of the net effect a little in the "beneficial" direction, but my previous estimate was that having HPMOR on the front page is clearly harmful on net and my current guess is that 1 isn't close to being enough to change this.

Finally, I'll note again that your "Having said all that" comment seems like it ends up quite close to my position; but I don't want to overstate the extent of our agreement. I'm not sure that a similarly-prominent "Rational Fiction" box would be an improvement on the HPMOR box; anyone actually investigating will see HPMOR as soon as they do, and the other stuff that would be at the far end of the link is probably mostly weirder than HPMOR (and mostly lower in quality, too). My guess is that it's also less likely to inspire than HPMOR is, so one of the more intriguing pro-HPMOR arguments gets weaker if what's being proposed is a "Rational Fiction" collection rather than just HPMOR.

comment by Raemon · 2017-10-08T07:29:22.430Z · score: 5 (1 votes) · LW(p) · GW(p)

Cool. I'm going to attempt to wrap up threads where possible (avoiding this turning into a many-headed-hydra). I think we're mostly in agreement about what the problem is.

Things that still stand out as worth noting:

Arguably I could simply declare victory because you end up by saying that we should do something less visibly-weird than making HPMOR prominent on the front page

I want to push back a bit on "declare victory" being a thing that's relevant here. (This is sort of semantic but I think it's actually pretty important). A key element of Productive Disagreement is shifting away from "someone gets to win" to "we get to figure out the right/true/most-useful thing."

And in this case I think we have similar enough goals that we're actually able to do that (whereas in some disagreements, you have to recurse all the way to "do we even both believe in consequentialism?" or "do we even both believe in objective measurable truth, or on what counts as evidence about that?")

I suspect the great majority of HPMOR readers are not inspired by it in a way that makes a substantial difference to their lives, and that those who are were probably already rationalist types before they read HPMOR

I'm not sure about the total numbers either, but this point is very salient to me because HPMOR radically changed my life trajectory, when several previous "why not change the world?" type people and books failed to do so. I read HPMOR before sequences and am not sure what it'd have been like if it'd been reversed, but my sense is "The Sequences are the System 2 content of LessWrong, HPMOR is the System 1 content."

I'm guessing that "don't pay too much attention to traditional prestige markers" wouldn't be that near the top of the list, and even if it were "put some Harry Potter fanfiction on the front page of our website" probably wouldn't be near the top of the list of the ways to say it.

I pretty much agree with this. Insofar as HPMOR is necessary to have easily-accessible, I think it is a solvable problem to make it look somewhat classier. (I'm not committed to replacing it with the "rationalist fiction" page, but I'll note an advantage of that is if you aren't trying to explain it in a single paragraph in a quarter-of-the-front-page, you have more room to set the context of why HPMOR exists and why to give it a second look if you have an allergic reaction to it)

Sum Consequences

(Or, converting this into an empirical question that's answerable)

I don't think either of us would consider this definitive, but I think we'd at least both consider it evidence if a LW Survey attempted to solicit questions about how big an effect size reading the Sequences, HPMOR and Slatestar have been on people's ambition, life goals, etc.

Doing this properly is tricky. The ideal version of it would be a legit randomized control trial that included people outside this community. I think that's impractical, but it should be tease out something.

comment by gjm · 2017-10-08T09:56:17.685Z · score: 4 (1 votes) · LW(p) · GW(p)

I agree that it's wrap-up time, so just a few comments on your few comments.

First, let me push back on your push-back on the "declare victory" comment. In the very same sentence as that comment I added: "but arguments aren't meant to be about winning and losing around here". Please don't try to make it look as if I don't appreciate this, when I've made it explicit that I do. Thanks.

The fact that you read HPMOR before the Sequences and found that it changed your life is very interesting, and is evidence for your inspiration theory (though obviously less evidence than it would be if someone else reported the same experience).

I agree that one advantage of putting the link to HPMOR somewhere less space-constrained is that you get to explain it better when it's first seen.

And yes, I agree that we might get useful information from an LW survey if for some reason it prioritized this. Perhaps if no one but you reported having their life changed by HPMOR you'd change your mind; perhaps if 10% of readers did I'd change mine. I think it would be really difficult to get any handle on how many people see HPMOR on the front page, decide "I want nothing to do with these people", and badmouth rationalism to their friends, from any sort of survey, but perhaps it's fair to guess that the number who overreact so dramatically won't be large.

comment by Raemon · 2017-10-09T00:02:06.336Z · score: 9 (2 votes) · LW(p) · GW(p)

> First, let me push back on your push-back on the "declare victory" comment. In the very same sentence as that comment I added: "but arguments aren't meant to be about winning and losing around here".

Yeah, I think my comment came across stronger/differently than I meant it to (and re-reading both your comment and mine I think that's a mistake on my part).

I meant something like "I see that we're both arguing in good faith and trying to do a good thing, but it feels a little sad that the 'victory' mindset from traditional debate is still lingering at all."

For comparison: there were multiple times when I wrote recent comments on Double Crux that I accidentally wrote "your opponent." In both cases, you and I generally were approaching things in the right mindset, but I think it's a good habit, when one notices creeping "opposition-mindness" to flag it and let it pass.

Rereading your comment I think that is what you were intended to do, I just didn't initially read it that way. Sorry.

Wrapping Up

So it sounded like some final things potentially worth doing are:

a) Actually put some effort into operationalizing the survey thing. (It so happens that the survey is in-the-zeitgeist right now, but it looks like this year's survey was already pretty long).

I am interested in talking to the survey-folk about doing something with this next year. It doesn't feel pressing to me to continue with this in the immediate future but seemed at least worth considering.

b) Potentially, take what we've written here and turn it into something more easily digestible (or maybe just more easily findable) as a publicly-available transcript. (Basically, I think turning our series of comments into a single top-level post would be useful. Is that something you'd be okay with and/or interested in doing?)

comment by gjm · 2017-10-09T01:19:32.237Z · score: 8 (2 votes) · LW(p) · GW(p)

Apology accepted, obviously.

I've no objection to making our comments into a top-level post. My only concern (which has nothing to do with its being our comments rather than anyone else's) is that this would fall firmly into the category of discussion of the community rather than discussion of the things the community is about, and maybe that's a thing we want less of rather than more.

comment by Raemon · 2017-10-09T02:45:31.670Z · score: 5 (1 votes) · LW(p) · GW(p)

Ah, gotcha. Maybe make it a Meta post in this case.

comment by gjm · 2017-10-09T11:07:07.597Z · score: 4 (1 votes) · LW(p) · GW(p)

Yeah, I think it should be Meta.

It occurs to me that I said something ambiguous. By "I've no objection to making our comments into a top-level post" I meant "I've no objection to our comments being made into a top-level post"; I wasn't saying anything about my willingness or unwillingness to do the work myself.

... I suppose I should answer the obvious followup question. I don't mind doing the work myself but if I do it's likely to be quite some time before I get round to it. If someone else does the work I don't mind offering constructive criticism, corrections, etc., and would probably be quicker about doing that.

comment by Raemon · 2017-10-09T16:25:48.305Z · score: 5 (1 votes) · LW(p) · GW(p)

I actually went ahead and did it last night (wording was ambiguous but seemed to imply you weren't up for it in near future anyway)

So here it is [LW · GW]. If you feel any of my commentary is misrepresenting you let me know.

comment by ZeitPolizei · 2017-10-09T10:24:48.203Z · score: 1 (1 votes) · LW(p) · GW(p)

Is anything known about how many people who weren't already rationalists have been inspired by HPMOR to make a serious effort at being rational and changing the world, and (even harder to find out) what they have actually done as a result?

I have been keeping track of which people have read at least parts of HPMOR either directly or indirectly because of my recommendation, so I think I can give at least a rough idea of what the answer may look like.

All of this is as far as I know, I haven't directly asked many of the people about this.

Including myself, I know of 14 people who read (parts of) HPMOR (excluding the people I've met through LW/EA of course). Of those:

  • 3 (including myself) I would consider actively involved in the LW community, having read a lot of the rationalist materials and going to real-life meetups

  • 3 are interested in rationality but haven't actually looked into it that much

  • 6 don't seem interested in learning more about rationality

  • 2 I don't know anything about

The first group of people is probably has the highest percentage of "people who […] have been inspired by HPMOR to make a serious effort at being rational and changing the world", but it's not inconceivable that those people also exist in the other groups. And it is only the first group, that it would be possible to capture with a LW survey.

comment by gjm · 2017-10-09T11:03:37.810Z · score: 4 (1 votes) · LW(p) · GW(p)

Thanks! Just to check I've understood, these are all people who had no previous exposure to the LW community before you said "hey, why don't you read this thing"?

Do you have any conjectures about how the results might have been different if HPMOR hadn't existed and instead you'd pointed them at the LW website, or the "Sequences", or some non-LW resource with related content (e.g., one of the various books about irrationality and cognitive biases and the like)?

comment by Raemon · 2017-10-07T20:40:37.871Z · score: 4 (1 votes) · LW(p) · GW(p)

Meta: In the process of replying to this. I noticed that I felt it a bit hard to keep everything in working memory to respond, which I was initially attributing to the site's font-size and other choices.

I tried playing around with that in Stylish to suit my own preferences. What I found was that that did help a little, but that it was still somewhat hard to parse the paragraphs here. Might be helpful in the future to break things into smaller sections (esp. for paragraphs that naturally lend themselves to it, i.e. ones with subheaders like "(a) example 1, (b) example 2", like the one beginning:

(i.e Case 2 (newcomers of various sorts): What's HPMOR-on-the-front-page going to do to these people? (a) Some people will be immediately put off by it....

comment by gjm · 2017-10-07T21:20:52.796Z · score: 7 (2 votes) · LW(p) · GW(p)

I did actually consider splitting things up into smaller paragraphs, and it's entirely possible that I made a suboptimal tradeoff between bittiness and unclarity. Part of the trouble, though, is just that there are a bunch of different people and effects and possibilities to consider, and keeping track of them all is a burden on working memory. And it's possible that my own thoughts are insufficiently organized and that if I'd been cleverer I could have been clearer :-).

comment by Raemon · 2017-10-07T23:33:45.688Z · score: 4 (1 votes) · LW(p) · GW(p)

Yeah, not confident this was worth the time to stress about. (although I do think people tend to undervalue "how much time it's worth to reduce the cost of other people reading a thing more easily")

(Note sure what you mean by "bittiness")

This is not meant to be a "you should do it this way", but FYI here is how I'd have broken up that paragraph, which I think was fairly straightforward to implement (at least as a second pass after you're done brainstorming)

Case 2 (newcomers of various sorts):

What's HPMOR-on-the-front-page going to do to these people?

(a) Some people will be immediately put off by it - That's bad if they would otherwise have become valued members of the community; it's bad if it leads them to badmouth LW or rationalists to others; it's neutral or slightly positive if they were going to ignore us anyway and just do so slightly faster, or if they would have tried to participate in the LW community but been annoying or stupid and this scares them off.

(b) Some people will be intrigued, read some or all of HPMOR, and like it. That's good if they would otherwise have bounced off; it's neutral or slightly good if they would have been valued LW contributors anyway; it's bad if they turn out to be annoying or stupid.

(c) Some people will be more or less indifferent to it. By definition it doesn't have much net impact on them.

Here, everything depends on the relative numbers of these people, and how much we'd have wanted them to stick around here.

(the bold-weight on the site isn't quite bold enough to do it's job of making the highlight-point of each paragraph stand out, which was my intent here, so I also italicized them. In any case, point being, with that structure, I can more easily re-skim it to remember each point)

comment by gjm · 2017-10-08T00:32:00.312Z · score: 4 (1 votes) · LW(p) · GW(p)

By "bittiness" I mean that the downside of splitting things up into short paragraphs, bullet points, tables, etc., is that the reading experience becomes less "smooth" and more "jerky" somehow. More an aesthetic than a practical thing, though there are practical aspects too. For instance, when you split paragraphs up more, the same material takes up more space, which means you can't see as much of it at once, which means more load on the reader's working memory. For another instance, sometimes you want structures with a couple of levels of nesting, and making that nesting visible via new paragraphs or bulleted lists may require multiple levels of indentation, which requires either that the "inner" text be very narrow and therefore hard to read or that the "outer" text be very wide and therefore hard to read.

In general, though, I'm a fan of splitting things up to make their logical structure more visible, and I think you're correct that in this case I should have listened more to that particular inner voice. (In particular, the example downsides I listed don't really apply; the fiddly bits of what I wrote weren't that long, and as your proposed rewrite shows it's possible to indicate the outer-level structure via boldfaced heading lines or similar.)

comment by Raemon · 2017-10-05T19:16:38.499Z · score: 4 (1 votes) · LW(p) · GW(p)

(adding: I will most likely respond on the weekend. Trying to err on the side of "let people know when I've ended up putting something off")

comment by gjm · 2017-10-05T19:57:25.853Z · score: 4 (1 votes) · LW(p) · GW(p)

It's good to have that information; thank you for erring on that side.

comment by panickedapricott · 2017-10-02T19:46:21.867Z · score: 4 (2 votes) · LW(p) · GW(p)

There isn't much of a difference between Frequentist statistics and Bayesian statistics.

comment by SilentCal · 2017-10-02T19:43:34.839Z · score: 3 (1 votes) · LW(p) · GW(p)
  • In economic policy, redistribution measures (e.g. UBI) are a better idea than trying to change the initial distribution (e.g. minimum wage).

  • It is not especially irrational to forego cryonics.

comment by panickedapricott · 2017-10-03T01:30:07.154Z · score: 2 (3 votes) · LW(p) · GW(p)

On your first point. If better is defined as affect on crime, dependency, poverty, and mental-illness I would expect NO to "negligible" difference between the two. It's a minor disagreement I guess.

On your second point. I feel like the answer to this question is subjective and depends largely on how much someone values the future. I'm pretty optimistic about it so I think it's worth the 0.05% chance it would give me as opposed.

comment by SilentCal · 2017-10-03T16:19:10.295Z · score: 3 (1 votes) · LW(p) · GW(p)

I don't consider the second point a disagreement, since we're both sort of ambivalent. I'm pretty sure there are people who would think I'm unambiguously wrong not to be signed up, and they're who I was looking for.

On the first point--this actually seems substantial, maybe worth pursuing. I think initial-distribution measures carry a substantial risk of backfiring and making the poor poorer, while redistribution does not--seems hard to expect the same results if this is the case. This isn't necessarily a crux for me, but I'll hear more about your position before I try to find a proper DC.

comment by panickedapricott · 2017-10-07T06:51:40.767Z · score: 2 (1 votes) · LW(p) · GW(p)

If I believed this to be true I think I would take your position. But because you would not change your mind if you believed this was false I too, do not believe this counts as the crux of our disagreement.

I'll give it a shot this time. My proposed crux is that much of what we believe about the causes of poverty (crime... ect. ) are likely false in such a way that we are completely missing something conceptual in our models (including the one you stated above) or the causes are more powerful than our greatest operational intitutions can influence. (Age, genetics, ect)