[Previous] Media Matters Memo: Their Plans To Control All Political Speech | Home | [Next] Objectivism and Empathy Discussion

Discussion Structure

Dagny wrote (edited slightly with permission):

I think I made a mistake in the discussion by talking about more than one thing at once. The problem with saying multiple things is he kept picking some to ignore, even when I asked him repeatedly to address them. See this comment and several comments near it, prior, where I keep asking him to address the same issue. but he wouldn't without the ultimatum that i stop replying. maybe he still won't.

if i never said more than one thing at once, it wouldn't get out of hand like this in the first place. i think.

I replied: I think the structure of conversations is a bigger contributor to the outcome than the content quality is. Maybe a lot bigger.

I followed up with many thoughts about discussion structure, spread over several posts. Here they are:


In other words, improving the conversation structure would have helped with the outcome more than improving the quality of the points you made, explanations you gave, questions you asked, etc. Improving your writing quality or having better arguments doesn't matter all that much compared to structural issues like what your goals are, what his goals are, whether you mutually try to engage in cooperative problem solving as issues come up, who follows whose lead or is there a struggle for control, what methodological rules determine which things are ignorable and which are replied to, and what are the rules for introducing new topics, dropping topics, modifying topics?


it's really hard to control discussion structure. people don't wanna talk about it and don't want you to be in control. they don't wanna just answer your questions, follow your lead, let you control discussion flow. they fight over that. they connect control over the discussion structure with being the authority – like teachers control discussions and students don't.

people often get really hostile, really fast, when it comes to structure stuff. they say you're dodging the issue. and they never have a thought-out discussion methodology to talk about, they have nothing to say. when it comes to the primary topic, they at least have fake or dumb stuff to say, they have some sorta plan or strategy or ideas (or they wouldn't be talking about). but with stuff about how to discuss, they can't discuss it, and don't want to – it leads so much more quickly and effectively to outing them as intellectual frauds. (doesn't matter if that's your intent. they are outed because you're discussing rationality more directly and they have nothing to say and won't do any of the good ideas and don't know how to do the good ideas and can't oppose them either).

sometimes people are OK with discussion methodology stuff like Paths Forward when it's just sounds-good vague general stuff, but the moment you apply it to them they feel controlled. they feel like you are telling them what to do. they feel pressured, like they have to discuss the rational way. so they rebel. even just direct questions are too controlling and higher social status, and people rebel.


some types of discussion structure. these aren’t about controlling the discussion, they are just different ways it can be organized. some are compatible with each other and some aren’t (you can have multiple from the list, but some exclude each other):

  • asking and answering direct questions
  • addressing unstated, generic questions like “thoughts on what i just said?”
  • one person questioning the other who answers vs. both people asking and answering questions vs. some ppl ignoring questions
  • arguing points back and forth
  • saying further thoughts related to what last person said (relevance levels vary, can be like really talking past each other and staying positive, or can be actual discussion)
  • pursuing a goal stated by one person
  • pursuing a goal stated by two people and mutually agreed on
  • pursuing different and unstated goals
  • 3+ person discussion
  • using quotes of the other discussion participants or not
  • using cites/links to stuff outside the discussion or not
  • long messages, short messages, or major variance in message length
  • talking about one thing at a time
  • trying to resolve issues before moving on vs. just rushing ahead into new territory while there are lots of outstanding unresolved points
  • step by step vs. chaotic
  • people keeping track of the outline or just running down rabbit holes

i’ve been noticing structure problems in discussions more in the last maybe 5 years. Paths Forward and Overreaching address them. lots of my discussions are very short b/c we get an impasse immediately b/c i try to structure the discussion and they resist.

like i ask them how they will be corrected if they’re wrong (what structural mechanisms of discussion do they use to allow error correction) and that ends the discussion.

or i ask like “if i persuade you of X, will you appreciate it and thank me?” before i argue X. i try to establish the meaning X will have in advance. why bother winning point X if they will just deny it means anything once you get there? a better way to structure discussion is to establish some stakes around X in advance, before it’s determined who is right about X.

i ask things like if they want to discuss to a conclusion, or what their goal is, and they won’t answer and it ends things fast

i ask why they’re here. or i ask if they think they know a lot or if they are trying to learn.

ppl hate all those questions so much. it really triggers the fuck out of them

they just wanna argue the topic – abortion or induction or whatever

asking if they are willing to answer questions or go step by step also pisses ppl off

asking if they will use quotes or bottom post. asking if they will switch forums. ppl very rarely venue switch. it’s really rare they will move from twitter to email, or from email to blog comments, or from blog comments to FI, etc

even asking if they want to lead the discussion and have a plan doesn’t work. it’s not just about me controlling the discussion. if i offer them control – with the caveat that they answer some basic questions about how they will use it and present some kinda halfway reasonable plan – they hate that too. cuz they don’t know how to manage the discussion and don’t want the responsibility or to be questioned about their skill or knowledge of how to do it.

structure/rules/organization for discussion suppresses ppl’s bullshit. it gives them less leeway to evade or rationalize. it makes discussion outcomes clearer. that’s why it’s so important, and so resisted.


the structure or organization of a discussion includes the rules of the game, like whether people should reply more tomorrow or whether it's just a single day affair. the rules for what people consider reasonable ways of ending a discussion are a big deal. is "i went to sleep and then chose not to think about it the next day, or the next, or the next..." a reasonable ending? should people actually make an effort to avoid that ending, e.g. by using software reminders?

should people take notes on the discussion so they remember earlier parts better? should they quote from old parts? should they review/reread old parts?

a common view of discussion is: we debate issue X. i'm on side Y, you're on side Z. and ppl only say stuff for their side. they only try to think about things in a one-sided, biased way. they fudge and round everything in their favor. e.g. if the number is 15, they will say "like 10ish" or "barely over a dozen" if a smaller number helps their side. and the other guy will call it "around 20" or "nearly 18".

a big part of structure is: do sub-plots resolve? say there's 3 things. and you are trying to do one at a time, so you pick one of the 3 and talk about that. can you expect to finish it and get back to the other 2 things, or not? is the discussion branching to new topics faster than topics are being resolved? are topics being resolved at a rate that's significantly different from zero, or is approximately nothing being resolved?

another part of structure is how references/cites/links are used. are ideas repeated or are pointers to ideas used? and do people try to make stuff that is suitable for reuse later (good enough quality, general purpose enough) or not? (a term similar to suitable for reuse is "canonical").


I already knew that structural knowledge is the majority of knowledge. Like a large software project typically has much more knowledge in the organization than the “payload” (aka denotation aka direct purpose). “refactoring" refers to changing only the structure while keeping the function/content/payload/purpose/denotation the same. refactoring is common and widely known to be important. it’s an easy way for people familiar with the field to see that significant effort goes into software knowledge structure cuz that is effort that’s pretty much only going toward structure. software design ideas like DRY and YAGNI are more about structure than content. how changeable software is is a matter of structure ... and most big software projects have a lot more effort put into changes (like bug fixes, maintenance and new features) than into initial development. so initial development should focus more effort on a good structure (to make changes easier) than on the direct content.

it does vary by software type. games are a big exception. most games they have most of their sales near release. most games aren’t updated or changed much after release. games still need pretty good structure though or it’d be too hard to fix enough the bugs during initial development to get it shippable. and they never plan the whole game from the start, they make lots of changes during development (like they try playing it and think it’s not fun enough, or find a particular part works badly, and change stuff to make it better), so structure matters. wherever you have change (including error correction), structure is a big deal. (and there’s plenty of error correction needed in all types of software dev that make substantial stuff. you can get away with very little when you write one line of low-risk code directly into a test-environment console and aren’t even going to reuse it.)

it makes sense that structure related knowledge is the majority of the issue for discussion. i figured that was true in general but hadn’t applied it enough. knowledge structure is hard to talk about b/c i don’t really have people who are competent to discuss it with me. it’s less developed and talked through than some other stuff like Paths Forward or Overreaching. and it’s less clear in my mind than YESNO.

so to make this clearer:

structure is what determines changeability. various types of change are high value in general, including especially error correction. wherever you see change, especially error correction, it will fail without structural knowledge. if it’s working ok, there’s lots of structural knowledge.

it’s like how the capacity to make progress – like being good at learning – is more important than how much you know how or how good something is now. like how a government that can correct mistakes without violence is better than one with fewer mistakes today. (in other words, the structure mistake of needing violence to correct some categories of mistake is a worse mistake than the non-structure mistake of taxing cigarettes and gas. the gas tax doesn’t make it harder to make changes and correct errors, so it’s less bad of a mistake in the long run.)


Intro to knowledge structure (2010):

http://fallibleideas.com/knowledge-structure

Original posts after DD told me about it (2003)

http://curi.us/988-structural-epistemology-introduction-part-1
http://curi.us/991-structural-epistemology-introduction-part-2

The core idea of knowledge structure is that you can do the same task/function/content in different ways. You may think it doesn’t matter as long as the result is (approximately) the same, but the structure matters hugely if you try to change it so it can do something else.

“It” can be software, an object like a hammer, ideas, or processes (like the processes factory workers use). Different software designs are easier to add features to than others. You can imagine some hammer designs being easier to convert into a shovel than others. Some ideas are easier to change than others. Or imagine two essays arguing equally effectively for the same claim, and your task is to edit them to argue for a different conclusion – the ease of that depends on the internal design of the essays. And for processes, for example the more the factory workers have each memorized a single task, and don’t understand anything, the more difficult a lot of changes will be (but not all – you could convert the factory to build something else if you came up with a way to build it with simple, memorizable steps). Also note the ease of change often depends on what you want to change to. Each design makes some sets of potential changes harder or easier.

Back to the ongoing discussion (which FYI is exploratory rather than having a clear conclusion):

“structure” is the word DD used. Is is the right word to use all the time?

Candidate words:

  • structure (DD’s word)
  • design
  • organization
  • internal design
  • internal organization
  • form
  • layout
  • style
  • plan
  • outline

I think “design” and “organization” are good words. “Form” can be good contextually.

What about words for the non-structure part?

  • denotation (DD’s word)
  • content
  • function
  • payload
  • direct purpose
  • level one purpose
  • task
  • main point
  • subject matter

The lists help clarify the meaning – all the words together are clearer than any particular one.


What does a good design offer besides being easier to change?

  • Flexibility: solves a wider range of relevant problems (without needing to change it, or with a smaller/easier change). E.g. a car that can drive in the snow or on dry roads, rather than just one or the other.

  • Easier to understand. Like computer code that’s easier to read due to being organized well.

  • Made up of somewhat independent parts (components) which you can separate and use individually (or in smaller groups than the original total thing). The parts being smaller and more independent has advantages but also often involves some downsides (like you need more connecting “glue” parts and the attachment of components is less solid).

  • Easier to reuse for another purpose. (This is related to changeability and to components. Some components can be reused without reusing others.)

  • Internal reuse (references, pointers, links) rather than new copies. (This is usually but not always better. In general, it means the knowledge is present that two instances are actually the same thing instead of separate. It means there’s knowledge of internal groupings.)

Good structures are set up to do work (in a certain somewhat generic way), and can be told what type of work, what details. Bad structures fail to differentiate what is parochial details and what is general purpose.

The more you treat something as a black box (never take it apart, never worry about the details of how it works, never repair it, just use it for its intended purpose), the less structure matters.

In general, the line between function and design is approximate. What about the time it takes to work, or the energy use, or the amount of waste heat? What are those? You can do the same task (same function) in different ways, which is the core idea of different structures, and get different results for time, energy and heat use. They could be considered to be related to design efficiency. But they could also be seen as part of the task: having to wait too long, or use too much energy, could defeat the purpose of the task. There are functionality requirements in these areas or else it would be considered not to work. People don’t want a car that overheats – that would fail to address the primary problem of getting them from place to place. It affects whether they arrive at their destination at all, not just how the car is organized.

(This reminds me of computer security. Sometimes you can beat security mechanisms by looking at timing. Like imagine a password checking function that checks each letter of the password one by one and stops and rejects the password if a letter is wrong. That will run more slowly based on getting more letters correct at the start. So you can guess the password one letter at a time and find out when you have it right, rather than needing to guess the whole thing at once. This makes it much easier to figure out the password. Measuring power usage or waste heat could work too if you measured precisely enough or the difference in what the computer does varied a large enough amount internally. And note it’s actually really hard to make the computer take exactly the same amount of time, and use exactly the same amount of power, in different cases that have the same output like “bad password”.)

Form and function are related. Sometimes it’s useful to mentally separate them but sometimes it’s not helpful. When you refactor computer code, that’s about as close to purely changing the form as it gets. The point of refactoring is to reorganize things while making sure it still does the same thing as before. But refactoring sometimes makes code run faster, and sometimes that’s a big deal to functionality – e.g. it could increase the frame rate of a game from non-playable to playable.

Some designs actively resist change. E.g. imagine something with an internal robot that goes around repairing any damage (and its programmed to see any deviation or difference as damage – it tries to reverse all change). The human body is kind of like this. It has white blood cells and many other internal repair/defense mechanisms that (imperfectly) prevent various kinds of changes and repair various damage. And a metal hammer resists being changed into a screwdriver; you’d need some powerful tools to reshape it.


The core idea of knowledge structure is that you can do the same task/function/content in different ways. You may think it doesn’t matter as long as the result is (approximately) the same, but the structure matters hugely if you try to change it so it can do something else.

Sometimes programmers make a complicated design in anticipation of possible future changes that never happen (instead it's either no changes, other changes, or just replaced entirely without any reuse).

It's hard to predict in advance which changes will be useful to make. And designs aren't just "better at any and all changes" vs. "worse at any and all changes". Different designs make different categories of changes harder or easier.

So how do you know which structure is good? Rules of thumb from past work, by many people, doing similar kinds of things? Is the software problem – which is well known – just some bad rules of thumb (that have already been identified as bad by the better programmers)?

  • Made up of somewhat independent parts (components) which you can separate and use individually (or in smaller groups than the original total thing). The parts being smaller and more independent has advantages but also often involves some downsides (like you need more connecting “glue” parts and the attachment of components is less solid).

this is related to the desire for FI emails to be self-contained (have some independence/autonomy). this isn't threatened by links/cites cuz those are a loose coupling, a loose way to connect to something else.

  • Easier to reuse for another purpose. (This is related to changeability and to components. Some components can be reused without reusing others.)

but, as above, there are different ways to reuse something and you don't just optimize all of them at once. you need some way to judge what types of reuse are valuable, which partly seems to depend on having partial foresight about the future.

The more you treat something as a black box (never take it apart, never worry about the details of how it works, never repair it, just use it for its intended purpose), the less structure matters.

sometimes the customer treats something as a black box, but the design still matters a lot for:

  • warranty repairs (made by the company, not by the customer)
  • creating the next generation production
  • fixing problems during development of the thing
  • the ability to pivot into other product lines (additionally, or instead of the current one) and reuse some stuff (be it manufacturing processes, components from this product, whatever)
  • if it's made out of components which can be produced independently and are useful in many products, then you have the option to buy these "commodity parts" instead of making your own, or you can sell your surplus parts (e.g. if your factory manager finds a way to be more efficient at making a particular part, then you can either just not produce your new max capacity, or you could sell them if they are useful components to others. or you could use the extra parts in a new product. the point was you can end up with extra capacity to make a part even if you didn't initially design your factory that way.)

In general, the line between function and design is approximate.

like the line between object-discussion and meta-discussion is approximate.

as discussion structure is crucial (whether you talk about it or not), most stuff has more meta-knowledge than object-knowledge. here's an example:

you want to run a small script on your web server. do you just write it and upload? or do you hook it into existing reusable infrastructure to get automatic error emails, process monitoring that'll restart the script if it's not running, automatic deploys of updates, etc?

you hook it into the infrastructure. and that infrastructure has more knowledge in it than the script.

when proceeding wisely, it's rare to create a ton of topic-specific knowledge without the project also using general purpose infrastructure stuff.

Form and function are related.

A lot of the difference between a smartphone and a computer is the shape/size/weight. That makes them fit different use cases. An iPhone and iPad are even more similar, besides size, and it affects what they're used for significantly. And you couldn't just put them in an arbitrary form factor and get the same practical functionality from them.

Discussion and meta-discussion are related too. No one ever entirely skips/omits meta discussion issues. People consider things like: what statements would the other guy consent to hear and what would be unwanted? People have an understanding of that and then don't send porn pics in the middle of a discussion about astronomy. You might complain "but that would be off-topic". But understanding what the topic is, and what would be on-topic or off-topic is knowledge about the discussion, rather than directly being part of the topical discussion. "porn is off topic" is not a statement about astronomy – it is itself meta discussion which is arguably off topic. you need some knowledge about the discussion in order to deal with the discussion reasonably well.

Some designs actively resist change.

memes resist change too. rational and static memes both resist change, but in different ways. one resists change without reasons/arguments, the other resists almost all change.


Discussion and meta-discussion are related too.

Example:

House of Sunny podcast. This episode was recommended for Trump and Putin info at http://curi.us/2041-discussion#c10336

https://youtu.be/Id2ZH_DstyY

  • starts with music
  • then radio announcer voice
  • voice says various introductory stuff. it’s not just “This is the house of Sunny podcast.” It says some fluff with social connotations about the show style, and gives a quick bio of the host (“comedian and YouTuber”)
  • frames the purpose of the upcoming discussion: “Wanna know what Sunny and her friends are thinking about this week?”
  • tries to establish Sunny as a high status person who is worthy of an introduction that repeats her name like 4 times (as if her name matters)
  • applause track
  • Sunny introduces herself, repeating lots of what the intro just said
  • Sunny uses a socially popular speaking voice with connotations of: young, pretty, white, adult, female. Hearing how she speaks, for a few seconds, is part of the introduction. It’s information, and that information is not about Trump and Putin.
  • actual content starts 37 seconds in

This is all meta so far. It’s not the information the show is about (Trump and Putin politics discussion). It’s about the show. It’s telling you what kind of show it’s going to be, and who the host is. That’s just like discussing what kind of discussion you will have and the background of a participant.

The intro also links the show to a reusable show structure that most listeners are familiar with. People now know what type of show it is, and what to expect. I didn’t listen to much of the episode, but for the next few minutes the show does live up to genre expectations.

I consider the intro long, heavy-handed and blatant. But most people are slower and blinder, so maybe it’s OK. I dislike most show intros. Offhand I only remember liking one on YouTube – and he stopped because more fans disliked it than liked it. It’s 15 seconds and I didn’t think it had good info.

KINGmykl intro: https://www.youtube.com/watch?v=TrN5Spr1Q4A

One thing I notice, compared to the Sunny intro, is it doesn’t pretend to have good info. It doesn’t introduce mykl, the show, or the video. (He introduces his videos non-generically after the intro. He routinely asks how your day is going, says his is going great, and quickly outlines the main things that will be in the video cuz there’s frequently multiple separate topics in one video. Telling you the outline of the upcoming discussion is an example of useful meta discussion.)

The Sunny intro is so utterly generic I found it boring the first time I heard it. I’ve heard approximately the same thing before from other shows! I saw the mykl intro dozens of times, and sure I skipped it sometimes but not every time, and I remember it positively. It’s more unique, and I don’t understand it as well (it has some meaning, but the meaning is less clear than in the Sunny intro.) I also found the Sunny intro to scream “me too, I’m trying hard to fit in and do this how you’re supposed to” and the mykl intro doesn’t have that vibe to me. (I could pretty easily be wrong though, maybe they both have a fake, tryhard social climber vibe in different ways. Maybe i’m just not familiar enough with other videos similar to mykl’s and that’s why I don’t notice. I’ve watched lots of gaming video content, but a lot of that was on Twitch so it didn’t have a YouTube intro. I have seen plenty of super bland gamer intros. mykl used to script his videos and he recently did a review of an old video. He pointed out ways he was trying to present himself as knowing what he’s talking about, and found it cringey now. He mentioned he stopped scripting videos a while ago.)

Example 2: Chef Heidi Teaches Hoonmaru to Cook Korean Short Rib

https://www.youtube.com/watch?v=EwosbeZSSvY

  • music
  • philly fusion overwatch league team intro (FYI hoonmaru is a fusion twitch streamer, not a pro player)
  • slow mo arrival
  • hoonmaru introducing what’s going on (i think he lied when he said that he thought of this activity)
  • hoonmaru talking about his lack of cooking experience
  • hoonmaru says he’ll answer fan questions while cooking
  • says “let’s get started”
  • music and scene change
  • starts introducing the new seen by showing you visuals of hoonmaru in an apron
  • now we see Chef Heidi and she does intro stuff, asks if he’s ready to cook, then says what they’ll be doing.

The last three are things after “let’s get started” that still aren’t cooking. Cooking finally starts at 48s in. But after a couple seconds of cooking visuals, hoonmaru answers an offtopic fan question before finally getting some cooking instruction. Then a few seconds later hoonmaru is neglecting his cooking, and Heidi fixes it while he answers more questions. Then hoonmaru says he thinks the food looks great so far but that he didn’t do much. This is not a real cooking lesson, it’s just showing off Heidi’s cooking for the team and entertaining hoonmaru fans with his answers to questions that aren’t really related to overwatch skill.

Tons of effort goes into setting up the video. It’s under 6 minutes and spent 13.5% on the intro. I skipped ahead and they also spend 16 seconds (4.5%) on the ending, for a total of 18% on intro and ending. And there’s also structural stuff in the middle, like saying now they will go cook the veggies while the meat is cooking – that isn’t cooking itself, it’s structuring the video and activities into defined parts to help people understand the content. And they asked hoonmaru what he thought of the meat on the grill (looks good... what a generic question and answer) which was ending content for that section of the video.

off topic, Heidi blatantly treats hoonmaru like a kid. at 4:45 she’s making a dinner plate combining the foods. then she asks if he will make it, and he takes that as an order (but he hadn’t realized in advance he’d be doing it, he just does whatever he’s told without thinking ahead). and then the part that especially treats him like a kid is she says she’s taking away the plate she made so he can’t copy it, he has to try to get the right answer (her answer) on his own, she’s treating it like a school test. then a little later he’s saying his plating sucks and she says “you did a great job, it’s not quite restaurant”. there’s so much disgusting social from both of them.


Elliot Temple on August 29, 2018

Comments

What do you think?

(This is a free speech zone!)