IGCs are a way of introducing Yes or No Philosophy and Critical Fallibilism. I'm posting this seeking feedback. Does this make sense to you so far? Any objections? Questions? Doubts? Ideas that are confusing?
Ideas cannot be judged in isolation. We must know an idea’s goal or purpose. What problem is it trying to solve? What is it for? And what is the context?
So we should judge IGCs: {idea, goal, context} triples.
The same idea, “run fast”, can succeed in one context (a foot race) and fail in another context (a pie eating contest). And it can succeed at one goal (win the race) but fail at another goal (lose the race to avoid attention).
Think in terms of evaluating IGCs not ideas. A core question in thinking is: Does this idea succeed at this goal in this context? If you change any one of those parts (idea, goal or context) then it’s a different question and you may get a different result.
There are patterns in IGC evaluations. Some ideas succeed at many similar goals in a wide variety of contexts. Good ideas usually succeed at broad categories of goals and are robust enough to work in a fairly big category of contexts. However, a narrow, fragile idea can be valuable sometimes. (Narrow means that the idea applies to a small range of goals, and fragile means that many small adjustments to the context would cause the idea to fail.)
There are infinitely many logically possible goals and contexts. Every idea is in infinitely many IGCs that don’t work. Every idea, no matter how good, can be misused – trying to use it for a goal it can’t accomplish or in a context where it will fail.
Whether there are some universal ideas (like arithmetic) that can work in all contexts is an open question. Regardless, all ideas fail at many goals. And there are many more ways to be wrong than right. Out of all possible IGCs, most won’t work. Totally random or arbitrary IGCs are very unlikely to work (approximately a zero percent chance of working).
Truth is IGC success – the idea works at its purpose. Falsehood or error is an IGC that won’t work. Knowledge means learning about which IGCs work, and why, and the patterns of IGC success and failure.
So far, this is not really controversial. IGCs are not a standard way of explaining these issues, but they’re reasonably compatible with many common views. Many people would be able to present their beliefs using IGC terminology without changing their beliefs. I’ve talked about IGCs because they’re more precise than most alternatives and make it easier to understand my main point.
People believe that we can evaluate both whether an idea succeeds at a goal (in a context) and how well it does. There’s binary success or failure and also degree of success. Therefore, it’s believed, we should reject ideas that will fail and then, among the many that can succeed, choose an idea that will bring a high degree of success and/or a high probability of success.
I claim that this is approach is fundamentally wrong. We can and should use only decisive, binary judgments of success or failure.
The main cause of degree evaluations of ideas is vagueness, especially vague goals.
I'll stop there for now. Please post feedback on what it says so far (rather than on e.g. me not yet explaining vague goals).
Messages (9)
I commented on this article here: http://justinmallone.com/2020/10/comments-on-elliots-article-on-igcs/
> So we should judge IGCs: {idea, goal, context} triples.
I like and agree with this.
I think we might have talked in the yes/no tutorials (12 and 13 from memory) about convergence and the way I'd thought about legislation as goal, explanation, environment, and predictions. IGCs feels like a good codification or standardisation of whatever my legislation idea was converging with.
on that note: should *predictions* be included? They're sort of implicit in the {goal, idea} pair, but can include stuff that's not otherwise explicit. predictions are an important way of judging ideas, though.
example: if I have some new theory of gravity then my goal is to explain some relevant things about the universe*, I have my explanation as the idea, and context is like the universe as a domain and all the measurements we have and theories of those measurements, etc as constraints.
*: I'm particularly being vague b/c it's quick and I don't think it's that important
if my new theory of gravity predicts weird stuff or stuff we know is wrong then that's a criticism of the IGC. arguably it's sort of superfluous to include it in the IGC. if it were in the ICG then problems might be more obvious.
> If you change any one of those parts (idea, goal or context) then it’s a different question and you may get a different result.
Thinking more about predictions: if you change the predictions it doesn't really change the IGC. It's our understanding of how the IGC is applied or our reasoning about the effects or something like that. The ICG isn't wrong, just our *judgement* of the IGC was wrong. *(Are all predictions also judgements of an IGC?)*
I've changed my mind: predictions shouldn't be included. It might make sense for legislation to add them to the IGC b/c it means there's a permanent record and it's easier to evaluate. While good IGCs might have some incorrect predictions (one WRT special relativity and spheres comes to mind), they'll be the exception not the rule. Bad IGCs will have lots of incorrect predictions, and only have correct predictions by luck or convergence (possibly due only to the IGC using good constituent ideas).
> *Truth* is IGC success – the idea works at its purpose. *Falsehood or error* is an IGC that won’t work. *Knowledge* means learning about which IGCs work, and why, and the patterns of IGC success and failure.
I really like how succinct this is and that the building blocks are really simple. Connecting *truth*, *falsehood*, and *knowledge* together feels like essence of epistemology. I imagine other epistemologies (like baysianism/inductivism or justificationism) usually end up using complex / theory-laden terms and couldn't put things this simply -- at least that's what happens when I try to describe them in 3 sentences.
> should *predictions* be included?
a prediction is a type of idea. it has a purpose. you make the prediction for a reason. as with any idea, you can sometimes think of a prediction as the main idea in an IGC, sometimes as a sub-idea (aka part) of some other idea, and sometimes as context/background.
http://justinmallone.com/2020/10/comments-on-elliots-article-on-igcs/
> Your ideas about the appropriate use of commas are useful for writing, editing and offering writing criticism, but not useful for dealing with a house fire. So you can’t judge your comma ideas in a way that applies independently of the context that they’re intended to deal with. If someone asks you for help with dealing with a fire and you say that commas are mandatory to set off a long introductory clause or phrase, and do so on the basis that that comma-related statement is a perfectly good and true idea, that would be unhelpful and unreasonable.
yes
> If you speak of an idea being good without stating a context, you have to have at least some implied context in mind (like it’s useful for solving certain problems, it meets certain criteria of elegance you have for an idea in some field, whatever). Talking in a loose manner like that is okay if you understand what’s really going on.
yes. reduce loose talking when there are problems: difficulty resolving a disagreement with someone, difficulty communicating with someone, difficulty making progress on an issue, when you suspect more precision would be valuable, etc.
this reduce-when-problems advice also applies to e.g. using positive arguments and to tons of approximations and shortcuts in general.
In the original blog post, the quoted sentence *Does this idea succeed at this goal in this context?* is in italics.
Tip: You can copy/paste into Ulysses so it preserves italics from webpages (and books). Use the "Paste From" feature and select either HTML or Rich Text (both work and will sometimes result in different formatting, though they treat italics the same).
On Mac, you can find "Paste From" via the Edit menu or right click context menu. You can also use cmd-shift-v to bring up a small UI thing that lets you select which format to paste from.
This way you'll get quote formatting right without having to manually look for and add the italics, which is a hassle.
You're also missing line breaks between paragraphs in some quotes including this one. And in this one you have a line break between sentences from the same paragraph, too!
> Some ideas seem like they’d be irrelevant for some contexts, so I’m not sure what it would mean for an idea to work in a context for which it is irrelevant.
By working in all contexts I meant an idea that gives correct answers in all contexts. That's a different issue than relevance. The issue, which is pretty tangential and unimportant here (maybe I shouldn't have mentioned), is more like e.g. that maybe "different laws of physics, logic and computation" is a context where "2+2=4" (or the way we think plus works) is wrong.
Having irrelevant stuff in the context/background term is common and harmless. The issue is more like is there some idea you can *always* include in the C part of IGC and it'll never be wrong/problematic no matter what actual context you're dealing with?
In general we could make mistakes about what's relevant so we don't exclude stuff from C just because it seems irrelevant. Instead we don't focus attention on it, but if someone else thinks it's relevant you don't debate them about whether to include it, you just say "sure it can be included if you find a way to use it". Be very inclusive about what's *allowed* in C and more picky for which aspects of C actually get your attention.
> On the one hand I think that it can be reasonable to have a range of outcomes that you treat as a broad category of “succeeding at goal”. There doesn’t seem to be anything inherently wrong with that to me.
It causes trouble in technical epistemology. It's reasonably OK (rather than a disaster) as an approximation that you back off from when there's trouble, as discussed above. But it's usually, practically, a bad idea because it ignores bottlenecks and breakpoints (which I've talked about before so you may have an idea of what I'm talking about, but I'm not going to preemptively explain here. you can ask questions though if you want).
> On the other hand, it seems like you could think of things in terms of multiple different goals
Yes, you construct many IGCs with the same I and C but different G, and you figure out which goals the I works for. This is more precise and accurate. And then you look at some other I's, for the same set of G's and the same C. And this gives you a better picture of your options. And then you pick one specific IGC to act on (which is actually usually a meta idea of some sort, which makes things more complicated).
There is more to consider. Each IGC can have unintended consequences or tradeoffs. Goals should not be considered in isolation.
> Truth is IGC success – the idea works at its purpose. Falsehood or error is an IGC that won’t work. Knowledge means learning about which IGCs work, and why, and the patterns of IGC success and failure.
A person may “run fast” to “win the race” and accomplish their goal. But as they cross the finish line they could fall down and die because they didn’t train and prepare properly. When you narrow the scope of the goal you have success by “winning the race”. When you expand the scope and the number of goals you can have failure by “winning the race and being healthy”.
There are tradeoffs. There are goals which are being ignored. And there are better ideas which account for more goals.
#4:
> http://justinmallone.com/2020/10/comments-on-elliots-article-on-igcs/
>> Your ideas about the appropriate use of commas are useful for writing, editing and offering writing criticism, but not useful for dealing with a house fire. So you can’t judge your comma ideas in a way that applies independently of the context that they’re intended to deal with. If someone asks you for help with dealing with a fire and you say that commas are mandatory to set off a long introductory clause or phrase, and do so on the basis that that comma-related statement is a perfectly good and true idea, that would be unhelpful and unreasonable.
> yes
>> If you speak of an idea being good without stating a context, you have to have at least some implied context in mind (like it’s useful for solving certain problems, it meets certain criteria of elegance you have for an idea in some field, whatever). Talking in a loose manner like that is okay if you understand what’s really going on.
> yes. reduce loose talking when there are problems: difficulty resolving a disagreement with someone, difficulty communicating with someone, difficulty making progress on an issue, when you suspect more precision would be valuable, etc.
> this reduce-when-problems advice also applies to e.g. using positive arguments and to tons of approximations and shortcuts in general.
>
> In the original blog post, the quoted sentence *Does this idea succeed at this goal in this context?* is in italics.
> Tip: You can copy/paste into Ulysses so it preserves italics from webpages (and books). Use the "Paste From" feature and select either HTML or Rich Text (both work and will sometimes result in different formatting, though they treat italics the same).
Ya I used this feature. The issue was an error in my CSS, which has been fixed.
> On Mac, you can find "Paste From" via the Edit menu or right click context menu. You can also use cmd-shift-v to bring up a small UI thing that lets you select which format to paste from.
> This way you'll get quote formatting right without having to manually look for and add the italics, which is a hassle.
> You're also missing line breaks between paragraphs in some quotes including this one. And in this one you have a line break between sentences from the same paragraph, too!
One potential cause of the line break issue:
When I do paste from HTML as opposed to just copy/paste, the breaks get changed. This is regular copy/paste:
And this is Paste from HTML:
:(
>> Some ideas seem like they’d be irrelevant for some contexts, so I’m not sure what it would mean for an idea to work in a context for which it is irrelevant.
> By working in all contexts I meant an idea that gives correct answers in all contexts. That's a different issue than relevance.
Ah okay. So e.g. addition *works* in a dealing-with-house-fire context and a cooking context and an elementary math school test context but might not be relevant in all of those contexts.
> The issue, which is pretty tangential and unimportant here (maybe I shouldn't have mentioned), is more like e.g. that maybe "different laws of physics, logic and computation" is a context where "2+2=4" (or the way we think plus works) is wrong.
Ahhh okay, yeah I wasn't really thinking about that sort of issue at all.
I forgot to sign comment #6.
#5 Any additional goals one has, which aren't in the goal term (which may be of the form "A and B and C"), are in the context term.
IGCCs? Qualitative differences between context elements
I think there might be qualitative differences between some context elements.
Certainly there's less and more relevant context in the evaluation of an Idea against (Goal,Context) pairs. Like what I had for breakfast is context to any IGC I evaluate, but it rarely matters.
But for most problems, there's some specific context that matters in a qualitatively different way to most contexts. Like for dietary things, what I had for breakfast will usually matter.
I was trying to reason about bouldering "problems" (as they're called). That is: a particular configuration of holds on a wall, usually with an intended solution. the wall can often have multiple surfaces at different angles, and they often include "volumes", which are basically extra geometric bits you screw to the wall that can be used in any problem (and make the wall more interesting). here's an image of some pyramid-y volumes.
the walls and volumes are general context that apply to any of the problems (aka routes) that run close enough. but the holds are specific context that mostly applies to the routes that they compose. (they can sometimes get in the way of other routes, and you're meant to use only the holds on the route you're doing, so holds are a sort of general context for the other routes.)
When thinking about it, I kept using (or 'reaching for') the word 'constraint'. The idea would be that a 'constraint' is context that always applies to any solution to a problem. Other context may or may not apply depending on the solution, but constraints are qualatatively different.
Though I'm using that word, it's mb a bit different to Goldratt's use in ToC. Like there are many holds for a bouldering problem, is each a constraint? if so, then it's a diff usage of the word b/c there's >2 constraints. but we could look at the set of holds as the single constraint, so maybe it's a similar enough usage to make sense.
Using the word 'constraint' was the reason for using "IGCC" in the post title. like 'idea, goal, constraint, context'.
In any case, even without using the word 'constraint', i think there is a qualitative difference between some context elements given any specific problem.