I wrote a dense paragraph about Critical Fallibilism (CF), particularly ideas inspired by Eli Goldratt's Theory of Constraints (TOC). This article breaks it down and explains it. Here's the paragraph with small changes from the original:
Whatâs it saying?
Overall I think the paragraph is saying: People want to solve their big problems using big solutions, but really theyâre making things more complicated than they need to be.
What do the details mean?
People want to solve their big problems using big solutions. One way that people try to solve their big problems is by weighing and adding 50 factors that are qualitatively different.
I think
means to come up with a conclusion or make a decision.
The âThenâ in the beginning means when people want complex solutions they then optimize a bunch of local optimaâŚ
The âthey end upâ is what happens after people want complex solutions
the âfactoringâ part i think means like trying to think of whatâs important for the conclusion or decision
âshouldnâtâ means donât count on the result changing. Like it can but logically itâs not going to.
âlocal optimaâ means stuff that doesnât happen in a wider scale. Like if you were to change it to help you with your goal, it wouldnât help you get good results
âexcess capacityâ means when you have enough of one factor that getting more wonât help you achieve your goal or thing youâre trying to do using the factor.
âThey donât knowâ means they havenât figured out how excess capacity works and applies to their decision making
âto always updateâ means to change the credence based on some new information or new evidence.
You can quickly check if a few factors are important to consider changing, but besides that you really only need to focus on a few factors for your solution.
Imma write more and answer more of the questions later
Local optima can be seen on graphs like this:
Every peak besides the big one is a local optima. Itâs a point where moving in any direction (left or right) gives you a worse amount. You can only go down if you make a small change. But itâs not a global optima b/c a different point is way better (the big peak), so itâs possible to move along the graph and come out ahead, but you have to go a decent distance to do that (leave the local area).
Another way to think of it is climbing a mountain. IIRC Dawkins gave this example in Climbing Mount Improbable. Mountains have places where you arenât at the very top but any direction you go is down. Those are local optima. The only way to get higher up the mountain is to go down the mountain some first. This is relevant to how evolution works (animal genes can get stuck at local optima because if they go down at all their survival and replication chances are worse).
Every other peak being local optima is a relative thing it seems. Like if the big peak wasnât there then things would be different.
The big peak is different than the small peaks cuz you lose or gain less with a small peak.
Every small peak is local optima and âItâsâ is referring to local optima so if youâre at the peak at the local optima and you move left or right it gives you a worse amount. âworse amountâ means less than if you would go left or right on the big peak.
I donât know what âgo downâ means Im thinking itâs going down from a small peak or maybe any peak. Cuz if you move in any direction of a peak you go down and you can only go down. I donât know what âmake a small changeâ means but I think itâs talking about moving from a peak.
I think the sentence is talking about us moving around in the graph. Itâs kind of hard to see the application of the ideas in irl stuff cuz I donât think i practiced small peaks and big peaks before.
âway betterâ i think could be like 5x better. Like wouldnât one want that for their goal? Like they want the big solution for their problem but if some of their solution is making that big peak wouldnât they want to focus on that instead of the small peaks of their solution?
When you do come out ahead, it seems like you gotta like measure it in a way so you donât get lost in the small stuff, cuz like a person can make a lot of changes to solve their problem but they need a good way to find which is getting them to the high peak.
leaving the local area is the donât focus on local optima. It seems like a skill to see optima for their right level
oh yeah those are local optima then.
Im thinking thatâs what
from the last paragraph means cuz you have to go down first before you go up again.
Thatâs interesting how you can get stuck at local optima. Before you reach the big peak you can get stuck in the small stuff.
A small change means moving a small amount left or right. When you change the horizontal position (x value) a small amount, while following the graph line and starting from a peak, then the vertical position (y value) goes down.
Ok, a small change is you moving a little to the left or a little to the right on the graph.
ok when you move a small amount on the graph the x-value changes. And when you do that off a peak the y-value will always go down.
That means that the inverted looking peaks on both sides of the graphs arenât peaks cuz the y-value goes up on those when you move. Theyâre called something else.
What background knowledge is it relying on?
I donât know for sure but âcomplex solutionsâ doesnât sound too uncommon but is similar to something FI would say and talk about.
âcorrectly weighting and adding 50 factorsâ is a topic about weighting to make decisions and is something I think Yes or No philsophy and CF addresses.
I donât really see(not that I read a lot) qualities being hard to add as points on anything on the internet and as a point for a major idea. Y/N talks about problems with that I think.
Things being further updated by any piece of evidence, reminds me of BOI. Like, talking about an idea or theory being supported by any piece of evidence means the idea is wrong or bad
Local optima and excess capacity comes from Theory of constraints and CF. I donât know about excess capacity and local optima being applied to factors that would change the result is something from TOC. Sounds to me more like something CF would say.
any good or bad evidence updating an idea or credence sounds to me like something BOI or FI would say. And the excess part is TOC, but I think CF is the one that applies it to ideas not TOC.
Background knowledge of âquickly checking that many factors donât ruin an optionâ maybe comes from Y/N but for sure CF. Itâs the topic of breakpoints and I think breaking the analog down to digital and dealing with discrete stuff.
â(giving pass/fail grades to things that are not near the borderline between pass and fail)," & âyou should focus your more detailed attention on just a few important factors, which is simpler than trying to use 50 factors.â is from CF. Maybe from FI.
Iâm unsure of a lot of the background knowledge cuz 1 i dont use the words background knowledge a whole lot so I donât know exactly what they mean and 2 have to check again what philosophy or school of thought says what.
What questions do you have?
Is there more kinds of complex solutions? Is a complex solution in hero shooters: trying to get to top 500 and trying to improve every little thing you see(e.g. something from positioning, or aim, or resource management)?
Is a complex solution in doing dark souls challenges trying to improve at one boss attack at a time and letting your subconscious deal with the rest?
Are all complex solutions bad? Like maybe some are better than others? Or really thatâs missing the point cuz weighing 25 different factors instead of 50 is still trying to focus on too many details instead of a few?
Is the point of bringing up qualitatively different factors to show how difficult complex solutions can become? Like instead of it being 2 times as hard(cuz weighting and adding 50 factors) itâs really way harder like 98 times as hard? I was thinking you would have to find a way to convert each qualitatively different thing 49 times and that itself is 49 different things you gotta know to do.
Is CF like the only philosophy that takes into account optima and excess capacity? LIke are they concepts that are not practiced much in the world and difficult to understand? They seem really important to consider when solving problems.
Is excess capacity related to the graph? Like letâs say a factor has excess capacity would it even form a peak on the graph if you changed it? Would it just not move the y position at all? Would it make one of the small peaks? Since the small peaks are local optima and local optima are factored in then when they are changed the graph will just show a small peak.
Does the local optima even have to be changed? or is it when you factor them in is when you get the small peak on the graph?
letâs say you keep factoring in local optima would the graph just keep making small peaks? Is that why itâs called excess capacity? Youâre just not coming out ahead like you would if you factored in something else, a global optima.
Local optima is a common, well known idea. I donât know where it originates from.
FI would criticize it, not say it. I donât think BOI talks about it.
Oh ok im looking it up:
I got a response from gemini to âWhere did the phrase local opitma originate?â:
While searching for this i learned that optima is plural for optimum and that local optima can still be treated as solutions. Like, I thought since they didnât get the same effect as global optima that you would have to discard them completely.
Yeah, I didnât write that part right. I think FI would criticize it too.
I think the topic of updating credences on any good or bad evidence is the topic of weighting truths and I donât think BOI talks about that.
Which parts could you explain to someone else, and which couldnât you?
I wouldnât be able to explain the details of this sentence to someone. For example, I wouldnât be able to to give someone a picture of what a complex solution looks like. It would be hard to explain to someone what weighting and adding 50 different factors looks like. It would be hard to say that thatâs an example of a complex solution.
I could tell them what a quality is like the color purple or a sweet taste. I donât know if this works: people want to solve their problems in a complex way. One way is they try to do 50 different things at once but fail. They end up figuring an answer, but that answer is not so good cuz they have to keep going back to fix it.
Have you seen examples of reviewing cars, colleges, video games or other products with large lists of factors, which are then weighted and combined for an overall rating?
For video games yes. The other stuff not as often. Im ngl the factors not intuitive to recall. Like, video game A received a 9.5 rating based on level design, soundtrack, graphics, story, etc.
Which parts could you explain to someone else, and which couldnât you?
The complex solution of weighting and adding 50 factors I could explain to people like: A complex solution can be when someone rates a video game by using many factors like artstyle, graphics, level design, performance, story, etc.
I could probably say that people use the complex solution to find a thing. That thing is the credence but I donât use that word often so i donât really get it. I looked it up and ox languages said itâs a belief or accepted truth.
If credence is an accepted truth then does that mean when the person uses the complex solution and finds an answer, do they accept it as true after?
When the person wants to use a complex solution like weighting and adding all the video game factors, it makes them take things into account that shouldnât impact the overall score that much.
Why it shouldnât impact the overall score much idk maybe like someone focuses too much on graphics on a souls game that itâs missing the point. Like the fun isnât so much about looking at high quality graphics. Idk
Itâs interesting you say if you write or speak your thoughts out, youâll learn more effectively from this article. When I think about efficiency I think about taking shortcuts and using less resources than you need to so you can still achieve something. Is it more effective cuz youâll already have stuff written out and you can compare with the article? That way you can start with something rather than having nothing written down
Credence (statistics) - Wikipedia
Effectiveness and efficiency are different things.
Basically effectiveness is like benefit and efficiency is like cost/benefit ratio.
I read that article n i kind of learned that credence is based on probability. One can say that they are 50% credent that something is gonna happen. An example is: based on what i know n feel i am 50% credent that my boss wonât show up for work today.
I also learned a bit what a fractional odd is. Like if the prize pool for guessing my boss will not show up is $100 and I stake 50 dollars my fractional odds i think are 1/1.
That makes sense i didnât know those two words were different. If effectiveness is like benefit then i think itâs like success and like gain too. Itâs like success cuz if you learn more effectively then you might be successful at learning x, y, and z. Like maybe those werenât your direct goals but thatâs what you gained. You benefited from it
I would not have thought these words were next after people want complex solutions. At least not the silver bullet part. Like I didnât think talking about what TOC advocates is a good response or leeway for an analysis.
I think you probably talk about what TOC advocates cuz you have a solution to propose.
To me the response in the quote above is like inputting something in the function f(x) and you get a number I wouldnât have gotten on my own.
Key issues is a common word people use when talking about problems. Is it easier to think of constraints as key issues? Key issues sounds easier than just thinking about issues cuz sometimes thereâs lots of issues to look at.
I know some changes that will have negative large downstream effects. Sometimes I try to think of them. Like when trying to beat a boss in a souls game whatâll have a large downstream effect is you put the controller down and stop using it. Youâll die. Or maybe every time you want to press an input you wait a sec before you do it. Itâll affect how you walk to the boss, how you hit the boss, and how you dodge the boss.
What matters most is something people can find sometimes I think. Itâs a commonly used word. Are people usually good at finding what matters most? If you find what matters most is that where youâll find global optima?
High impact changes sounds to me like getting results. Like thereâs better ways to change something than others.
Do people usually make small optimizations? or do overly complex ones? If yeah itâs so hard to tell whatâs small or big. Overly complex ones sounds like a lot of work, but I bet it happens a lot.
Itâs like if I were to make a building with many floors. The chances of consecutive floors going wrong if i somehow make the first one right gets worse.
Isnât it sometimes you have to do the complex changes how else are you going to get something done? Are we talking about the same thing?
I notice you say man AND/OR complex changes. You can make complex changes but you can also make many changes and many complex changes.
That would make sense to me cuz youâre spreading yourself thin among doing many things.
I havent read the choice yet only a little bit of the goal. Inherent simplicity sounds important for understanding how much work you should do so itâs not extra
Arrogance? I usually think about arrogance as a gotcha from people. Looking it up i found: ox languages: âthe quality of having an exaggerated sense of oneâs own importance or abilities.â I think in the block quote it means overestimating oneâs abilities.
So complex solutions doesnât go well and we canât control so many factors very well. Complex solutions are when we try to control so many factors. Controlling that many factors doesnât go very well cuz we canât. Like our ability to do it is not there i think.
âindividuallyâ sounds important too like it sounds like we canât take so many things on without working in chunks.
This is relatable. There is so much sometimes to work on that it becomes too hard. Like thereâs gotta be an easier way instead of just focusing on a 100 hundred to get good at.
Oh I didnât think about these being what factors are. It didnât come to mind to define what a factor was. Factors are like qualities
Factors look like things you can measure like with a ruler. It seems important to factor in things you care about. Like itâs not intuitive to me to pick factors that are tailored for you and your goals.
Naming examples of factors isnât that intuitive for me, but I wanna try:
Stuff I care about: Price, Taste, how sweet something is, how long will it take to do an activity, what is my accuracy in an fps, how fast is the strafing in a shooter, what is the serving size of a meal, how tall the roof is to my new home, how much storage a new mac has, how many hz a monitor have, how long does a product lasts, how many reports of a faulty product are there, how many uses does the analog stick of a controller have, how many calories does a food have, what use does an app have.
Donât you have to make sure you made your research every time to factor in something? Like canât it go really wrong if u donât get the nuances? It sounds like a lot of work.
I want to try to think of an example where you can think of one factor split it into multiple for a category:
Mini-Post mortem:
Error: I thought I could name mini-factors for every factor.
How did that error happen? Well first I thought about âwhat is a category I can name as a factorâ. Then I thought a product like Mac will do. Then I thought what are some factors in buying a mac. I wrote storage because I thought there were factors you can break apart, like how good the storage is and what parts make a storage good. I know you can name something you care about like how much storage does a pro mac have vs a regular mac but still.
I thought you can make price into multiple factors similar to how you can break down car safety into multiple kinds of safeties like in the quote above.
The error happened similarly to the other main factors without subtrees. I thought I could use the skill of breaking down any factor into multiples of that category(like breaking down the safeties). I donât think it was that bad an error but I wanted to use it to practice post mortem. I hope itâs ok to do postmortems like this in this post.
Goldratt says lots of companies do lots of initiatives that, if they work correctly, will only provide a small benefit like 1%. Some of them are complex too.
Maybe but often a simple change works. A car is a complex system. When a car breaks, one simple change will often fix it. You need to understand the system, understand the problem, and figure out the right change.
If the car breaks and you start optimizing random stuff, you probably wonât fix it. You could give it new paint to help prevent rust. You could clean all the trash out of it to reduce excess weight to make it go slightly faster. You could change the oil, get premium gas, get brighter headlights, have a better radio put in, and have a roof rack added. Those all make the car âbetterâ but theyâre probably irrelevant to why it broke.
This example gets into breakpoints. I think 120hz is good enough. I like more hz (120 is better than 60) but once it reaches 120hz I donât really care about more.
This is another case where the factor generally sounds like a good thing but more isnât necessarily better. If a home is too tall, then itâs more vulnerable to wind or earthquakes and maintenance costs are too high.
Thatâs doable.
For Mac storage, here are some points you could put nested under it:
For Mac applications:
Other ways to break down price:
