Table of Contents
In order to make progress, you must find and fix errors. A key is being able to successfully judge, for yourself, what is an error. You’re not going to be able to fix errors if you can’t find them. Finding errors is a more important issue to focus on first. You can already do this well for some topics, such as identifying common animals. You don’t mix up a dog, bird or cow. You can tell when an animal label is correct or not. You wouldn’t double or triple check that and keep being unsure or wrong.
There are more difficult cases, e.g. some small dogs look similar to some cats. Sometimes you only get a glance at the animal. Sometimes you only hear the animal but don’t see it. Sometimes you’re asked the particular species of bird but you don’t know much about birds. Reasonable people can tell when they’re unsure. You can say “I’m confident that this one is a sheep, but while I can make a guess about that other one, I’m not confident.” Knowing when you know, or not, is an important part of the skill of being able to check for errors. If it was really important to identify the exact species of bird, and you knew you were unsure, you could look it up (but being overconfident could prevent you from looking it up).
If you try to learn philosophy, you’ll probably find that it’s hard to tell what’s an error. Philosophy seems much less clear cut than arithmetic, animal species identification, or learning the names and locations of every country on a globe or map. You can write something philosophical but not know whether it’s good or bad. It’s easy to reread it and still be unsure. You can edit it but not know if you made it better or worse.
Criticism from other people is a great tool but it can’t fix the need for finding and fixing errors yourself. External feedback needs to be an extra thing to help with some stuff you miss, not the primary way you learn.
So, to learn philosophy, you need to take areas where you can find (and fix) your errors and expand on those. Build up from there. Don’t jump into stuff where you can’t effectively judge what’s correct or working.
To work on learning something, you don’t need to be perfect at identifying errors, but you do need to be fairly good at it. If you’re bad at it, then you need to find a way to develop some other skills to work your way up. If you can’t tell what’s good or bad, right or wrong, then you’ll be lost and unable to guide your learning effectively (you might learn and practice wrong ideas).
This also applies to action rather than learning. If you can’t really tell what a good outcome is or judge when things have gone wrong, then your actions may be ineffective or counter-productive and you wouldn’t know.
People often get stuck because they dismiss many errors as unimportant or unworthy of their time or attention. They claim many errors are minor and try to ignore them. They can find the error, but then they don’t care or refuse to deal with it. But people ignore some important errors. They won’t listen to some valuable criticism. They don’t find some errors on purpose because they don’t value finding those errors. They end up blocking themselves from making much progress.
People are worried about pedantic hair-splitting in critical discussions. To defend against that, they ignore lots of criticism and don’t think of lots of criticism in the first place. But lots of the criticism they ignore is important. They choose what to ignore based on their biases and intuitions, not with a good method. There is a real problem of some criticism being unimportant, but no good solution for how to decide which criticism to focus attention on.
Besides ignoring errors, people also go wrong by trying to worry about every little maybe-error in unreasonable ways. This is often called perfectionism. They count anything besides perfection as an error and it bothers them. This attitude can also block progress.
It’s hard to tell what’s important before you understand an issue well. But you have to learn about things that you don’t already understand well.
Before you know the solution to a problem/error, you don’t know what impact the solution will make. What will go wrong if you ignore the issue? What benefits will you get from this solution? Will the proposed solution even work or would you need to come up with some other solution? These are the kinds of things you usually know in retrospect, after taking action and seeing the result, not at the beginning.
What can you do about this?
There’s no perfect solution. You have to make fallible judgment calls.
The ideal thing to do is automatize dealing with easy problems. Figure out the standard patterns of error and deal with them in ways that you can reuse with little effort. Then you can deal with minor errors cheaply, and you’ll have much less reason to try to ignore errors. However, getting to the point where you have things under control like this can take a while. People often have trouble getting started on this process because they are overwhelmed by errors now, and they don’t know which to start trying to deal with and which to set aside. Bad choices about how to begin can prevent getting to a good situation where you don’t run into too many hard, time-consuming problems to deal with.
Also, there’s a common problem when people start setting some issues aside to come back to later. They find the list grows much faster than they actually go back to and address any issues they saved for later. Eventually they may realize they’ll never do most stuff that they don’t prioritize immediately. Most things being “now or never” is a bad situation to be in.
To judge which errors are important well, you have to set appropriate goals. Say you’re trying to throw a ball to someone. What exactly is the goal? A reasonable goal is throwing it within arm’s reach of your partner. You might also want to throw it somewhere he can catch it if he runs. But you could have a goal of throwing it to an exact spot in the air in front of him. You could aim for millimeter precision or even literally perfect precision. If your goal is too ambitious, you’re always going to have errors because you can never throw a ball to a perfectly exact location even once, let alone learn the skill of throwing that accurately every time.
Your goal defines which errors are important. If you have a good and clear goal, then all errors/problems – things that would result in failure at your goal – are relevant and worth working on. But if your goal is vague or too perfectionist, then you’ll have to ignore problems (or, better, adjust your goal).
You can also make a goal that’s too easy and unambitious like “I want the ball to go forward when I throw it”. Unless you’ve never thrown a ball before, that’s something you can already do, and it will give you no guidance about what improvements are worth making (since you’re already able to achieve the goal, no improvements matter according to that goal).
Perfectionist goals make all improvements matter because you can always be more perfect. Or you can be somewhat perfectionist, and overly ambitious, and set a goal which makes way too many improvements matter, at small levels of detail.
By contrast, trivial goals make no improvements matter because the goal is easy to achieve without getting better at anything. Underly ambitious goals make most improvements not matter.
Both perfectionist and trivial goals prevent progress. Neither gives you a way to judge which problems/issues are worthwhile to improve. You need to set goals which help guide you. A goal should dictate that some improvements are worthwhile but others are not part of your goal. The goal needs to help distinguish what matters and what doesn’t. If a goal says everything or nothing matters, then it isn’t doing its job. Less extreme goals that say almost everything or almost nothing matters have similar flaws.
Reasonable goals have margins for error for physical quantities like measurements of objects (like their speed, location, size, mass, or temperature). The physical world is not something we can control perfectly. Similarly, we need margins for error whenever dealing with time. You don’t know exactly how long things will take. Those are a couple broad guidelines/tips for setting reasonable goals. You probably already intuitively know those tips and many others. Not having enough margin for error can make goals too ambitious/perfectionist (and having huge margins of error can make them unambitious/trivial).
Perhaps the most important concept for judging which errors/problems are worth working on is to look at issues decisively in terms of which goals will this issue cause failure at? Here we’re talking about hypothetical or abstract goals, not necessarily your goals. They are goals you could have, but you’ll have to decide which ones to use as your own personal goals and which not to.
If you know what goals will fail due to a problem, you can consider if you actually have any of those goals. You might be fine with failing at those things while succeeding at some other goals that you do care about. That’s not failure for you since you succeeded at all the goals you aimed for.
Instead of trying to consider “How big a deal is this error?” you can change the question to “Which goals will this error cause failure at and which goals will succeed anyway?” In other words, which goals is it actually an error for? It’s not an error in terms of the goals it doesn’t cause failure at.
This requires thinking about many clear goals which specify a clear line between success and failure, rather than overly broad or ambiguous goals. This is a mindset shift away from some common thinking.
Typical thinking about goals is like “My goal is to get a good dinner.” That is a single, broad goal that doesn’t define what counts as success or failure very well. Then people try to manage their dinner-acquiring project while only considering that one goal.
Critical Fallibilism looks at many more specific goals with unambiguous criteria for success, like “Get sushi for dinner”, “Get pizza for dinner” or “Spend under $10 on dinner”. For each of those goals, you would be able to confidently judge whether you succeeded or not. You’d be able to tell if there’s an error/failure or not. Those dinner goals are about as easy to judge as “Is that a cow or a chicken?”.
You can think of many goals, evaluate which plans/solutions will succeed and fail at which goals, and decide which goals you’ll actually pursue in your life. That’s a big improvement over having one vague goal. Deciding on clear goals, and using binary goals (goals that label all outcomes as either success or failure), lets you figure out which issues/problems/errors are actually relevant and important.
A criticism is a statement that tries to point out an error. Good criticism says why an issue/problem will cause failure at one or more goals. It says which goals will fail and which won’t. A criticism should say why an idea/action/plan will fail at a specific objective/purpose/goal (or at multiple goals). And that will enable you to judge whether the criticism/error is important to you (since you know what goals it will cause you not to succeed at).
This perspective only works well if you divide big goals up into separate pieces. If you have goals like “Do X and Y and Z” and a criticism says you’ll fail at that goal, you won’t know if the failure relates to X, Y or Z. Instead, you should separate them into three goals so criticism can be more specific.
When you actually act, you’ll have multiple goals – or, equivalently, a combined goal with “and” in it multiples times. But you can think of criticism as best focused on sub-goals or minimally-sized goals. Criticism should only be aimed at a group of three things if it actually means that you won’t be able to succeed at all three together. Most criticism says an idea/plan/action fails at a specific thing, so it should be aimed at that one thing, not at a broader group. Many criticisms say that an idea/plan/action will fail at multiple specific things, in which case you can have a list of goals that the criticism applies to.
Note that criticism is often valuable despite being incomplete. People don’t have to tell us everything to be helpful. You can often figure some stuff out, and fill in some gaps, yourself. You can use a critical comment as a lead to point you in the right direction. But if you’re unable to fill in the gaps, then you shouldn’t try to act on the criticism, since either it’s mistaken or you don’t understand it.
Overreach typically involves doing stuff where you aren’t able to judge what’s an error well. That results in not fixing your errors well, which results in an error rate significantly above your error correction rate, which ruins progress. It results in e.g. making a new, second error while trying to fix an error, and then making a third error while trying to fix the second error, and then making three more errors while trying to fix the third error, and so on. A common reason people do this is because they don’t realize how many errors they’re making. They can’t tell what’s an error, so they don’t see their errors, so they think everything is going well. This actually incentivizes people to do extra hard stuff where they’re fully confused – if they did something that was more medium difficulty, they’d have more ability to spot some errors. Doing really hard stuff can also make it harder for other people to judge whether you’re doing good work or getting tons of stuff wrong.
Let’s look at a pretty typical overreach scenario.
Joe wants to discuss a difficult intellectual topic like consciousness or politics. Joe begins. Joe makes a mistake which I see. The mistake is what would normally be considered a matter of fact or logic, not opinion. It’s not a controversy; it’s not something that some experts believe; Joe just screwed up. I tell Joe.
Next, Joe wants more information. He doesn’t see the mistake right away. He has some questions. He argues back. This is OK so far. But now we’re having a second discussion. In order to make the first discussion work well, we need to fix the mistake. Now we’re having a second discussion about the mistake. In order to fix the mistake, the second discussion needs a successful outcome. (Including, possibly, that we determine Joe didn’t make a mistake and I did. If I’m mistaken, that also needs fixing. Either way, the second discussion is important to the first discussion.) Having two discussions plus a relationship between them gives us three things to keep track of, which makes things more complex and difficult.
In the second discussion, Joe makes a mistake. This creates a third discussion about the mistake in the second discussion. Now we need a successful outcome in the third discussion in order to get a successful outcome in the second discussion in order to fix the mistake in the first discussion.
In the third discussion, Joe makes a mistake.
In the fourth discussion, about the mistake in the third discussion, Joe makes two mistakes.
In the fifth and sixth discussions, Joe makes three total mistakes.
In the seventh, eighth and ninth discussions, Joe gives up and stops discussing.
And none of the discussions ever got very far. Perhaps Joe would have made 30 mistakes in the first discussion before it finished. So we might have needed 30 separate successful secondary discussions (not counting the additional sub-sub-discussions they branch to). But we got stuck on the first one so the other 29 never came up.
In this example, no mistakes were ever fixed. That’s pretty common. Sometimes a few things get fixed, but not enough to keep the number of discussions under control or even to get back to the original discussion. Or we might make it back to the first discussion, but then one or two sentences later Joe would make another mistake. Then he probably wouldn’t want to deal with a bunch more sub-discussions, so he might refuse to talk about the mistake or he might give up.
Why does this happen? Because Joe is making mistakes at a rate of around one every few paragraphs or higher. Mistake rates near one per sentence are common. Higher isn’t rare. When mistake rates are that high, sub-discussions usually have one or more mistakes before they’re resolved.
Also, note that I’m only talking about mistakes that will cause failure at a goal a discussion participant has (which we could call “decisive mistakes”), not minor or perfectionist “mistakes”.
Mistakes cause additional sub-discussions, in which more mistakes are made, which cause additional sub-discussions, in which more mistakes are made. The only way discussions are productive is if that process is kept under control and doesn’t happen too much. People need to be able to have some sub-discussions with zero mistakes so that the discussion can resolve successfully without creating more sub-discussions. The standard way people attempt to accomplish that is by ignoring mistakes, but the rational, truth-seeking way to do it is to get good enough at stuff to make fewer mistakes.
If each discussion averages 5 mistakes, and each mistake creates a sub-discussion that averages 5 mistakes, and each of those mistakes creates a sub-sub-discussion that averages 5 mistakes, then the discussion will never get anywhere.
You don’t make many (decisive) mistakes while walking. You really can learn and practice things so that your mistake rate is low. You can build up your skills. This is common with people who succeed at anything. For example, what normally happens with the people who actually get good at math? They learn counting and make mistakes initially, but with practice they get better at counting and make counting mistakes rarely. Then they learn addition, and make mistakes at first, but with practice they learn to be pretty reliable at adding. And their occasional addition mistakes don’t turn into many levels of sub-discussion. They can get it right when double checking, and they also do fine when breaking addition into parts, like counting, to discuss individually. Then they learn multiplication, then algebra, and so on. Each time they learn a new thing, they gain the ability to fairly accurately judge what is an error or not, and they get their error rate pretty low before moving on. (It’s also common to learn several things at once. You can go ahead a few steps even if you aren’t done mastering the current step. But you can’t go too far past something you aren’t done learning. This is a main reason people get stuck: they never finished learning some stuff a few steps earlier than what they’re stuck on. Being stuck often isn’t actually about the thing you’re stuck on.)
Sub-discussions about errors tend to be about prior, easier topics. If your background knowledge is solid, then you shouldn’t make a ton of mistakes in most sub-discussions. Many sub-discussion mistakes are due to missing prerequisites. And this becomes truer the more the discussions are nested. Sub-sub-sub-sub-sub discussions tend to be about simple, basic stuff. (If people were creating sub-discussions that were more complicated than the prior discussion, then complexity would escalate and there’d be no realistic hope of resolving the issues. The occasional sub/tangential/branch discussion can increase complexity, but most shouldn’t.)
When learning a new thing, you should have lots of sub-discussions that are one layer below the original topic (you make lots of mistakes), but you shouldn’t have a lot of trouble with more distant sub-discussions (you shouldn’t be making many mistakes when dealing with your mistakes). Each level – sub-discussions, sub-sub-discussions, sub-sub-sub-discussions, etc. – should have a lower error rate than the one before it. The more levels removed from the original discussion you are, the higher the chance should be that this sub-discussion is resolved quickly and successfully with no decisive mistakes to spawn more sub-discussions. If it’s not working that way, then you are going to struggle to have productive discussions that resolve anything. And note that thinking things through by yourself is the same sort of process as discussion (it involves trying to rationally address conflicts between ideas) – it’s a one-person discussion or self-discussion – so most of the same issues apply.