Critical Fallibilism Introduction

This is a companion discussion topic for the original entry at
1 Like


Critical Fallibilism is the philosophy I developed over the last 20 years. It’s about rational thinking and learning. We’ll call it CF for short.

This is an essay summary video. The essay is on the right for reference and linked below.

I named CF similarly to my biggest inspiration, Critical Rationalism by Karl Popper. I’m building on Popper’s themes like fallibilism, evolutionary knowledge creation, and error correction.

I also found useful ideas in Eli Goldratt’s Theory of Constraints. Goldratt emphasized achieving goals by figuring out the right places to focus our limited attention where we can achieve big wins. That means finding and optimizing constraints, not optimizing factors with excess capacity.

Ayn Rand’s Objectivism helped me understand how learning works. Rand explained integration and automatization. To summarize her conclusions, we build up knowledge in layers from small parts, and we need to practice.

It’s important to improve on existing knowledge, not start from zero. Those three thinkers are my biggest inspirations, especially Popper.

Now let’s talk about CF’s most distinctive idea.

CF says to use binary evaluations of ideas as refuted or non-refuted instead of evaluating ideas by degrees or amounts of anything, like goodness, credence or justification. That means there are no strong or weak arguments. For now, I’ll focus on explaining how this works, not why it’s awesome nor what’s wrong with alternatives.

Ideas are refuted if we accept a decisive criticism of them. Decisive criticisms contradict their target ideas, so the criticism and the idea can’t both be right. They don’t merely try to weaken the idea.

Ideas have purposes or goals which they either succeed or fail at. Decisive refutations explain that an idea fails at a goal.

We should never act on or believe an idea, in order to achieve a goal, if we have a decisive criticism of the idea for that goal. In other words, never use ideas that you don’t think will work for what you’re trying to use them for.

How can a binary system offer nuance? We’re basically scoring ideas as 1 or 0. What if two ideas are good but unequal? 0 is for failure, so both ideas score a 1. That’s a tie even though one idea is better.

CF’s solution is using multiple evaluations for ideas. Instead of trying to give an idea one complex evaluation, it can have multiple simpler evaluations. All the binary evaluations together offer complexity and tie-breaking.

An idea can succeed at one goal but fail at another. Instead of evaluating ideas, we should evaluate pairings of ideas with goals. We can also add the context.

Eating pizza works for satisfying hunger, getting enjoyment, and providing your mitochondria with energy. It fails at curing cancer, cleaning your room, or causing you to win the lottery.

To say one idea is better than another, you need to find a goal that it succeeds at which the other fails at. For example, eating pizza and eating plain, leftover rice would both satisfy your hunger, but the pizza would be enjoyable and the rice wouldn’t.

You could make up hundreds of arbitrary goals, so make sure to use relevant goals that you actually care about. Failing at a goal only matters if someone actually has that goal.

CF uses binary goals. Goals define success and failure with no concept of partial success. This works like binary logic, which has true and false, but no three-quarters-true.

Trying to use degrees of success leads to maximization goals. More success is better, so it’s implied that you should use the idea that will give you the most success, and using any other idea would be an error. This gives a binary distinction between best and not-best, so it doesn’t offer a non-binary alternative to CF.

Also, maximization goals are rarely a good idea. In general, our goal should be about good enough, not about maximization.

Most factors we consider have excess capacity. They’re already good enough, so a little more or a little less isn’t important.

To come up with binary goals, try asking yes-or-no questions about your topic.

Breakpoints are points on a spectrum where there’s a qualitative difference. In other words, they’re amounts that make the difference between success and failure at some goal.

Changes in quantities that don’t change any goal from failure to success, or from success to failure, aren’t important. Those are changes in excess capacity which don’t cross a breakpoint.

Whether a breakpoint is crossed is a binary issue. Either it’s crossed or not. So breakpoints let us deal with quantities in a binary way. Breakpoints also help us focus our attention on important, qualitative differences.

CF has some other ideas, like about how to debate rationally or learn effectively, but the thinking method using binary evaluations is the most original and I consider it the most important. CF also has some arguments that basically say non-binary methods run into logical problems and can’t work.

The full essay is linked below, and I have more free essays at