Evolution and Epistemology

How evolution relates to intelligent thinking and error correction

Table of Contents

How can knowledge be created from non-knowledge? Where can knowledge come from? This is the problem of design. The only known answer is evolution.

A wolf, or a wolf’s eye, appears to be designed. This “appearance of design” – of purpose – is knowledge. It looks like it was designed because it’s not random or purposeless. It’s good at things like hunting or seeing.

Evolution creates knowledge. Knowledge is information which is adapted to a purpose. A purpose is a goal or objective – some kind of design. Evolution is not an intelligent designer, but it creates things that look designed by adapting information to purposes. Evolution does a kind of problem solving.

Human thinking creates new knowledge somehow. Modern science or philosophy wasn’t contained in our genes. A wristwatch or smartphone wasn’t designed by biological evolution. But it has the appearance of design. So where could it come from? How does human thinking work?

Since no alternatives are known, we should look at the one answer we do have (which has no known refutations): humans create new knowledge (design, adaptation to a purpose) by evolution – by evolving ideas inside their heads. So there’s a second kind of evolution. It’s evolution of memes rather than genes. (By “meme” I mean the original more technical definition – any idea which is a replicator – not a mind virus or internet joke.)

Note that inventing alternatives to evolution is really hard. People have tried a lot but have failed. Intelligent design isn’t a viable alternative because it assumes intelligence is present, and does the designing, but it doesn’t explain how intelligence functions. Creationism has the same issue – it basically says complexity and knowledge were created by a creator who already had complexity and knowledge … but then where did that creator come from? Again there’s a major gap where something isn’t explained. Spontaneous generation says maybe stuff just appears out of nothing for no reason, which is about as unsatisfactory as proposing magic as an alternative to evolution. Those are, sadly, the best alternatives to evolution we’ve come up with. It’s hard to come up with any candidate ideas at all for how knowledge can be created from non-knowledge. We don’t have a single reasonable alternative to evolution.

Why can evolution work with multiple different things but still be evolution, be the same process? Because genes and memes have something in common. They are both replicators. Evolution can work with any replicator. The logic of evolution involves replication with variation and selection. Genes and memes are also complex replicators that are capable of doing interesting stuff in a pretty general purpose way. Other replicators can be like that too, but not all are (some are simple or domain-specific).

Thinking with Evolution

If evolution is the thing underlying our intelligent thinking at a subconscious level, what does that mean for us? What should we do with that insight? How should we adjust our conscious thinking to be more compatible with evolution?

Evolution generates lots of new things (genes or memes) by replication with variation. Variation means there are copying error so some new stuff is created. The new stuff could be an improvement, but more often it’s worse. The reason evolution is effective is error correction. The random variation part of evolution creates many errors, but the selection part of evolution is not random – it’s more likely to get rid of errors than to get rid of improvements.

In conscious human thinking, we can mirror this process by brainstorming many ideas (replication with variation) and then using criticism to rule out ideas (that’s selection – some ideas survive criticism and some don’t). Critical thinking is the key reason that intelligent thinking is effective. As Karl Popper put it, we learn by “conjectures and refutations”. I think calling it brainstorming (or guessing) and criticisizing is easier to understand, but it’s the same idea.

Note: I’m drawing on ideas about evolution and knowledge from many people including Richard Dawkins, Karl Popper, and David Deutsch. They learned some of it from earlier thinkers like Charles Darwin and William Paley.

Brainstorming, Criticism and Intellectual Tolerance

Brainstorming is the less important part of thinking because random change isn’t knowledge creation or adaptation to a purpose. Actual conscious brainstorming is far from random, but that’s because it relies on subconscious evolution. Our brains are powerful computers and can evolve a bunch of ideas in under a second.

To facilitate effective thinking, we should try not to suppress brainstorming. In other words, be tolerant of weird ideas. Don’t be like the inquisition. Don’t burn heretics at the stake. Don’t suppress your own ideas to try to fit in and be normal. Don’t believe every outlier idea you think of, but don’t automatically dismiss them either.

The more important part of thinking is criticism because it differentiates good ideas from bad ideas. That’s harder than just coming up with different ideas.

Rational criticism should be the way we reject ideas. We shouldn’t reject ideas due to e.g. saying they’re taboo, different, unintuitive or believed by another tribe that we regard as an out-group. We also shouldn’t judge ideas by the social status of their inventor or advocate.


There’s no great problem about how new things could be created at all. The wind blows a rock to a new location and now there is a new arrangement of rocks that is different than before. Water erodes a rock and now there is a new rock that’s a different shape than before. The hard problem that people wondered about was the source of design (purpose, knowledge). That includes things that appear to be designed, whether they were actually designed or not.

Natural forces like wind and rain wouldn’t randomly make a smartphone or monkey. But we see purposeful complexity in the world around us (such as animals having legs that let them walk, rather than random, useless appendages). The biggest key to that is not replication or brainstorming – creating random new things – but selection which actually addresses what is better or worse. Wind and rain don’t judge, evaluate, or anything equivalent. They don’t preferentially change poor designs while leaving good ones alone. They aren’t progressive – they can’t preserve what’s good and make incremental progress over time. The survival of the fittest plants and animals is biased towards the fitter ones – they have an evolutionary advantage. The wind and rain lack any bias towards progress.

Genetic evolution has a fixed goal – basically survival and replication. Animals evolve to be good at eating, reproducing and caring for their young while not dying.

Memetic evolution doesn’t have a fixed goal. We can choose our own goals. You can pick an objective – a purpose, goal or problem – and then brainstorm ways to accomplish that objective. Then you can criticize ideas when you find errors – ways they will fail at your objective. You can also have many objectives at once. The current objective may be to get lunch, but you would criticize an idea that would get you injured. That’s because staying healthy is an objective you also have at the same time.

Genetic evolution does selection/criticism relative to the goal of, roughly, having a lot of grandchildren. (This isn’t actually intentional or planned. It’s just a description of what happens due to the laws of physics. It’s the logic of the situation.)

Intelligent thought does selection/criticism relative to our goals, which we can change. We can brainstorm new goals. We can also criticize goals (relative to some of our other goals, or sometimes relative to themselves when a goal is self-contradictory).

You cannot judge which ideas are good or bad with no context, no goals. Evolution adapts things to purposes. A purpose must be specified. A wolf is designed to hunt, breed, etc. It’s bad at other purposes like flying or living in outer space. Human thinking is more flexible. We can choose purposes like going to the moon and surviving there, or building airplanes and gliders to enable us to fly through the sky.

Overlap Between Goals

The purpose that knowledge is adapted to matters less than you might expect. Some purposes are bad and cause trouble. But many overlap. There’s convergence on what sorts of ideas and actions are able to achieve many different purposes.

For example, there are many goals where “be honest and hardworking” is part of the answer. There are, in some sense, more goals like that than goals where “be a lazy murderer” is part of the answer. It’s hard to count goals, partly because there appear to be infinitely many logically possible goals. But being a lazy murderer is effective at little besides goals that specify murder and laziness in the goal. If the goal appears to be about something else, then being a lazy murderer probably won’t work for it. However, honesty and hard work are effective for many goals other than being honest or hardworking. E.g. they’re useful for farming lots of wheat or building a car. They have some widespread, generic applicability to other goals. I talk about this more in my dialog about moral foundations and maximizing squirrels.

Error Correction

So we pick goals to adapt ideas to, but they aren’t arbitrary. And we generate random ideas and then criticize them relative to our goals. The criticism is the error correction step. Having a way to correct errors is crucial for making evolution of ideas function. It’s also crucial because errors are inevitable. There are no processes that can guarantee errors will never come up. We can’t avoid all errors preemptively. There’s no way to brainstorm only good ideas or do activities only perfectly correctly. We’re fallible. We make mistakes. And the physical world has random variance. We can’t have perfect, total control over it. Atoms wiggle around and sometimes end up in the wrong places. We have to build things with margins of error, engineering tolerances, resiliency, robustness, etc. And we have to think in resilient ways too. Being able to correct errors is how you get resiliency, since if something goes wrong it can be fixed.

Rationality is also about being good at error correction. If your mistakes can be corrected effectively, then your thinking is rational. If they can’t then that’s irrational – it means you’re doing some things to block progress and improvement.

This focus on error correction in thinking differs from a lot of widespread beliefs. Not everyone thinks this way. People often try to figure out how good ideas are. They try to judge arguments and evidence as strong or weak, and figure out how much weight is behind each idea. They try to figure out which ideas are superior more than they try to find errors. They look for good traits of ideas a lot. They’re more focused on positives than negatives. That’s a mistake. That approach doesn’t resolve conflicts between ideas.

Error isn’t a matter of degree. For a well-defined goal which defines what counts as success and what doesn’t, we can judge an idea as either succeeding or failing. An error is a reason it fails. No positive traits can guarantee success or prevent errors from causing failure. So instead of considering how good an idea is, or how strong the arguments for it, or how much evidence it has (or negative versions of those like how poor an idea is), we should consider if we know of a (decisive) error: a reason the idea will fail at a specific goal we have.


Understanding evolution helps us understand epistemology better. This understanding encourages us to brainstorm ideas more. And it encourages us to focus more on error correction rather than degrees of goodness, which is a more productive way to think.