Self-Improvement Without a Self: Actual vs Perceived Causes

24 Dec, 2025

I want to explore the idea of actual causes and percieved causes a bit more. Link to my previous blog where I started this discussion.

With actual causes we have direct causation.

"If actual causes, then action."

Perceived causes can fail when we have

"If perceived causes, but no action."

This can't happen with actual causes, by definition. This can happen with perceived causes because they can fail to map on to actual causes in some way.

So to measure how well our perceived causes match up to actual causes, we need to measure how often perceived causes precede the corresponding action.

The problem now becomes a problem of measurement. Basically, what we have to do is split both our perceived causes and our action into something we can measure. Let $x_1,...x_n$ be our measured perceived causes, then $y$ is the measured action. I choose many variables because to determine if an action occurs because we may have many causes leading towards an action. I assume countably many because that seems easier to reason about for now.

What we believe we have is a function. $f(x_1,...,x_n)=y$

Every time we measure our $x_i$ and $y$ we can update $f$ to increase prediction accuracy, but if $f$ never converges to a model that gives us good prediction accuracy we know that our perceived causes don't map onto actual causes.

I'm not sure if I'm making too much sense anymore, lol.