Solomonoff Induction (SI) focuses on short code. What’s short in English is not necessarily short in code, and vice versa. Most of our intuition in favor of short, simple explanations is from our experience with English, not code. Is there literature arguing that code and English brevity usually or always correspond to each other? If not, then most of our reasons for accepting Occam’s Razor wouldn’t apply to SI.
Another way to think of the issue may be that coding is a low level or reductionist way to deal with an issue, while English is a high level approach that uses high level tools like explanations. Ideas can be represented in many ways, including at different levels of abstraction. It’s unclear that length or simplicity is consistent for the same idea across different levels of abstraction. That is, if you have two ideas X and Y, and X is simpler and shorter when you compare both ideas at a one level of abstraction, it may be unsafe to assume that X will also be simpler and shorter than Y when you compare them at a different level of abstraction. Is there any literature which addresses this?