# Background

Some have lamented that it is difficult to distinguish randomness from probability, but I seem to be having a contrary difficulty: are they even logically dependent notions?

## Probability

While introductions to Statistics abound that will take for granted a relationship between probability and the notion of randomness, I am finding this relationship to be more optional as I read into the foundations of probability theory. It seems that probability theory doesn't require a notion of randomness to satisfy the axioms of Measure Theory and Kolmogorov. Some disciplines, such as Ergodic Theory, even compute expectations (which involve probability measures) on purely deterministic systems.

## Randomness

Randomness itself doesn't appear to be a precise notion. Most often I see definitions similar to "the property of being unpredictable". One can notice that some systems can be extrinsically random in the sense that there are models that reduce previously-unpredictable phenomena into a pseudorandom process. According to Chaos theory, this can simply be a matter of knowing the governing equations and the relevant initial conditions with enough precision. Perhaps aspects of Quantum Mechanics are instrinsically random, but even that is not a given in my view.

The closest I have found to a formal notion of randomness is information entropy, which itself goes back to being a consequence of probability theory.

All it seems to take is complexity resulting from not knowing the state-transition rules or the state of a system and we seem to get randomness.

## Summary

In summary, (1) probability theory doesn't seem to need a notion of randomness and (2) it is not clear whether randomness is anything more than limited predictability when relevant details are missing. At a glance these notions seem inseparable, but after a long gaze they seem to be at-most in a casual relationship.

# Question

Who can I read that really cuts into the meat of this problem?

• There is also so-called "algorithmic randomness", "Martin-Löf randomness", etc. Oct 27, 2021 at 0:34
• @Aksakal That is a powerful simile. The difficulty expressed above is not about intuition or a desired use case of probability theory. The axioms of probability theory as developed by Kolmogorov simply don't require the notion. Oct 27, 2021 at 19:11
• I am afraid kolmogorov would disagree. He was very well grounded in experiment. In fact in the preface to The book he says that the axioms are based on experiment in a broad sense: the theory needs to explain reality. The axioms need to correspond to what we know about the world. It’s very similar to geometry where axioms are extracted from what we learned about space around us. Yes, once the axioms are written one can forget about reality and just work with theory. However it is the connection with reality that brings most interesting developments in the theory again and again Oct 27, 2021 at 19:34
• @Aksakal Kolmogorov's axioms of probability theory are not the same as the whole of what Kolmogorov thought about probability theory, and you are appealing to his views beyond the axioms themselves. My concern doesn't invoke any of his views beyond the axioms, nor does his informal treatment seem to adequately unpack my concerns. I'm not advocating that we disregard reality (i.e. observation), or theory for that matter. I simply want to have the clearest understanding of the relations between these concepts as I can practically obtain. Hence a reference request to further that understanding. Oct 27, 2021 at 19:54
• You'd find the same studying Geometry or Mechanics: any bit of Maths only refers to sets in the end, & bridge principles are needed to connect it with the notions to which we want to apply it. Gillies (2000), Philosophical Theories of Probability is a good introduction; & IIRC makes the point that the Kolmogorov axioms could just as well be applied to the lengths of a collection of sticks as to anything having to do with random processes (or the credibility of propositions). Jan 1, 2022 at 16:12

Returning to our problem, one probability that corresponds to a lack of structure is a 50% probability of 1's and 0's, but we need a lot more than that, for example, the sequence $$10101010101010101010101010\dots$$ has an equal probability of 1's and 0's but it is structured, i.e., ever other number is a 0. So in sum, randomness is a pertinent lack structure within a context, which context, although it may seem to be contradictory for it to be so, can be highly structured.