Ergodic or non-Ergodic, that is the question.

Ergodicity is my new favourite word.

Some things are ergodic, some are not, they’re non-ergodic.
And rather than a definition (see also post-note below),
it deserves a riff …

A ubiquitous question is, does the end always justify the means or maybe no end ever justifies any & all means?

You might achieve the same physical state, some “objective” end, but we all know intuitively that sometimes the collateral damage to the rest of the cosmos and collective human psychology might outweigh the (local) objective result. For some things, when it comes to getting stuff done, we already know that the how is as important as the what.

The more complex the situation, the more we may need to consider, maybe also need to be seen to have considered, all the possible options and steps, and the more ingenuity, imagination, skill or craft may be needed to pick a best route to the desired solution.

At the level of society, politics, culture and psychology – the humanities – I doubt that would be considered remotely contentious, probably more like second nature. Duh – obvs!

But what about physics, science generally, the real world of natural philosophy? Surely “atoms” – the particles of matter and energy – don’t care how they got to some arrangement, do they? The same arrangement is the same result, surely? You’d be mad – it’s inconceivable, impossible – to think otherwise. That’s what a hard-determinist would say. In reality and in thought experiments, running repeated cases from the same starting condition to the same final state, must achieve the same result, identical in absolutely every way. Indeed, even the thought experiment’s stock-in-trade, “if” the situation was reset to an identical starting state, we can safely ignore questions of how the reset could conceivably, possibly, let alone tractably or physically be achieved, after all it’s only a thought experiment. Physics is deterministic, therefore … all bets on variations beyond statistical uncertainty are off.

Well the hard-determinists, the greedy-reductionists, are wrong. Many processes in the natural world exhibit both route-dependent and route-independent properties. The properties of states depend on their histories as well as their arrangements. The net result of their histories cannot be reduced to the arrangements of the component parts in their end states. Think about that.

Nassim Nicholas Taleb uses Ergodicity in the context of downside risk-management in a chapter he’s pre-published from his forthcoming book Skin in the Game. Where downside risks in probability tails involve not simply cumulative damage but binary all or nothing, dead or alive, result or ruin, success or failure, it makes a hell of a difference how you arrive at a population of cases to derive and interpret your stats. One person taking the same risk many times is quite different from a distributed population of individuals each risking their own skin in their own games. “Average” risk is pretty meaningless in the former case where the sequence of repeated risks clearly matters unlike the latter.


… a situation is deemed non-ergodic here when observed past probabilities do not apply to future processes. There is a “stop” somewhere, an absorbing barrier that prevents people with skin in the game from emerging from it ”and to which the system will invariably tend. Let us call these situations “ruin”, as the entity cannot emerge from the condition. The central problem is that if there is a possibility of ruin, cost benefit analyses are no longer possible.

Nassim Nicholas Taleb

Stuart Kauffman chooses non-Ergodic as the key concept more people really need to understand in his response to the 2017 Edge Question. It’s very brief and profound, so presented with acknowledgement in its entirety here:

“Non-ergodic” is a fundamental but too little known scientific concept.
Non-ergodicity stands in contrast to “ergodicity. “Ergodic” means that the system in question visits all its possible states. In Statistical Mechanics this is based on the famous “ergodic hypothesis, which, mathematically, gives up integration of Newton’s equations of motion for the system. Ergodic systems have no deep sense of “history.” Non-ergodic systems do not visit all of their possible states. In physics perhaps the most familiar case of a non-ergodic system is a spin glass which “breaks” ergodicity and visits only a tiny subset of its possible states, hence exhibits history in a deep sense.
Even more profoundly, the evolution of life in our biosphere is profoundly “non-ergodic” and historical. The universe will not create all possible life forms. This, together with heritable variation, is the substantial basis for Darwin, without yet specifying the means of heritable variation, whose basis Darwin did not know. Non-ergodicity gives us history.
Stuart A Kauffman

I’m pretty sure now I’ve seen Dennett use ergodicity in his evolutionary explanations towards consciousness, and I know now Kauffman must have used it in his Reinventing the Sacred which I’ve read and reviewed before, but it never really registered – as a word – until I saw those two references above within 24 hours yesterday.

There are so many corollaries from appreciating the distinction, that I’d probably better stop and leave the concept as food for thought.



[Post Note: Firstly, a “working definition” might be useful:

Ergodicity is a technical concept about the probability of states at the fundamental entropy level, after Boltzmann. Ergodic theory is about the relationship between any one state of a system and all the possible states it might otherwise be in or have been through on its “path” to that state.

Essentially when the probability of a given (outcome) state has no dependency on previous or alternative states – ie they’re all equally likely and independent of any actual path(s) through states to get there – a system is “ergodic”.

Most evolving macro systems are non-ergodic and actual history matters. Obvious to the humanities, not so obvious to greedy-reductionist scientists.]

[Post Note: Below are a growing lists of (very rough) thoughts sparked-off already.
Shout up if you’d like me to elaborate.

Conscious, intentional stuff happening within physics, but not supported by physics ?!?
Because – info underlies both material / energetic physics AND information processing (conscious or otherwise).

Reality. Ontology. Possibility. Conceivability. (Marletto in Edge 2017)
(Spookily, if one substitutes Shakespeare’s original verb “to be” it’s remarkably close to the – objective, deterministic – existential point, but not the reason I posted with that allusive title. Was the Bard on this case already?)

Evolved or Divinely created. Hardly matters to a determinist, ironically, but natural history should be fundamental to normal rational people.

Causal dependency can be backward.
Kauffman reverse causality …. Taleb too points out similar error.

Bayesian approaches look more sophisticated than simple chance, where knowledge gained from earlier choices / chances affect current decisions, but we really need to be aware what kind of ergodic / non-ergodic systems we are dealing with first. Bayes can be as misleading as naive single-choice chance.

Is there any genuinely objective reality (that matters)?
Realities that matter have subjectivity – “skin in the game”

Reversibility. History matters in irreversible incorporation – integration is more than arrangement of components. (Tad Bonicziewski – quality management). Entropy, 2nd Law and probabilistic mechanics are fundamental – ubiquitous – here.

So … Integrated Information Theory?

(Irony of historicity given Taleb / Beard spat? I won’t mention it if you don’t 😉

Think also about those fuzzy areas between “hard” science and “soft” humanities …
(eg Engineering, psychology, statistics and risk?)
Engineering humanities – “Rivets in formation” (Ingenuity is the root of engineering.)
However many things and layers in our ontology, EVERY thing and every interface comes in 3-layers.
Thing <> Interface <> Thing = 3 things
Interface <> Thing <> Interface = 3 things
Difference is simply the choice of edge vs node view in system networks?

Evolutionary design-space (in Dennett B2BnB sense) clearly closely related to evolution not visiting all possible states (in the Kauffman sense above) – directedness of evolution, as opposed to a random walk, is not just an illusion.

Here, Anthony Garrett Lisi, another physicist from Edge 2017 that sees the limits to greedy-reductionist determinism. Emergent objects are better treated as independent of their components parts!?!]

[Post Note August 2018: Taleb’s brutal style is making the running in public consciousness, but this is a real and deep issue:

(Boltzmann acknowledged in there too.)
I really do have to create a coherent piece

based on the above rough notes.]