Most examples of non-ergodicity are activities in which the outcome of a person completing them many times is lower than the outcome of many people completing it once. For example, Russian Roulette.
But there are cases in which its higher.
1/9
2/ First, if you don’t know about ergodicity, I suggest reading this thread:
3/ One classic example of non-ergodicity is Russian Roulette. The expected outcome of 600 people playing it once is 100 dead and 500 winners, whereas the outcome of a single person playing it 100 times is 1 dead and no winner.
4/ However, there are activities with an opposite effect. The outcome of 100 people cooking one tart for the first time is 100 terrible tarts, whereas the outcome of one person cooking 100 tarts is a 5 terrible tarts (at the beginning) and 95 great tarts (once learning kicks in).
5/ Why? And why does it matter?
6/ One interpretation of the field of ergodicity is that it concerns itself with the study of the “phantom consequences” of actions, that do not matter if the action is performed once but affect the *next* time it is performed.
7/ For example, if a skier breaks his leg in a race, the consequences are “only” pain and losing the race. But in the context of a championship there’s an additional consequence: losing all future races (because he cannot participate). This is the phantom consequence.
8/ Similarly, if I cook a tart and it’s the only time in my life I will ever cook one, the fact that I learned how to make better tarts is irrelevant. A “phantom consequence”.
But if cooking more tarts is a possibility, then the learning becomes a real and valuable consequence.
9/ Hence, in all cases in which repeated actions are a possibility, it’s important to consider “phantom consequences” that are impossible to observe by studying a single action in a vacuum.
How come that as we get better tools to be more productive (software, …), in some jobs, productivity didn’t increase too much?
More tasks that don’t add value, of course. But why do we choose to engage with them, rather than being productive?
1/N
2/ To explain this phenomenon, called productivity homeostasis (which roughly translates to “stays the same”), we must first look at a similar phenomenon: risk homeostasis.
3/ The Fence Paradox (see image below) is an example of risk homeostasis: the idea that, when an activity becomes safer, people often react by increasing their risk taking.
A peek inside my adaptive systems course starting on the 23rd of February.
In this thread, a list of what participants will learn.
1/N
MODULE #1: HARNESSING ANTIFRAGILITY
The organic is both antifragile (we lift weights → our muscles grow) and fragile (we lift too much → we injure ourselves).
What determines antifragility?
What's the relationship between it and fragility?
What to do about it? 2/N
3/ Antifragility can make us stronger (exercise → stronger muscles) or weaker (no exercise → muscles atrophy).
It can make us adapt (famine → we adapt by storing more nutrients) or maladapt (lack of famine → we take risks & store less nutrients, making us more fragile).
Societies are adaptive systems. What a policy does is less important than how people adapt to it.
Our body is an adaptive system. We lift weights not to move them, but for how our muscles adapt to it (they grow).
(thread, 1/N)
Teams are adaptive systems. In the short-term, a manager's decision matters for what it does. In the long-term, it matters for how the team adapts to it. What behaviors does it make more likely?
2/N
Markets are adaptive systems. Many strategies only work until the market adapts to them.
Marketing, sales, and strategy are about adaptive systems. In the long-term, what matters is how customers, competitors, and suppliers adapt to a new product.
3/N
A Nobel prize can tell us two things: how good is the recipient or how bad is the committee.
“Unless you have confidence in the ruler’s reliability, if you use a ruler to measure a table you may also be using the table to measure the ruler.” – @nntaleb
2/ I used to express Wittgenstein’s ruler as follows: the more the free parameters, the less you know what is being measured.
For example, last spring COVID mortality could have been informing us about how aggressive is the virus or how good is a country’s testing
3/ In addition, and this is the point of this new thread, it just dawned to me that Wittgenstein’s ruler is not just about the precision of the ruler but also about its choice.
Those who put too much ego in their car do things that are good for their car (eg spending Saturdays afternoon washing it) rather than things which are good for them (eg hanging out with their friends and family).
We do what is good for what we invested our ego in.
1/4
2/ Those who put too much ego in their job stop doing things that are good for them and instead do things that are good for their job.
Those who put too much ego in their political party stop doing things that are good for them and instead do things that are good for their party
3/ Those who put too much ego into racial discrimination stop doing things that are good for them and instead do things that are good for racial discrimination.
When Twitter banned Trump, I wrote that even though I don't like Trump and thought that the world would be better off without him, I was also against its censorship for fear of a slippery slope.
3 weeks later, did it happen?
Yes. Examples & implications 👇
1/8
2/ After Twitter, FB banned him. Then, Google suspended Parler from its store, Apple did it too, and finally Amazon banned it from its infrastructure.
3/ Two comments:
– What looks inconsequential if one small company does it is very consequential if all major players do it.
– It started with banning a few bad apples, it ended up with banning full categories of users.
(the latter should give the chills; also see tweet #5)