A new way of defining metacrisis
I was watching Rufus’s video on metacrisis.info and something occurred to me that I think might provide a more precise way of defining what we mean by the metacrisis.
Rufus has described the problem in terms of three causal layers:
Surface layer → Intermediate layer → Root layer
This maps almost perfectly onto the Logical Thinking Process:
Undesirable Effects → Intermediate Causes → Root Cause(s)
But I think we can generalize the model one step further.
The important thing may not actually be that there are three layers.
There can be an arbitrary number of layers.
Start with what we can observe
Suppose we begin with some set of things happening in the world that we consider undesirable.
Call these Undesirable Effects (UDEs).
For example:
UDE₁, UDE₂, UDE₃ ... UDEₙ
For the moment we don’t need to agree on what belongs on this list. Climate change, war, political instability, loneliness, biodiversity loss, AI risk, whatever. Different people could construct different lists.
Now make a very weak assumption:
These effects have causes.
Take the immediate causes of a UDE and call them Intermediate Causes (ICs).
So we might have:
IC₁ → UDE₁
But then we can ask exactly the same question about IC₁:
What causes IC₁?
Perhaps:
IC₂ → IC₁ → UDE₁
And then:
IC₃ → IC₂ → IC₁ → UDE₁
There is no particular reason to stop after one “intermediate layer”. Intermediate causes can themselves have intermediate causes, nested to whatever depth our evidence allows us to trace them.
So rather than a three-layer model, we now have a recursive procedure:
-
Identify an undesirable effect.
-
Identify its causes.
-
Treat those causes as effects requiring causal explanation.
-
Repeat.
Now something interesting happens
If we did this independently for lots of UDEs, we would initially get lots of apparently separate causal chains.
Something like:
A → B → UDE₁
C → D → UDE₂
E → F → UDE₃
But eventually we may discover that apparently separate chains share causes.
For example:
A
/ \
B C
| |
UDE₁ UDE₂
A single upstream cause can produce multiple downstream causes.
Those causes can themselves produce multiple further causes.
So the thing we are constructing is not really a set of chains. It is a causal graph.
And as we work backwards from many apparently unrelated undesirable effects, parts of that graph may begin to converge.
That convergence, I think, is where the idea of metacrisis becomes much more precise.
A possible definition
Here is my first attempt:
The metacrisis is the shared upstream causal structure discovered by recursively tracing the causes of multiple significant undesirable effects as far upstream as our evidence reasonably permits.
Or slightly more technically:
Let a causal graph be
G = (V, E)
where the vertices V are conditions or phenomena and an edge
x → y
means that x causally contributes to y.
Let U ⊂ V be the set of Undesirable Effects we are attempting to explain.
We recursively identify the causal ancestors of U.
The interesting objects are then those upstream nodes — or groups of nodes — that are ancestors of multiple UDEs.
The farther upstream a cause is, and the larger and more diverse the set of downstream UDEs it generates, the more plausible it becomes to describe it as part of the metacrisis-generating structure.
This gives us something I think the word “metacrisis” usually lacks: an actual discovery procedure.
We don’t have to decide in advance what the metacrisis is.
We try to find out.
This changes the meaning of “root cause”
There is one important complication.
The analysis does not require there to be one ultimate Root Cause.
There might be one.
There might be several.
There might also be a mutually reinforcing causal loop:
A → B → C → A
In that case asking which member of the loop is “the root cause” may be the wrong question. The causal unit we care about would be the whole self-reinforcing structure.
So when I say “continue backwards towards the root”, I don’t mean that we are guaranteed to eventually discover one metaphysical First Cause of Everything Bad.
“Root” is relative to the causal model, its scope and the available evidence.
We simply keep asking “what causes this?” until further causal decomposition becomes unsupported, unhelpful, or takes us outside the system we have chosen to study.
The result could therefore be:
-
one highly generative root cause;
-
several independent root causes;
-
a set of interacting root causes;
-
or one or more self-reinforcing causal systems.
That result should be determined by the analysis rather than assumed by the definition.
Why I think this matters
At present, “metacrisis” often functions a little like saying:
There are all these symptoms, so there must be some kind of meta-disease generating them.
Which may be directionally correct, but “meta-disease” isn’t much of a diagnosis.
What I’m suggesting is that metacrisis should initially be treated as a placeholder for a causal structure we have not yet adequately mapped.
We start with the symptoms and work backwards.
If the causal analysis eventually tells us that climate instability, political polarization, AI races, institutional failure, ecological destruction, etc. share some set of generator functions, then those generator functions give us a progressively more descriptive account of what we have been calling “the metacrisis”.
Instead of:
The world has a metacrisis.
we might eventually be able to say something like:
These N apparently different undesirable effects are substantially generated by causal structure X, through intermediate mechanisms A, B, C and D.
At that point “metacrisis” has stopped being the diagnosis.
It has become the name of the research question that produced the diagnosis.
This also gives us a falsification condition
There is another reason I like defining it this way.
It is possible that we carry out the analysis and discover very little causal convergence.
Perhaps the things we currently group together under “metacrisis” turn out to have mostly unrelated causes.
If so, that is an interesting result.
It would suggest that there isn’t much of a metacrisis in the strong sense at all. There may simply be a large number of simultaneous crises — something closer to a polycrisis.
On the other hand, if very different UDEs repeatedly converge on the same upstream generators, then we have empirical grounds for talking about a metacrisis.
So the existence, structure and depth of the metacrisis become questions to investigate rather than premises of the investigation.
And this connects directly to intervention
There is an obvious practical consequence.
Suppose:
A → B → C → UDE₁
and
A → D → E → UDE₂
We can intervene on C and improve UDE₁.
We can intervene on E and improve UDE₂.
Both interventions may be worthwhile.
But if A really is a common generator, an intervention on A potentially changes both causal chains.
This is basically the intuition behind looking for root causes in the Logical Thinking Process: a cause becomes strategically interesting when it has a large amount of undesirable downstream consequence.
So rather than arguing abstractly about whether capitalism, Moloch, modernity, human nature, Cartesian metaphysics, collective action problems, the meaning crisis, etc. is the metacrisis, we could represent these as candidate nodes in the causal graph and ask:
What does this actually cause?
Through which intermediate mechanisms?
Which UDEs does it explain?
What causes it in turn?
What evidence supports each causal link?
That seems to me like a much more productive argument.
And potentially a collaborative research programme.
The really short version
Polycrisis describes the collection and interaction of crises we observe downstream.
Metacrisis describes the common causal structure we discover upstream.
And the way we get from one to the other is not by choosing the correct grand theory in advance.
We start with observable undesirable effects and recursively ask:
Why?