[ExI] a complicated patient is the bee

Anders Sandberg anders at aleph.se
Sun Jul 28 19:23:46 UTC 2013


On 2013-07-28 18:01, spike wrote:
> The honeybee is the bug equivalent of the complicated patient.  Too many
> inputs, some of them unknown, the collective effect clearly harmful and
> possibly fatal to both the patient and her dependents.

Moving this to the meta-level (as we philosophers always do), this seems 
to be the hallmark of the Big Problems we are facing. A while ago I 
wrote about the apparent similarity between the obesity epidemic and the 
possible stagnation problem ( 
http://lists.extropy.org/pipermail/extropy-chat/2013-March/076102.html ) 
- both are very nonlinear, complex systems with tangled causality, yet 
might show coherent undesirable trends. Same thing might be happening in 
a lot of domains like finance, nuclear arms-races and disarmament, or 
(the classic) the Middle East.

Complex problems where there are many opposing and supporting adaptive 
feedback loops, of varying level of visibility and understanding, and 
where the overall causal structure is unknown and perhaps not knowable. 
Some of them are even more vexed, in that it is not clear what 
constitutes a solution state (consider the Middle East).

This is an interesting domain... especially since I am planning to work 
a bit on questions of systemic risk this autumn. Most systemic risks are 
far nicer: contagion of bad stuff from one company/subgrid/ecosystem to 
the next, instability in collective modes, and so on. They are due to 
densely interconnected systems, but the nodes are all more or less of 
the same kind: the mess happens because a global instability is reached, 
but the causal structure of the disaster is fairly straightforward. Not 
so in the complicated patient risks.

Hmm... Looking at it from an abstract perspective, what we have here is 
a partially observable nonlinear system with a lot of internal 
complexity. The reason we have a Problem is that some of the state 
variables or functions of them move in a bad direction. We can send 
various inputs to it, including changing some rules for some internal 
components. In the case of some chaotic systems there are methods to 
bring them under control, exploiting the fact that unstable periodic 
orbits are dense in strange attractors, but it seems that in this case 
we are typically not looking at anything with a strange attractor, but 
rather a big unwanted trend. It seems that there are lots of methods in 
system identification, operations analysis and other domains that might 
give interesting insights. But the core problem seems to be that we only 
get a very limited time series of trending data rather than a bigger 
exploration of their state space.


-- 
Anders Sandberg
Future of Humanity Institute
Oxford University



More information about the extropy-chat mailing list