I would argue that it is not prediction that is failing with economic systems; it is the lack of interest to change anything (from those who could) that triggers the consequences. After all, rich didn't get any poorer with the last crisis.
Thanks for that. I'm not convinced, it look like they try to correlate energy and lattice structure and then extrapolate from there to dynamic system collapse. Scanning it, I felt like it read like one of those 'autogenerated' papers, which is always a red flag for me.
It seems unlikely this would be generalizable in much of a useful way. "We conjecture that this phenomenon is universal for a class of complex dynamical systems and discuss implications." the paper says, I'd like to see more on why they think it's reasonable to conjecture that.
Let me try to summarize the paper, which I only slightly understand. They take a 2-D grid with a magnetic domain at each point, sort of like a hard drive. It starts out with random N/S orientations. Each domain "wants" to match up with its neighbors. As domains flip, they will start to match up more and more. Eventually you have a phase transition where you suddenly go from mostly-random to mostly-aligned, kind of like water freezing. This is a standard physics model called an Ising model.
They look at the "information flow" between neighboring domains. This is sort of how much influence a domain has on its neighbors, with an incomprehensible transfer entropy measure. At the beginning, when everything is random, there's not much correlation. At the end, when everything is aligned, there's no information flow since neighbors are already aligned. The peak influence is right in the middle when domains are snapping into alignment with each other. The paper measures this information flow via simulations of the grid and finds the peak information flow is in the disordered phase shortly before the phase transition to order. This seems like a reasonable conclusion.
They then conjecture that this information flow peak isn't just for the specific Ising model they investigate but may happen in general for financial crises, epileptic seizures, and other complex models, so if you can measure information flow you can tell when the phase transition will happen. I don't see anything in the paper that actually supports the conjecture, but it makes for nice headlines. (I don't want to get into credentialism, but it makes me a bit suspicious that the authors aren't statistical physicists (who usually do Ising model stuff), but work in a center for consciousness science, civil engineering, and other random departments.)
Apologies to anyone who actually understands Ising models, because I've probably totally mangled the explanation, but I figured since nobody else had posted a summary it might be useful.
This is going to sound funny, but HOW did they get the bottom bar to (unwittingly?) bring up the mobile Safari chrome around the website when tapped? Try it on an iphone... it's an incredibly nice effect
From what I understand of chaos theory, and from what I'm hearing of recent trends in scientific overreaching and poor peer reviewing, I am inclined to dismiss this paper as yet another overambitious claim to know what can't be known. Divination is the world's oldest and most alluring scam. What makes this different?
If you have a method to predict epileptic seizures you can intervene. If you have a (published) method to predict stock market crises then the agents in the stock market will take it into account, presumably breaking the method.
15 comments
[ 3.8 ms ] story [ 36.5 ms ] threadThey look at the "information flow" between neighboring domains. This is sort of how much influence a domain has on its neighbors, with an incomprehensible transfer entropy measure. At the beginning, when everything is random, there's not much correlation. At the end, when everything is aligned, there's no information flow since neighbors are already aligned. The peak influence is right in the middle when domains are snapping into alignment with each other. The paper measures this information flow via simulations of the grid and finds the peak information flow is in the disordered phase shortly before the phase transition to order. This seems like a reasonable conclusion.
They then conjecture that this information flow peak isn't just for the specific Ising model they investigate but may happen in general for financial crises, epileptic seizures, and other complex models, so if you can measure information flow you can tell when the phase transition will happen. I don't see anything in the paper that actually supports the conjecture, but it makes for nice headlines. (I don't want to get into credentialism, but it makes me a bit suspicious that the authors aren't statistical physicists (who usually do Ising model stuff), but work in a center for consciousness science, civil engineering, and other random departments.)
Apologies to anyone who actually understands Ising models, because I've probably totally mangled the explanation, but I figured since nobody else had posted a summary it might be useful.