Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

MPI Kolloquium

Wednesday, 04.12.2013 14:15 s.t.

Criticality of control in man and markets

by Prof. Dr. Klaus Pawelzik
from Institute for Theoretical Physics, University of Bremen, Bremen

Contact person: Jan Nagler


Ludwig Prandtl lecture hall


Stabilization of a dynamical system annihilates observable information about its structure. We discovered that this basic mechanism can induce critical points as attractors when the system is coupled to an adaptive controller. Investigations of human balancing behavior demonstrated that this principle can reproduce experimentally observed critical behaviour in unprecedented detail. Our results suggests, that the observed fat-tail error distributions of human balancing behaviour may reflect an optimal compromise between the elimination of random local trends and rare large errors (published in PRL). We went on and wondered if speculative markets, which also absorb self-generated information, can exhibit a similar effect: evolution towards efficient equilibrium states and their subsequent destabilisation. We introduced a minimal agent-based market model where the impacts of trading strategies naturally adapt according to their success. The latter provably implements a learning rule for the whole market minimising predictable price changes, and an extreme susceptibility at the point of perfect balance. The model quantitatively reproduces the real heavy-tailed price change (log return) distributions as well as the so called volatility clusters. Our results demonstrate that the notorious market instabilities might be a fundamental consequence of the very mechanisms that lead to a market efficiency (published in Scientific Reports). Currently we are investigating if the working point of the cerebral cortex in the brain might also correspond to a critical state where every input is adaptively balanced by self generated predictions.

back to overview