ABSTRACT
This chapter examines the generic form of learning curves in the framework of non-linear complex systems. It presents simulations that demonstrate how observed learning curves can be reproduced by a number of different methods that can lead to testable hypotheses about the learning mechanism. Theoretical models based on an algorithmic information processing approach used a "chunking" hypothesis which lead to a powerlaw prediction for the learning curve. If the learning curve is a power law then the rate is not constant but rather decreases continuously. For certain classes of complex systems scaling exponents play a central role and are "universal" in the sense that their value does not depend on system details. The negative values indicate the influence of stochastic variations. The approximately constant learning rate is consistent with an exponential form of the corresponding learning curve.
