James P. Crutchfield - Is Anything Ever New? Considering Emergence (1999)
James Crutchfield is a veteran of the Santa Fe institute and director of UC's Complexity Sciences Center. From an information-theoretic standpoint, he here considers the optimal approach for an observer to explain the behaviours emerging from a black-box natural system. The solution put forward here is to attempt to built a machine which generates a corresponding output, minimising:
- the model size, and
- the error margin between our model and the observed data
From the complexity of this model (which here takes the form of an FSA-like ε-machine), we can deduce the structural complexity of the underlying natural system. These ideas form the core of the computational mechanics field, behind which lie Crutchfield, Shalizi and others.
It's an incredibly dense yet engaging paper, itself a reduction of The Calculi of Emergence (pdf), probably the most essential piece of work on quantifying emergence and effective complexity.