Evolution, Learning and Popper's Epistemology

N. Cristianini
Royal Holloway - University of London, 1996
nello@dcs.rhbnc.ac.uk

Abstract:

Artificial Life algorithms are extensively used for Knowledge Acquisition tasks in Artificial Intelligence, but no theoretical justification for this fact has yet been provided.

Evolutionary Epistemology represents the ultimate development of Karl Popper's analysis of knowledge acquisition processes, and stretches to its extreme consequences his view of theory formation as a process of trial-and-error. In the most general case, Popper proved, every knowledge acquisition activity must be regarded as a sequence of casual conjectures and successive refutations, in tight analogy with evolutionary processes of mutation and natural selection.

Such analogy can even be regarded as an identity, after having blown the definition of knowledge up to include the implicit knowledge embodied in the physical structure of organisms.

Genetic Learning Systems, hence, turn out to be the straightforward implementation of a core epistemological principle, which not only connects biology to the theory of knowledge, but also provides a firm conceptual background to this technology.

• Contents
• Introduction
• Evolutionary Epistemology
  • Implicit and Explicit Knowledge
  • Blind Conjectures and Informated Refutations
• Genetic Learning Systems
  • Inferring Implicit Knowledge
  • Inferring Explicit Knowledge
• Prior Knowledge in EE Model
• Conclusion
• References