IDEAL

Iterative Dialectic Engine for Automated Learning


The process of iterative development

What came first - theory or experiment?  The theory of induction puts experiment first: 

induction.  A method of reasoning by which a general law or principle is inferred from observed particular instances. (1)

In setting out his theory of falsification, Karl Popper (with characteristic emphasis) took the opposite view:

The theory to be developed in the following pages stands directly opposed to all attempts to operate with the ideas of inductive logic.  It may be described as the theory of the deductive method of testing, or as the view that a hypothesis can only be empirically tested - and only after it has been advanced. (2)

The solution to this disagreement (a fine example of Hegelian dialectic, though Popper would be loath to admit it) lies in the original question, 'What came first - theory or experiment?', so-phrased to suggest a parallel with the classic hypothetical dilemma, 'What came first - chicken or egg?'.  The answer to the chicken/egg dilemma is, of course, 'An earlier chicken and an earlier egg', and likewise the solution to the theory/experiment question is, 'An earlier theory and an earlier experiment', thus:

Therefore the critical process is iteration (or iterative development) between theory and experiment, and the theories of induction and falsification each tell only half the story.

The process of iterative development appears in many fields.  The following examples are taken from psychology and military studies, respectively:

David Kolb's learning cycle

John Boyd's OODA loop

A key feature of Kolb's theory is the appreciation that, while different individuals tend to prefer to spend the majority of their time in one part of the learning cycle only (thereby distinguishing 'pragmatists', 'reflectors', 'theorists' and 'activists'), the most effective learning takes place when each individual is able to undertake all parts of the cycle with ease.  A key feature of Boyd's theory is the appreciation that a combatant can defeat a stronger opponent simply by maintaining a greater operational tempo, i.e. by undertaking the OODA loop faster than the opponent.  In both cases the emphasis is on the overall process of iterative development rather than any individual component.

It is concluded that any design of an automated learning engine must be based on the process of iterative development.

References

  1. Antony Flew (ed.), A Dictionary of Philosophy (Pan 1979).

  2. Karl Popper, The Logic of Scientific Discovery (Hutchinson 1959).

Click here to go back


Copyright © Roger Kingdon 2004