Thursday, February 19, 2009

Theory vs Observation -- Philosophy of Science

There are essentially two ways of conducting research ... OK, everything can be made binary.

But in this respect, I am talking about what I am starting to call Theory-based research and Experiment-based research ... and I am thinking specifically about IT research.

Theory-based research involves determining an idea or a structure, learning the state of the art, then forming a new theory, perhpas expressed mathematically. Then, in computer science (computing science, web science, IT?) an instantiation of the theory may be created (i.e. a program). The program, however, is a demonstration of the theory in practice...what is desired is already known, the outcome pre-determined (if the theory was fully thought out)

Experiment-based research involves determining an idea or structure, learning the state of the art, then constructing experiments and observing the results. The theory is developed later, from generalizing observations and deriving rules based on them. The program created is the experiment and its outcomes are not known because the point of the exercise is to see what happens.

Both approaches are valid and science mores forward due to both. The first method is more deductive and the second more inductive, although these are somewhat arbitrary descriptions.

Experiment-based research is more to do with engineering in its purest sense as it expects to iteratively refine a design and to discover a rule through observation. Theory-based research is more to do with mathematics and thought experiments, any engineering being, to some degree, superfluous to the point of the research.

They are two sides of the same coin, attempts to come at ideas and discover truths in different manners and each approach has its strengths and weaknesses.

I am fascinated by emergent behaviour in systems. Emergent behaviour is systems is very hard to explore (perhaps impossible to explore) apriori via theory. The 'system' must be created and allowed to operate to observe emergent its very nature, emergent behaviour cannot be predicted, although it can be encouraged.

Consequently, most of my research is experiment-based ... build it and truths will come. I like that approach. it feels more ... satisfying ... to me.


1 comment:

Daniel Sepulveda said...

Although I agree with you on the fact that the experimental way of discovering is more satisfying, I can venture to say that this may be because it is more intuitive. However, this intuitive nature of the discoveries will largely be bounded by experience and what is even more limiting: perception.
The theoretical approach will in many cases open whole new areas, challenging paradigms and setting completley new environments from where to understand reality.
It is however satisfying to see scinetific pondering such as what you expose. Greetings!