Grounding the Term 'Intelligent'

As I have discussed in preceding publications (cf. [], [71]) it is not clear from the outset how to use the term intelligent in a non-ambitious way. We follow here the widely accepted opinion that the primary source of the phenomenon of intelligence is rooted in certain properties of human behavior which in a deferred manner can also be applied to certain properties of animal behavior too. Analogously one can apply the term to properties of the behavior of engineered systems. With regard to the general paradigm of empirical measurement this implies that the widely fuzzy usage of the term 'intelligent' can only be turned into a measurable object for an engineering context, when one associates the term 'intelligent' with observable behavior in an explicit empirical experiment.

This idea of making the term intelligent 'measurable' is not new. At the end of the 19th century the discipline of psychology experienced a methodological crisis because psychologists were using observable behavior as well as not observable introspective phenomena as data for psychological statements (cf. Marx (1987)[208], Benjamin (1988)[24]). This inclusion of non-empirical introspective data hindered the development of real empirical theories. The following behavioral turn in psychology led to the development of psychological behavioral learning theories (cf. the overview by Bowler and Hilgard (1981) [33]). This radical shift to behavior enabled many important insights in the nature of learning as one aspect of intelligence. At the same time it revealed some methodological problems in the search for intelligence (cf.again Bowler and Hilgard (1981) and Bolles (1975)[30], Buss (2008)[37]). The key idea of the critical reviews can be summarized in the sense that to use 'observable behavior' as a primary hint for 'intelligence' is a good starting point, but because the observable behavior is enabled and driven by a complex machinery 'inside' the biological systems -some kind of neuronal network as part of a complex body- it became more and more clear that the observable behavior is not a one-to-one mapping of these underlying processes. Although the observable behavior described in some behavior model $ M_{S-R}$ is completely induced by the generating biological structure $ M_{Biol}$, one cannot deduce an unique biological structure as the only source of a certain kind of observable behavior, because the same observable behavior could be produced by quite different internal structures (cf. 2.2).


$\displaystyle M_{Biol}$ $\displaystyle \not\Longleftrightarrow$ $\displaystyle M_{S-R}$ (2.1)
$\displaystyle M_{Biol}$ $\displaystyle \Longrightarrow$ $\displaystyle M_{S-R}$ (2.2)

This reveals a dilemma: what we can observe is behavior, but the internal structures, which is generating the behavior, we can not observe. Because modern Neurophysiology (cf. e.g. Shephered (1994) [282]) has made some remarkable progress since the times of the classical behavioral learning theory we can today extend the scope of classical behavioral learning theory and include the study of the micro structure of neural networks in the body. But without the correlation with observable behavior these neural structures do not have any meaning. Moreover Biology has taught us since Darwin (cf. [313]) to embed the actual observable behavior into the wider scope of an evolutionary process where the structures of the individual systems are developing in continuous response to the actual environmental conditions. And as psychologists have demonstrated (cf. [177], [176]) the understanding of human intelligence changes when analyzed in such a wider developmental context.

Gerd Doeben-Henisch 2012-03-31