Margaret A. Boden
Honey-bees returning from new territory are in the happy position of being able to communicate unambiguously the desirability of what they found there. Students of animal behaviour who wander from the familiar paths of ethological and psychological research into the field of artificial intelligence (A.I.) will not find themselves similarly blessed. For there is both good news and bad to be broadcast to their fellows after such a foray. The good news is that animals must indeed be credited with the ability to form symbolic representations; that this is so even in the absence of communicative behaviour on their part; that there exists a source of concepts for clearly articulating the structure of and transformations between different representations; and that there exists (though at a great distance) the possibility of a principled matching of varying content and function to distinct representational forms. The bad news is that the problems of formulating plausible hypotheses about animals' representations are even more complex than is generally believed; that, given any such hypothesis, the possibility of testing it is more problematic than might at first appear; and that the most puzzling feature (though not all features) of consciousness remains unresolved by this approach as by all others.
This paper is not available online