Can robots have phobias?

This paper evaluates the use of synthetic modeling to investigate the relationship between organic and artificial forms of behavioral mal-adaptability. In particular, it addresses the character of organic phobias and the issue of testing the validity of artificial models of these phobias. The two main accounts of organic phobias, the biological or evolutionary and the associative learning explanation, are used as the starting points of this exercise. The learning approach is explored in terms of a probability based model which uses a discrepancy mechanism to represent the artificial phobia, while the endogenous aspect of artificial phobias is discussed in terms of the potential offered by evolutionary learning. Several methods of assessing the construct validity of artificial phobias are outlined.