This paper develops an inductive computation algorithm upon biological mechanisms discovered by the immunology. We build an evolutionary search algorithm based on a model of the immune network dynamics. According to it, the concentration of lymphocyte clone-like solutions is determined by the degree of recognition of antigens, as well as the extent of behavioral interaction with other members of the population. The antigen-like examples also change their concentration to gear up solutions matching slightly covered examples. These dynamic features are incorporated in the fitness function of the immune algorithm in order to achieve high diversity and efficient search navigation. Empirical evidence for the superiority of this immune version before the simple genetic algorithm on automata induction tasks are presented.