Inductive ambiguity and the limits of artificial intelligence
Satosi Watanabé · Computational Intelligence · 1987
Most artificial intelligence (AI) systems are based fundamentally on deduction. Such systems usually rely on a number of deductive heuristics. We can justify this reliance only when we can assume that the logical lattice of propositions is closed. In particular, this restriction forces us to concede that deductive AI systems cannot be capable of performing activities such as abduction and inductive evaluation, which arc at the heart of scientific activity. To illustrate, we will look at diagnostic practices in medicine and compare the inductive‐scientific and deductive‐technological processes. Deductive AI systems can help us only with the latter type of analysis.