Cognitive Relevance and Chance Discovery

We propose a formal theory to capture qualitatively the characteristics of utility-relevance and effect-uncertainty of chance. It sets a definite boundary between chance discovery and classical artificial intelligence planning: the former pursues partial realization of desire while the latter complete realization. The theory also provides a formal specification and deriving machinery for partial realization. The main formal properties and cognitive rationality of the theory are examined and justified. According to our theory, chance discovery is reduced to a class of computational problems, which can be solved on the basis of qualitative, e.g. common-sense, knowledge but cannot be solved by classical AI planning.