Introduction

Abstract

This is a study about inference. Traditionally, epistemology has investigated inference from an introspective point of view. From this perspective, inference is what is going on when I consciously believe that α is true and when this is the reason for which I start to believe consciously that β is true as well. An inference like that is justified if my belief that a is true is a good reason for my belief that β is true, i.e., if a conscious, internal or external, rational argumentation has caused me to believe β on the basis of believing a. This is usually called an internalist concept of justification. The arguments that are used for such a reasoning should be logically valid ones, in a broad sense of ‘logical’ which includes inductive logic, probability theory, etc. The ideal reasoner’s laws of thought are thus identical to logical laws, since the ideal agent explicitly uses logic as the method of reasoning. More recently, this “Boolean dream” has turned into the development of artificial cognitive architectures that draw inferences in the same way as logicians, mathematicians, or scientists in general do, when they present a proof on the blackboard: i.e., by the manipulation of symbols. Starting from a traditional “internalist” view of justification, we arrive at a view of inference on the “high level”; the agents drawing such inferences may be called ‘high-level agents’. By ‘level’ we mean the level of the cognitive complexity of an agent or an inference, but also the level of the epistemological standards that are to be applied.