Integrating Cognitive Principles From ACT-R Into Probabilistic Conditional Reasoning by Taking the Example of Maximum Entropy Reasoning

Abstract

Many modern artificial intelligence (AI) systems like human-interacting smart devices or expert systems adapt to specific users' information processes but the underlying AI methods commonly lack a theory of mind. Thus, there is a need to better understand human thinking and to integrate the resulting cognitive models into AI methods. By taking the example of maximum entropy reasoning (MaxEnt), we integrate cognitive principles from the cognitive architecture ACT-R into uncertain reasoning based on probabilistic conditionals. Therewith, we combine two powerful and well-established methodologies from probabilistic reasoning (MaxEnt) and cognitive science (ACT-R) and establish a blueprint for cognitive probabilistic reasoning.