An adaptive model for human syllogistic reasoning

Abstract

How humans reason in general about syllogisms is, despite a century of research and many proposed cognitive theories, still an unanswered question. It is even more difficult, however, to answer how an individual human reasons. The goal of this article is twofold: First, it analyses the predictive quality of existing cognitive theories by providing a standardized (re-) implementation of existing theories. Towards this, theories are algorithmically formalized, including their potential capabilities for adaptation to an individual reasoner. The implementations are modular with regard to the underlying mental operations defined by the cognitive theories. Second, it proposes a novel composite approach based on existing cognitive theories, resulting in a cognitive model for predicting an individual reasoner before s/he draws a conclusion. This approach uses sequences of operations, inherited and combined from different theories, to form its predictions. Among the existing models, our implementations of PHM, mReasoner, and Verbal Models make the most accurate predictions of the conclusions drawn by individual reasoners. The designed composite model, however, is able to significantly surpass those implementations by exploiting synergies between different models. In particular, it successfully combines operations from PHM and Verbal Models. Therefore, the composite approach is a promising tool to model and study syllogistic reasoning and to generate tailored cognitive theories. At the same time it provides a general method that can potentially be applied to predict individual human reasoners in other domains, too.