Intuitive modeling of inter-causal relationships based on cognitive experiments

Abstract

In this paper we address the issue of representation of cognitive processes of managers in organizations. This purpose derives from a need felt by theories that address strategic behavior of firms. We propose to represent the cognitive processes of managers (termed as beliefs, values, etc., in organizational science theory) as qualitative probabilistic networks (QPNs). QPNs have a topology similar to cognitive maps (directed-acyclic-graphs with qualitative signs attached to edges). Perceptions (concepts) of top management are represented as nodes, and beliefs about environmental uncertainties are quantified as probability estimates. We propose intuitive models on these QPNs. These models result from studying human behavior under uncertainty with the help of psychological experiments. During the experimentation we observe two patterns of inference that subjects resort to, while inferencing under uncertainty. We incorporate these patterns into the existing theories of inter-causality. This is done by defining patterns with the help of certain probabilistic criteria. Finally, we demonstrate the applicability of these intuitive patterns in quantitative belief propagation in a cognitive map abstracted from the annual report of a company from the Indian automobile industry.