Validation and Verification of Computational Models with Multiple Cognitive Agents

Wider issues of the validation of computational models - ascertaining that they are sound and consistent relative to some logical formalism and/or substantive theory - have not been a subject of the management science literature. In this paper, we demonstrate that computational models can be sound and consistent relative both to a fragment of strongly grounded autoepistemic logic (FOSGAL) and to theories of cognition without losing the expressiveness found in the informally oriented literature on organizational learning and business strategy. Validation is achieved by implementing models and their theoretical components in a programming language which corresponds to a known formal logic. The language used in this paper is SDML. The correspondence of SDML to autoepistemic logic is explained and justified. Issues associated with the verification of models - how well they correspond to observation - are also considered and extended. Benefits of explicit validation and verification of computational models are demonstrated by the implementation in SDML of a computational model of the critical-incident management organization of one of the largest public utilities in Europe. On the basis of the reported simulation results with the model, several research issues are identified both for the development of validation practices in the management sciences and for the analysis of crisis management.