In defence of deep modelling

Context: Since multi-level modelling emerged as a strategy for leveraging classification levels in conceptual models, there have been discussions about what it entails and how best to support it. Recently, some authors have claimed that the deep modelling approach to multi-level modelling entails paradoxes and significant weaknesses. By drawing upon concepts from speech act theory and foundational ontologies these authors argue that hitherto accepted principles for deep modelling should be abandoned and an alternative approach be adopted instead (Eriksson et al., 2013). Objective: We investigate the validity of these claims and motivate the need to shift the focus of the debate from philosophical arguments to modelling pragmatics. Method: We present each of the main objections raised against deep modelling in turn, classify them according to the kinds of arguments put forward, and analyse the cogency of the supporting justification. We furthermore analyse the counter proposal regarding its pragmatic value for modellers. Results: Most of the criticisms against deep modelling are based on mismatches between the premisses used in published definitions of deep modelling and those used by the authors as the basis of their challenges. Hence, most of the criticisms levelled at deep modelling do not actually apply to deep modelling as defined in the literature. We also explain how the proposed alternative introduces new problems of its own, and evaluate its merits from a pragmatic modelling perspective. Finally, we show how deep modelling is indeed compatible with, and can be founded on, classic work in linguistics and logic. Conclusions: The inappropriate interpretations of the core principles of deep modelling identified in this article indicate that previous descriptions of them have not had sufficient clarity. We therefore provide further clarification and foundational background material to reduce the chance for future misunderstandings and help establish deep modelling as a solid foundation for multi-level modelling. (C) 2015 Elsevier B.V. All rights reserved.