The SEXTANT software: A tool for automating the comparative analysis of mental models of dynamic systems

Highlights

• Most management situations are dynamic; and manager’s decisions are framed by their mental models.

• Mental models of dynamic systems (MMDS) have three levels of representation: elements, loops and the whole model.

• Distance ratios at each level reveal differences both between subjects and over time.

• SEXTANT calculates the distance ratios for elements, loops and the whole model for large samples of MMDS.

• SEXTANT enables researchers to strongly reduce the efforts to apply MMDS research.

Abstract

The comparison of mental models of dynamic systems improves our understanding of how people comprehend, interpret, and subsequently influence dynamic management tasks. Approaches to comparing mental models currently used in managerial and organizational cognition research, especially the distance-ratio and the closeness-approaches, have been criticized for not considering essential characteristics of dynamic managerial situations. This paper builds on a recent analysis method developed to compare mental models of dynamics systems, and introduces this mathematical approach to management and organizational researchers by means of the SEXTANT software. It presents the process of mental model elicitation, analysis, comparison, and interpretation. An example with four elicited mental models illustrates the software’s features to analyze and present the results. Then, the software is compared with existing software to map and compare mental models. Our conclusion is that SEXTANT marks a significant step in enabling large-scale studies about mental models of dynamic systems.

Introduction

Managers have to make purposeful decisions and implement actions in dynamic environments. Scholars in the field of managerial and organizational cognition (MOC) have studied how managers make sense of influence, causality, and systems’ dynamics in such environments (e.g., Eden et al., 1992, Hodgkinson, 2002, Huff, 1990). Perception and the mental process resulting in decisions are framed by mental models—an internal representation of conceptual and causal interrelations among elements that people use to understand phenomena (Groesser, 2012). In general, such models are represented by some form of conceptual diagram or cause map.

Cause mapping is an approach to representing managers’ mental models about a decision situation which has been used, for instance, to study the formation of beliefs (Markoczy, 1997), strategic change (Hodgkinson and Healey, 2008, Markoczy, 2000, Markoczy, 2001), top management team decisions (Ensley & Pearce, 2001), and team performance (Ellis, 2006, Mohammed and Dumville, 2001, Osborne et al., 2001). Research about managerial and organizational cognition has also shown that more accurate mental models can lead to more successful and more sustainable decisions, especially since a dynamic approach to decision making has the potential to make significant contributions to how managers adopt strategies and how well they perform (e.g., Gary and Wood, 2011, Kunc and Morecroft, 2010, Paich and Sterman, 1993, Sterman, 2010).

One of the latest developments in managerial and organizational cognition is the notion of mental models of dynamic systems (MMDS). Besides other common approaches to representing knowledge (e.g., Bougon, 1992, Brown, 1992, Clarke and Mackaness, 2001, Hodgkinson, 2002, Huff, 1990, Yukso and Goldstein, 2004), MMDS are useful for conceptualizing how humans think, reason, make sense, and learn about systems’ dynamics, that is, systems that change in response to decisions taken to influence them (Doyle, 1997, Doyle and Ford, 1999, Groesser, 2012, Groesser and Schaffernicht, 2012).

Once MMDS have been elicited, they can be compared. There are two purposes for MMDS comparison. First, pre- and post-comparison, i.e., differences between MMDS of individuals at several instances of time, allows one to estimate the learning of individuals over time (Schaffernicht, 2006, Sterman, 2010). And second, computing differences between several individual MMDS at the same point in time, for instance between novices and experts, allows one to assess the current level of individuals’ comprehension of dynamic situations as compared to a reference model which typically represents the understanding of an expert.

Groesser and Schaffernicht (2012) operationalize MMDS on three levels: elements, which contain variables and causal links, then individual feedback loops, and finally the complete model. This operational definition allows comparing MMDS to gain insights about the degree of similarity of such models. With this, research about MMDS can advance to levels which research about mental model of static systems has already achieved (Fiol and Huff, 1992, Hodgkinson and Healey, 2008, Tyler and Gnyawali, 2009). Since MMDS specifically address dynamic situations by means of conceptual components which traditional mental models do not take into account, a specific method for comparison has been developed by Schaffernicht and Groesser (2011). They define the levels of comparison as the Element Distance Ratio (EDR), a set of Loop Distance Ratios (LDRs), and a Model Distance Ratio (MDR). The measure of the dissimilarity of two MMDS is the distance between both. Existing mental model approaches and software tools do not operationalize the MMDS approach.²

The paper advances MMDS research by introducing and applying a sequence of steps to be carried out in MMDS comparison along with the SEXTANT software tool—a set of algorithms which automates most of these steps. This constitutes a progress, enabling researchers to undertake large-scale studies by automating the extensive calculations required by such comparisons of MMDS. Large sample sizes quickly exceed the research resources available, with the result of stagnating research in this area. In addition, statistical inferences can be made regarding patterns of global model configuration. The name “SEXTANT” refers to the purpose of the tool: it should help to analyse characteristics from many articulated MMDS, just as the old orientation device helped seamen to navigate across wide oceans. With SEXTANT, we respond to a call to advance MOC-research beyond small-scale, inductive studies (Hodgkinson, 2002, Huff, 1997, Walsh, 1995).

The paper is organized as follows: Section 2 introduces mental models and especially MMDS. Section 3 explains the process of comparing, analysing, and interpreting MMDSs. In addition, it presents the outlines of selected algorithms in the SEXTANT tool. Section 4 compares the SEXTANT software to other existing software tools. Section 5 discusses the benefits of our software for research and details its limitations. Statements about directions for future work conclude the article.