Potential benefits of model use in group model building: insights from an experimental investigation

Abstract

Abstract How do individuals discover the feedback structure of a situation? Does such knowledge help people design and select better decisions? Do decision aids such as causal loop diagrams help? These questions have so far received limited research attention in controlled studies in system dynamics. Here we test the effectiveness of different methods commonly used in group decision settings such as group model building workshops. We employed a hidden profile experimental design in which each individual in a group possessed private information that must be shared to arrive at a good solution—in this case, policies to improve performance. The task placed participants in the role of a manager in a department at the Saturday Evening Post. Processing of department‐specific information will be needed to discover the full feedback structure. We compare decision quality and structural understanding across four conditions: (i) a control setting in which participants only receive minutes of a meeting; (ii) a setting in which participants are actively engaged in qualitative modeling (developing a causal loop diagram); (iii) a setting in which participants receive a causal loop diagram as a decision aid; and (iv) setting (iii) but with all participants given full information. Model use had no effect on the importance participants attached to critical information, understanding of model structure, or confidence in their decisions. However, for participants actively engaged in modeling, we found a substantial correlation between importance attached to critical information, understanding, and confidence. The results have implications for group model building research and practice. © 2018 System Dynamics Society

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 40 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.