Simulation-based exploratory data generation and analysis (data farming): a critical reflection on its validity and methodology

Abstract

Data farming' is based on the idea that simulation models run thousands of times can provide insights into the possible consequences of different options. However, the validity of the models used for data farming, especially in the context of HSCB (human, social, cultural and behavioural) modelling for decision-making and future studies, is at least questionable. This paper first reflects on the epistemological aspects of this predicament in order to illustrate its fundamental severity. Then, a possible solution is presented that is based on the notion of 'bad models', the concept of plausibility, and the method of simulation-based weak point analysis. The approach can be complemented by interactive war gaming. Such a systematic approach appears more defendable than most attempts to use HSCB models for affirmative purposes, and is methodologically easier to implement since it solely requires focusing on the validation of empirically amenable microprocesses.

Using System Dynamics for Grand Challenges: The ESDMA Approach

“…In either case, the challenge for the analyst is to develop the computer simulation models such that the ensemble of dynamic hypotheses can be explored for its consequences. This in turn becomes a key issue in model testing, for the purpose of simulation models in exploratory modelling is to explore the consequences of uncertainty (Hoffman, ). A key question that therefore must be addressed in model testing is the extent to which the developed models and associated uncertainties cover the space of plausible models.…”

Inhalt Tagungsband ASIM 2022 – 26. Symposium Simulationstechnik

“…Die V&V stammt ursprünglich aus der Softwareentwicklung, ist aber aufgrund des wachsenden Einsatzgebietes von Modellen und datenbasierten Anwendungen nicht auf diese beschränkt (Scheidler und Rabe, 2019;Hofmann, 2013). Allgemein erfolgt durch die V&V eine Kontrolle unterschiedlicher Qualitätsmerkmale von Daten, Modellen bzw.…”

ASIM SST 2022 Proceedings Kurzbeiträge

“…Die V&V stammt ursprünglich aus der Softwareentwicklung, ist aber aufgrund des wachsenden Einsatzgebietes von Modellen und datenbasierten Anwendungen nicht auf diese beschränkt (Scheidler und Rabe, 2019;Hofmann, 2013). Allgemein erfolgt durch die V&V eine Kontrolle unterschiedlicher Qualitätsmerkmale von Daten, Modellen bzw.…”