Graphical models and the challenge of evidence-based practice in development and sustainability. (August 2020)

• Record Type: Journal Article
• Authors: Calder, Ryan S.D.; Alatorre, Andrea; Marx, Rebecca S.; Mallampalli, Varun; Mason, Sara A.; Olander, Lydia P.; Jeuland, Marc; Borsuk, Mark E.
• Is Part Of: Environmental modelling & software. Volume 130(2020)
• Publication Date: 2020-08
• Subjects: Results chain -- Bayesian network -- Logic model -- Evidence assessment; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2020.104734
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Governments and social benefit organizations are expected to consider evidence in decision-making. In development and sustainability, evidence spans disciplines and methodological traditions and is often inconclusive. Graphical models are widely promoted to organize interdisciplinary evidence and improve decision-making by considering mediating variables. However, the reproducibility, objectivity and benefits for decision-making of graphical models have not been studied. We evaluate these considerations in the setting of energy services in the developing world, a contemporary development and sustainability imperative. We develop a database of relevant causal relations (313 concepts, 1337 relationships) asserted in the literature (561 peer-reviewed articles). We demonstrate that high-level relationships of interest to practitioners feature less consistent evidence than the causal relationships that underpin them, supporting increased use of problem decomposition through graphical modeling approaches. However, adding such detail increases complexity exponentially, introducing a hazard of overparameterization if evidence is not available to match the level of mechanistic detail. Highlights: Practitioners desire evidence in support of high-level relations across multiple disciplines. Graphical models represent high-level relations as a network of detailed causal mechanisms. Frequency and consistency of articles cited in reviews are measures of evidential support. Detailed mechanisms have evidential support that is more reliable but less generalizable than high-level relations. A strong evidence basis may thus only exist for causal relations limited in scope or scale.