Quantifying hydrogen technology acceptance: Insights from Bayesian networks

Abstract

Traditional social science analyses often emphasise qualitative explanations, limiting the integration of insights into quantitative frameworks, which constrains predictive analysis and rigorous theoretical testing. We address this gap by showing how causal inference methods, specifically Bayesian networks, can strengthen technology acceptance theories through an explicit representation of hypothesised structural dependencies. This approach enables the principled exploration of hypothetical interventions (even in contexts with scarce data) by leveraging information in the parametrised network and adhering to a theoretically informed process called the do-calculus. We demonstrate the approach by examining hydrogen hub acceptance using survey data from 1682 Australian residents who were asked about hosting a hub in their communities. The resulting Bayesian network outperforms eight widely used structure-agnostic machine learning algorithms in predictive accuracy and identifies the strong causal influence of perceived risk and economic benefit on hub acceptance. By simulating ‘what-if’ interventions, the model delivers quantitative decision support under uncertainty, informing policy design and communication strategies for hydrogen-technology projects.