Towards causal reasoning in building performance modelling with multiple data sources: a viewpoint

Many building performance models learn statistical associations from historical data without explicitly representing underlying operating mechanisms. This can reduce reliability under changing controls or environmental conditions. This Viewpoint examines how multiple data sources (MDS) and causal reasoning can support building performance prediction and control-oriented modelling. MDS can broaden operating-condition coverage and improve measurement of potential confounders, while structural and semantic information can constrain plausible relationships. However, more data do not by themselves establish causality. Causal reasoning makes treatments, outcomes temporal ordering, and identification assumptions explicit, enabling confounding adjustment and counterfactual evaluation. We synthesize three published studies on causal-graph-based interpretation, confounding-adjusted indoor-temperature prediction, and simulation-supported surrogate modelling. One study is presented as an end-to-end example. We conclude with research priorities in data support and provenance, temporal and semantic alignment, validation, and simulation-to-operation transfer.