A relevant question is: given a series of competing models, bn1 and bn2, how can we choose the best model? There are several approaches for choosing the best one, and most of them rely on information metrics such as the Akaike information criterion (AIC). These can be computed using the score() function and an appropriate type= parameter. The first model (the expert-based one) has a lower AIC than the hybrid one; in consequence, we should keep that model:

score(dag, data, type = "aic")
score(dag2, data, type = "aic")

The following screenshot shows AIC values for both models: