Causal modeling to reduce risks and improve performance
Causal modeling helps in eliminating unreliable correlative relationships between variables. Eliminating such unreliable relationships reduces the risks of wrong decision-making across different domains of applications for machine learning, such as healthcare. Decisions in healthcare, such as diagnosing diseases and assigning effective treatment regimens to patients, have a direct effect on quality of life and survival. Hence, decisions need to be based on reliable models and relationships in which causal modeling and inference could help us (Richens et al., 2020; Prosperi et al., 2020; Sanchez et al., 2022).
Causal modeling techniques help in eliminating bias, such as confounding and collider bias, in our models (Prosperi et al., 2020) (Figure 15.1). An example of such bias is smoking as a confounder of the relationship between yellow fingers and lung cancer (Prosperi et al., 2020). As shown in Figure 15.1, the existence of...
