Multi-Criteria Decision Making (MCDM) with Causal Reasoning for AI/ML Applications – A Survey

Multi-criteria decision making (MCDM) refers to making the best possible decision out of different alternatives based on factors which can sometimes be vague. MCDM methods focus on structuring complex operations to choose the best possible decision out of the available alternatives by weighing and evaluating the different criteria. The process involves alternatives, criteria, weights, and the decision makers. It is utilized primarily for aiding in decision making with domain knowledge. However, the availability of domain knowledge for all the different alternatives can be a challenge, especially when using observational datasets. To overcome the lack of domain knowledge (sometimes the proper data) causal reasoning can be utilized to enhance the decision-making process by highlighting the hidden relationships with the data. Causal reasoning can discover the hidden relationship between the variables in observational datasets, and then provide estimates of treatment effects of the various alternatives. This study provides an overview of the causal reasoning with MCDM methods. We provide a detailed survey of MCDM and MCDM methods along with causal reasoning. We provide a detailed survey of how causal reasoning techniques have been used with MCDM methods for different applications. Finally, we highlight the challenges associated with the use of causal reasoning techniques with MCDM and provide some potential perspectives and solutions towards them.