Correlation as part of machine learning models
The majority of machine learning modeling and data analysis projects result in correlative relationships between features and output variables in supervised learning settings and statistical modeling. Although these relationships are not causal, identifying causal relationships is of high value, even if it’s not a necessity in most problems we try to solve. For example, we can define medical diagnosis as “The identification of the diseases that are most likely to be causing the patient’s symptoms, given their medical history.” (Richens et al., 2020).
Identifying causal relationships resolves issues in identifying misleading relationships between variables. Relying solely on correlations rather than causality could result in spurious and bizarre associations such as the following (https://www.tylervigen.com/spurious-correlations; https://www.buzzfeednews.com/article/kjh2110/the-10-most-bizarre-correlations...
