Kind of a terrible title, isn't it? Really, the article is asking "are complex models better predictors of natural processes than single factor correlations?" An answer other than "almost always" would be amazing.