“Causal Effects Versus Causal Mechanisms: Two Traditions With Different Requirements and Contributions Towards Causal Understanding”

Following up on our post from last year on the importance of mechanistic models, Jim Grace points to his above-titled paper in Ecology Letters. From the abstract:

When methods are described as ‘causal,’ emphasis is increasingly placed on statistical techniques for isolating associations so as to quantify causal effects. In contrast, natural scientists have historically approached the pursuit of causal knowledge through the investigation of mechanisms that interconnect the components of systems. . . . the case is made that an explicit assessment of existing mechanistic knowledge should be an initial step in causal investigations.

Grace also has this webpage.

I’m posting this because I continue to be interested in the connections between mechanistic and statistical models of causality, and I continue to think that the fields of statistics, economics, and policy analysis took a wrong turn by associating “causal inference” with black-box statistical analysis. I understand the appeal of the idea–indeed, the very first applied statistics class I ever took, 40 years ago, was Rubin’s course on causal inference–but in many ways it’s been a very long detour that only in recent years have we been able to take in a more useful direction.