Understanding the nature of inference: discerning deeper correlations and causation

In the physical sciences, the existence of empirically determined laws of nature provide connections between measurable quantities that parlays inference. These deterministic laws provide robust mappings between cause and effect. The universality of these laws has permitted for instance, study of the universe, wherein despite having only one universe and despite our being inside it, we have been able to chart its origin, formation and evolution very successfully. However, in other domains of knowledge where there are no known laws as yet or where laws may not exist due to the inherently probabilistic and indeterminate nature of the systems under study, inference pathways, generation of new knowledge and deriving truth value is significantly more complicated. The ability to perform controlled experiments, either in the laboratory or via simulations has played a crucial role in inferring causation and in disentangling correlation and causation. Our goal in this project is to deeply examine the process of inference of cause and effect across disciplines with a view to understanding the utility of methods that serve as proxies for controlled experiments. To do so, we plan to assemble a core group of experts with modeling experience in different intellectual disciplines ranging from cosmology, biology, organizational behavior to climate modeling as well as philosophers who study inference. We propose this as a pilot project -- one in which we learn about the various kinds of conceptual model building --- agent simulations, ab-initio time evolution modeling of complex non-linear systems like the universe, modeling the propagation of genetic transcription and expression, climate modeling --- to distill how causal effects are extracted from these efforts and how correlation, deeper correlations and causation are disentangled. We plan to interrogate these various frameworks, as case studies to understand and gain insight on how and why they have been effective in inference.