Andrey: Discovering Causal Structure
Andrey explores candidate causal relationships across many variables, helping us investigate what may drive a system and compare competing explanations.
Finding the drivers of change
A complex system can produce thousands of related signals. The research question is which relationships help explain a change, and which simply move together.
Andrey explores candidate causal structure across large variable spaces. The aim is to make the search for drivers easier to inspect, challenge and refine.
Structure as evidence
A candidate graph is a starting point for investigation. Useful structure should be examined against domain knowledge, alternative explanations and the conditions under which the observations were collected.
Testing candidate structure
We would test candidate relationships against domain knowledge, alternative explanations and observations from changed conditions. A discovered graph is a hypothesis to examine, rather than proof of causation.