On AI and ML

ML algorithms that learn DAG structure do so by analyzing a single dataset that has all relevant data measured together, an approach only realistic in settings such as manufacturing and other well-controlled, process-oriented applications. Causal explanation in complex, uncontrolled, real-world domains requires the ability to integrate evidence and findings from many separate experiments conducted in a piecemeal fashion, asynchronously over a period of years, and with varying levels of quality. Insights garnered from a tool such as CausaLens might be considered to be but one piece of the broader causal puzzle that Pallium allows users to assemble.