Tillers wrote:
>I agree that F = MA can be described by "Pearl-style causal diagrams."
AND
> the sort of complex relationships
> that one finds in even simple "real" causal hypotheses such as F = MA.
>
Just for clarification, F = MA is not the causal form that would be represented in a causal diagram, A = F/M is the form. A is caused by F and M. If one measures A (by observation) then one knows (actually infers) the ratio F/M. If some other source of information provides the M, (rsp., the F), then one can infer the F (resp., the M). If one obtains F and M, without observing A, one can predict (infer into the future) the A. Evidence enables inference when the effect is observed,and the inference (of cause) can be refined if a companion "cause" is also observed. But if one observes the causes, then the effect is predicted, which can be viewed as a special case of inference (into the non-negative time axis). In general, the term inference relates to conclusions about the non-positive time axis, taking the time of the latest observation as reference time 0, and assuming earlier measurements are still valid at reference time 0.
Paul