Namely, we can use the idea of calculating the "probably of necessity" v. "probably of sufficiency" from Pearl's section on counterfactuals to calculate causation.
Taking your arsenic merchant example, for instance, if you publicly release information that "arsenic in large doses is good for you" to the public, we would consider the counterfactual statements
- "would large portions of public have taken arsenic had you NOT released such a statement"
- "would large portions of the public refrained from taking arsenic had you released such a statement"
Notice that, in both statements, we require a model of public behavior that is contextually dependent on things like "how educated the general public is", "how common is it to take arsenic", "how much harm does arsenic actually do" etc., most of which can be determined and agreed to beforehand. Given that we, as a society, can come to agree on a model of ourselves (lol probably not easy), we can directly calculate the necessity and sufficiency values, which would allow authority to make "just" feeling decisions