The growing use of data and machine learning suggests that an automated program could rationally evaluate the impacts of a public policy. Is this kind of assessment of public policy in perspective? Then who would be in charge of training the program and with what legitimacy?
No public policy evaluation realized fully by a machine is in perspective. We need men (probably civil servants, or members of agencies specialized in the evaluation of public policies.
What exists so far, for example in France, is a platform where all the previous evaluations of public policies are gathered (L’observatoire de l’évaluation des politiquespubliques). There are however reasons to think, as you suggest, that a use of AI can be made to make these evaluation easier, faster, but the initiative comes from the central State (general inspections), the state or territories (cour des comptes), and I don’t feel such advanced methods are on the roadmap so far. This is an interesting question. For the experts though, is it the best to ask ?