Без заголовка

Спасибо за пояснение, теперь понятно почему вы упоминаете их часто вместе. Мне кажется, Дойч и Пёрл используют термин "counterfactual" в двух довольно разных смыслах, и это сильно запутывает. Смысл Пёрла кристально ясен (в рамках какой-то причинной модели), он дает конкретные математические выражения. Дойч не использует термин в The Beginning of Infinity ни разу. В первой работе по constructor theory (https://arxiv.org/pdf/1210.7439.pdf), "counterfactual" впервые используется вот в этом абзаце: Causation is widely regarded by philosophers as being at best a useful fiction having no possible role in fundamental science. Hume (1739) argued that we cannot observe causation and therefore can never have evidence of its existence. But here I shall, with Popper (1959, 1963), regard scientific theories as conjectured explanations, not as inferences from evidence, and observation not as a means of validating them, but only of testing them. So Hume’s argument does not apply. Nor does the argument (e.g. by Russell 1913) that the fundamental laws of physics make no reference to causes – for that is merely an attribute of a particular way of formulating those laws (namely, the prevailing conception) not of the laws themselves. Moreover, the prevailing conception itself is not consistent about that issue, for the idea of a universal law is part of it too, and the empirical content of such a law is in what it forbids by way of testable outcomes (Popper 1959, §31 & §35) – in other words in what transformations it denies can be caused to happen, including to measuring instruments in any possible laboratories. Explanatory theories with such counterfactual implications are more fundamental than predictions of what will happen. For example, consider the difference between saying that a purported perpetual motion machine cannot be made to work as claimed ‘because that would violate a conservation law’ and that it won’t work ‘because that axle exerts too small a torque on the wheel’. Both explanations are true, but the former rules out much more, and an inventor who understood only the latter might waste much more time trying to cause the transformation in question by modifying the machine. Из этого абзаца я вывожу, что он вкладывает вот что: counterfactual explanation = explanation (theory) that can generate testable predictions, that is, predictions of what can *not* be caused to happen. Получается, по Дойчу, counterfactual это P(Y|do(X)) = 0. И проверяем вот это do(X) в эксперименте. Что в терминах Пёрла вообще не counterfactual, а intervention, да еще и частный случай (P = 0).