Do _all_ or even _most_ of today's scientists suffer from an illusion that "the supposition that science is grounded in the plenitude of a concrete and lived experience"? In any case, I think, unlike Foucault, that you ought to distinguish the ideology of scientism from the truth work of science, unless you think there is no difference, for instance, between the theory of evolution and creationism.
Also, to dismiss, as Foucault does, as an illusion the idea that "the referent itself contains the law of the scientific object" without explaining what he thinks of as the true relationships between scientific laws and referents won't do, unless you want to ditch causality (and ideas of it) altogether, which noone is capable of doing.
Roy Bhaskar writes in _Plato Etc._:
***** ...To say that this acid turns litmus paper red or that this metal conducts electricity because all do is hardly explanatory. It merely redescribes the phenomena in generalizing it [sic]. This is to leave aside the question as to how we know the generalization is true, relevant and sufficient. Transcendental idealists and contextualists emphasize that what is required for a genuine explanation is the introduction into the explanatory context of new concepts and ideas, not already (explicitly or implicitly) contained in the explanandum, such as _models_ picturing plausible generative mechanisms for the production of the phenomena concerned. ...Realists go further than 'model theorists'...by allowing that, under certain conditions, these explanatory concepts or models can come to be known to denote newly identified deeper, more basic, inclusive or encompassing levels of reality. ...On this _vertical_ existential realism, science is seen as a continuous and reiterated _process_ in motion from manifest phenomena, via creative modelling by what the French philosopher Gaston Bachelard called 'scientific loans' and experimentation..., to their generative structures, which now become the new phenomena to be explained.
...If one is a vertical realist, it is natural to go on to ask: If explanatory structures _exist_ independently of their human identification, do they not _act_ so? That is to extend one's realism to _causality_, in what one could think of as a _horizontal_ direction. (In fact vertical realism presupposes horizontal realism both ontologically, insofar as things are constituted by their causal powers, and epistemologically, in that they are identified by their direct or indirect causal effects.) This is the move that transcendental realists make. The question they pose is: Is P1, the principle of empirical invariance, which transcendental idealists as well as classical empiricists accept, in fact true? The answer is: no. For empirical regularities are in general only forthcoming under experimentally, or otherwise locally..._closed_ contexts. It is only therein that the actualist formula 'whenever x, then y' applies. Conversely, the significance of experimental activity is precisely that, in _disrupting_ the course of nature (including any conjunction) that would otherwise have prevailed, we gain access to explanatory structures, generative mechanisms and laws which continue to operate, but _transfactually_ (transphenomenally, transsituationally, translocally), outside the experimental contexts in which they are identified. If one still wishes to maintain the Humean theory that laws are constant conjunctions (a position I have dubbed 'strong actualism'), then one will have to concede that there are no laws known to science. If, on the other hand, one holds them to be empirical but restrict them to closed systems ('weak actualism'), then one is left without a rationale either for experimental activity or for the practical applied diagnostic and exploratory work of science. That is to say, laws can be universal _or_ empirical (or more broadly, actual) but not both.
For transcendental realism, by contrast, the point of experimental activity is to identify the relatively enduring structures of nature and their characteristic ways of acting. Such structures may be classified into _natural kinds_, possessing causal powers, which, when triggered or released, act, as generative mechanisms, with _natural necessity_ and _universality_ (within their range) so as to codetermine the manifest phenomena of the world, which occur for the most part in open systems: that is, where constant conjunctions do not pertain.... The logical form of a law of nature is given by the concept of a transfactually efficacious _tendency_, which may be possessed without exercised, exercised without being actualized, and actualized without being empirically identified by human beings.
...What grounds induction, retroduction and falsification in science is _ontological stratification_. This is a condition of the possibility of science in general. This is entailed by the argument, from experimental activity, which demonstrates that science presupposes existentially independent and transfactually efficacious laws such that Dr > Da > De where Dr = the domain of the real, Da = the domain of the actual and De = the domain of the empirical. Science may be justified...by its causal efficacy and explanatory power. But a moment's reflection will show that existential independence and transfactual efficacy, and hence ontological stratification, is a condition of the possibility of the most mundane activities from making a pot of coffee to passing a football (both of which presuppose a structured, law-governed, not just empirical, world). That said, its corollary must be drawn out. Induction is _not justified_ at the level of the empirical or the actual. The surface course of nature is very far from uniform. (21-3) *****
While Foucault cannot explain (nor does he care to) the causal efficacy and explanatory power of science, Bhaskar's transcendental realism can. Foucault's error is that he felt he could dispense with the distinction between ideology and truth, ideology and science.
Yoshie