The Epistemic Palimpsest: Knowledge, Expert Authority, and the Architectural Conditions of Public Scientific Judgment

This chapter takes up the question that closed Chapter 5: if the political palimpsest requires an administrative-expert stratum whose deliverances are answerable to the legislative, judicial, and civic-deliberative strata without itself occupying the authorizing function, what architectural conditions in knowledge-production sustain that requirement? It argues that the major traditions of epistemology – Cartesian foundationalism, classical empiricism, the doctrine of sense data, coherentism, naturalized epistemology (Quine), reliabilism (Goldman), knowledge-first (Williamson), virtue epistemology (Sosa), the Popperian-Kuhnian-Lakatosian sequence in philosophy of science, and the contemporary social-epistemological turn (Hardwig, Longino, Fricker, Daston-Galison, Jasanoff, Douglas) – are best read as a sequence of attempts to address the same architectural problem at the epistemic level: how a knowledge-bearing structure can produce claims-to-knowledge without those claims collapsing the gap between the role that issues them and the role that determines their correctness. Read in this light, the regress of justification – the three modes Agrippa already isolates and the trilemma's classical responses of foundationalism, coherentism, and infinitism, together with the intermediate positions (Alston's minimal foundationalism, Haack's foundherentism) that recombine its horns – appears as the propositional form of the same architectural problem, so that the chapter's proposal figures not as a fourth answer to the trilemma but as a refusal of the framing that generates it. The chapter advances three claims. First, the classical figures of epistemic authority – the Cartesian cogito, the empiricist given, the naturalist's reliable process, the coherentist's web of belief, and, in its most articulated form, the reflective tier of Sosa's virtuous circle – fail in the same structural way as the political philosophies diagnosed in Chapter 5: by collapsing into a single role that both generates epistemic commitments and supplies the correctness conditions by which those commitments are assessed. The negative half of this result is shown to be available independently of the resolutions usually paired with it – Sellars's relocation into the space of reasons, Nagel's self-standing reason, and the communitarian completion of Wittgenstein's rule-following considerations – each of which is taken without the positive doctrine that would reinstate the single-role architecture one level up. Second, Popperian falsification is the closest structural analogue in epistemology to Raz's service conception in political philosophy: it locates the authorizing function in a position whose authority is not modally derived from the producing role (the theory) it evaluates, and the architectural reading offered here recovers this insight from the substantive controversies that have surrounded it. Third, the replication crisis in psychology and biomedicine (2011 onward) and the manufactured controversy on climate change are the epistemic analogues of the Joly–Dorpat configuration: the producing manifold of knowledge-production (peer review, journals, empirical disagreement) is preserved while the modal non-derivability of its strata is dissolved – in the first case by institutional incentive structures, in the second by the artificial multiplication of producing positions whose authorizing function is modally fixed by a single external source. Between the diagnosis and the proposal the chapter situates the question of expert authority and democratic legitimacy. The Foucauldian closure, in which a discipline produces its object and supplies the correctness conditions for claims about it, is read as the single-role architecture in its most concentrated form, and its contemporary realization is traced through the platform-mediated regulation of scientific knowledge (Palese's 'Artificial Truth'), the disavowed entanglement of fact-production and a