Research
My research is currently organized around my PhD thesis which defends one claim: epistemic grounding is a property of a formation history, not of language, and it does not close under composition. In a nutshell, it goes like this:
The instrument is the PRU framework (Pattern-Recognition Unity), a framework I developed during the past couple of years. PRU rests on a principle (that gives it its name) and an analytic tool modeled on neural networks functioning: Pattern-Constellation (PC), noted with capitals in curly brackets: \({X}\). The fundamental idea is that we learn things as complex structures of coherent sensing, feeling and motor patterns, that link together sensing and reaction. A child who has handled fire — felt its heat, been pulled back from it — has something no description gave her, the Pattern-Constellation \({FIRE}\): a multimodal, attractor-like organization sculpted by serial, locus-bound, concern-laden encounter, not a representation. The principle of Pattern and Recognition Unity states that there is no serial cognitive process of sensing followed by some sort of recognition of the sensum based on a template, that generates in turn a specific motor action. Instead sensing and reaction are learned together and activated together. \({X}\) realizes in basic minds what I call basic grounding, the fundamental connection between mind and world.
Language is treated similarly, but with some notable differences: words are learned initially as motor patterns (ways to produce sounds) that are part of the corresponding Pattern-Constellation; I call this mode of using a word, word-object, \(x^o\) (infant pointing to the burning matchstick and saying “fire” is \(<fire^o>\)). With time, we learn to use them abstractly, as concept-words, \(x^c\). Abstract use is developing what I call \(\langle x \rangle\), Linguistic Pattern-Constellation (LPC). The process gives a more analytical form to the Wittgensteinian language games in a practice, \(<fire>\) denotes the totality of acceptable ways that “fire” can enter into a sentence). Different LPCs tend to group together in distinct ways, and we introduce \(\mathcal{R}\), which names linking, the fundamental relation of the linguistic register. Epistemic grounding, \(\mathcal{G}_E\), is basic grounding together with word coordination: \(\mathcal{G}_E(FIRE) \equiv \{FIRE\} + \mathcal\{R\}(fire^c, fire^o)\).
The diagnostic follows. \(\mathcal{G}_E(A) \wedge \mathcal{G}_E(B) \nRightarrow \mathcal{G}_E(A,B)\): grounded parts do not deliver a grounded composite, which requires a formation history of its own. The assumption that grounding propagates through composition is the inheritance fallacy. Encounter presupposes a locus of organization for which differences make a difference. With language it becomes a center, the standpoint from which relevance, memory, belief, and value are possible. (this is what my thesis proposes to investigate)
Four domains test the claim. In artificial intelligence, text-trained models operate inside a detached linguistic register and hold no such center, against teleosemantic accounts of referential grounding (Mollo and Millière 2026) and enactivist claims about machine sense-making (Froese 2026). In rule-following, Carroll’s regress and Kripke’s paradox are the same closure failure, with Abrichtung as its resolution (Wittgenstein 1953, §198). In metaphysics, formative grounding is distinguished from metaphysical dependence (Fine 2012). In enactivism, Pattern-Constellations supply the structural units radical enactivism leaves unspecified (Hutto and Myin 2013).
The method is conceptual analysis with minimal formalism, supported by limited embedding experiments on the inherited-geometry hypothesis.