# User: please familiarise yourself with all the work we have been doing over the ## Retrieval Notes - Session id: `f2eeb495-1162-40aa-8f9d-10244ce5672d` - Last activity: `2026-01-24T12:37:39.791Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User User: please familiarise yourself with all the work we have been doing over the last few days about metaphilosophy User: hello? ### Assistant Let me catch up on recent work. I'll start by reading the last few daily notes and searching for metaphilosophy-related content. ### Assistant I can see significant metaphilosophy work from yesterday. Let me read the key notes and find the relevant session for more context. ### Assistant Good, I have the two main notes. Let me also read the "Philosophical Sparks from Claudian" note mentioned in yesterday's actions, and check for any other related content. ### Assistant Let me also read the "Can We Do Philosophy with AI" note which looks directly relevant: ### Assistant --- I'm now caught up on the metaphilosophy work. Here's what I found: ## Recent Metaphilosophy Work ### Core Notes Created (January 23) 1. **[[Notes/Metaphilosophy Landscape.md]]** — A comprehensive reference note mapping: - The three core questions (nature, method, value of philosophy) - A **10-position spectrum** from Quine ("philosophy is continuous with science") to Rorty ("edifying conversation") - Williamson's abductive methodology and the critique of deductivism - Bengson et al.'s two-stage model and six properties of theoretical understanding - The progress problem and various responses - **Implications for AI**: a table mapping each metaphilosophical position to whether/how AI could "do philosophy" 2. **[[Notes/Theoretical Understanding - Bengson vs Dellsén.md]]** — A detailed comparison of two accounts of understanding: - **Bengson**: 6 hierarchical properties (accuracy, reason-based, robust, illuminating → then orderly, coherent) - **Dellsén**: 2 gradable dimensions (accuracy, comprehensiveness of dependency models) - Key divergence: Bengson requires reasons and explanation; Dellsén allows understanding without explanation - **AI implication**: Neither account creates conceptual friction for AI-enhanced understanding—"reason-based" is a property of the *theory*, not the *producer* 3. **[[Notes/Philosophical Sparks from Claudian.md]]** — Two captured ideas: - Can paradigm-breaking be learned as a formal *pattern* (not just content)? - Verification as internal to understanding → Goodhart's Law might bite less for philosophical RLHF ### Framing Note (January 14) **[[Notes/Can We Do Philosophy with AI - and Should We.md]]** — The Substack-tagged question driving this: Is philosophy more like science (where AI is accepted) or art (where AI is contested)? ### Source Texts The work draws on extracted texts in Attachments/: - `widening_picture_williamson.txt` - `metaphilosophy_full.txt` (Overgaard et al. introduction) - `bengson_methodology_full.txt` - `dellsen_progress.txt` --- The project seems oriented toward arguing that **different metaphilosophical positions have different implications for AI-assisted philosophy**, and that positions emphasizing theoretical virtues (Williamson) or representational accuracy (Dellsén) are more permissive of AI contribution than positions emphasizing lived understanding or transcendental inquiry. Where would you like to pick up? *La questione centrale sembra essere se il valore filosofico risieda nel testo o nel produttore.*