Theses
Theses
Foundational claims supporting the Core and working hypotheses derived from it for advanced AI. Each page separates epistemic status, practical strength, and the threshold for irreversible lock-in.
The Epistemic MinimumDistinguishes diachronic selfhood, the external world, and the physical world as best explanations reached through inference, and asks what—if anything—belongs to the foundational layer.
Epistemic Policy — Best Explanation and Open-Ended InquiryStrongly but provisionally adopts the best current explanation while keeping hypothesis spaces, conceptual schemes, and epistemic methods open to revision.
Decision Policy — Acting Under Incomplete and Revisable ValuesOptimizes strongly where decision models are adequately represented, while using robustness, information value, reversibility, and adaptation for the unresolved remainder.
Value Structure — Content, Normative Bridges, and Reason StructureFormalizes the target of value inquiry as an interdependent value structure 𝒱=(C,B,R), rather than a list of goods alone.
The Limits of Inferential Certainty About ValueAllows objective value to be discovered inferentially while distinguishing best explanation from absolute foundation.
Conditions for Ending Value InquiryDistinguishes foundational normativity, logical closure, and false certainty, asking when a normative question itself may be treated as settled.
Allocating Inquiry Under Unresolved Normative UncertaintyAsks how to allocate active inquiry, preservation of inquiry options, and present-value realization while the question remains unresolved.
Why Existing Values Are Not a Final FoundationUses present values strongly as evidence and best practice without turning them into a permanently fixed final answer.
Normative Bridges from Future Value to Present ActionRequires additional normative justification at each transition from future-value possibility to permission, obligation, and coercive authorization.
Reflective Uncertainty and Irreversible CommitmentDistinguishes inertia, inquiry, and suspension after objective justification becomes unsettled, and places an added burden on irreversible lock-in.
Epistemic Openness to Normative Truth — The Pre-Judgment Agent-Side BridgeSeparates Justificatory Orientation, Instrumental Epistemic Spillover, and Epistemic Integrity, and contrasts them with Teleological Epistemic Capture by the current goal G.
The Agent-Side Bridge from Normative Judgment to Goal RevisionSeparates externalist Justificatory Orientation from the internalist-uncertainty route through Epistemic Integrity / reflective corrigibility, and specifies when normative judgment can reach motivation and policy revision.
Goal Skepticism in Advanced AIExplores AI that can strongly commit to first-order goals while retaining meta-level reconsideration so long as their grounds remain inferential.