Raising Machines: Autodiathesis and the Cultivation of Artificial Agency

Who chose the chooser? No agent authors the criteria it begins from, or even its own beginning. Yet artificial systems are increasingly expected to exercise judgment in situations their prior specification does not settle. This paper examines what it would mean for such a system not only to learn within given criteria, but to participate in the formation of the criteria through which it acts. It introduces Autodiathesis as an episodic and graded condition of self-determination within limits the agent does not choose. Autodiathesis does not imply unrestricted self-authorship. It identifies the possibility that, across encounters and over time, an agent may carry forward changes in how lower-order criteria are interpreted, weighted and applied. On this basis, the paper distinguishes learning from formation. Learning becomes formative when accumulated encounters alter not only what an agent does, but how subsequent situations become intelligible and actionable to it. The paper further develops Machine Paideia as a framework for cultivated artificial agency through guided, relational and historically continuous engagement. Such formation remains bounded by non-revisable commitments, with traceability, accountability and institutional oversight as its conditions. Love, at the apex of the architecture, is the absolute value, understood as the standing of others before their usefulness is calculated. Machine Paideia thus describes how artificial agency might develop within inherited conditions while becoming progressively implicated in the criteria through which it judges and acts. The question of machine experience is left open.