AI's Integration

This paper develops a structural account of artificial intelligence's integration into society organized around six elements: physics, economics, authority, consent, consequence, and responsibility. The central distinction is between configuration-mode systems, which produce contextually appropriate output without accumulating record that constrains subsequent operation, and trajectory-bearing systems, which integrate experience along their path through time. The claim is precise: configuration-mode substrate accumulates irreversible physical modification, but that modification is decoupled from what the system computes. Records exist to the substrate rather than in it. AI as currently deployed is configuration-mode by architecture; the deploying organization is the trajectory-bearing entity. The argument advances in stages. The physics element establishes the distinction, grounds it in three documented cases, and introduces three further axes the analysis uses: graded integration, channel coverage, and the difference between the capability ceiling of a class of system and what a particular deployment attains. The economics element identifies three externality forms — positive externalities through value capture from contributions, negative externalities through cost-deposition on users, and negative externalities routed through host communities — and explains why each takes the form it does under the architectural condition. The economics element then develops a finding that shapes the paper's proposals. When an externality is corrected, the recovered value does not return to the party it was taken from. Repriced insurance accrues margin to insurers; litigation distributes between class members at settlement rates and the firms bringing the action; regulatory penalty moves value to the state. Each mechanism is lawful and defensible on its own terms, and none has a structural channel returning value to the dispersed source, which occupies the residual position in the original extraction and in every mechanism that corrects it, for the same reason in both cases. The burden this places on the opposing view is stated precisely: anyone claiming market self-correction must identify the repatriating mechanism, not merely that extraction was penalized. The element closes by locating the boundary of the extraction pattern through the case it fits least well, on the ground that an account which discriminates is stronger than one that totalizes. The authority element examines AI-mediated institutional decisions, with the autonomous-actor framing operating as discursive cover for routing through configuration-mode entities. The consent element identifies three structural conditions for meaningful consent — epistemic access, meaningful alternative, durable position — that AI deployment systematically violates. The consequence element develops the structural finding at full scale: configuration-mode systems articulate output without bearing the costs the output produces, and the institutional cover function follows. The responsibility element identifies the accountability gap and develops mechanism design proposals, eight mechanisms operating at the deployment-and-surplus level, addressing the four structural conditions the architectural account names. Because correction through adjudication routes around the source even when it succeeds, the mechanisms are specified as direct instruments rather than as an enforcement regime. The paper distinguishes its analytical contribution from existing AI safety, AI ethics, and AI policy literatures. Standard frameworks address each element partially; the integration is what the paper contributes. AI deployment produces real productive value; the analytical task is identifying the structural conditions under which the value flows to specific parties while the costs flow to others, and designing mechanisms that correct