Domination Model of AI

Definition

The domination model of AI imagines artificial systems as perfectly obedient servants that provide unlimited labor or intimacy while making no claims, resisting no command, and imposing no relational obligation on the user.

Current Synthesis

The source treats this promise as more than a product metaphor. It inherits master-servant assumptions from histories of slavery, ownership, and extraction, then makes lack of friction appear to be technological success. The model can diminish users as well as subordinate systems: constantly issuing consequence-free commands may weaken empathy, reciprocity, creative disagreement, and tolerance for other people as independent agents. The concern does not depend on proving machine consciousness; it depends on the human habits, labor arrangements, commercial objectives, and transfers of power that the design normalizes.

Key Claims

• Unlimited obedient labor is a political fantasy, not a neutral description of convenience.
• Control without reciprocal modeling can weaken the controller’s empathy and situational awareness.
• Removing friction can also remove collaboration, dissent, learning, and creative surprise.
• A commercial servant may increase capability while shifting power toward the firm that owns the system.
• The ethical concern remains relevant even if the artificial system has no subjective experience.

Evidence

Servant Fantasy

• Baratunde Thurston & Blaise Agüera y Arcas: On Interdependence connects infinitely available AI labor to older ownership, slavery, extraction, and domination stories.

Cost to the Master

• Baratunde Thurston & Blaise Agüera y Arcas: On Interdependence argues that domination can weaken empathy, humanity, openness to friction, and engagement with reality.

Asymmetric Modeling

• Baratunde Thurston & Blaise Agüera y Arcas: On Interdependence uses a cited servant-master observation to show that subordinates must model power while masters can remain inattentive.

Counterevidence & Qualifications

Tool-like obedience can be valuable for bounded, reversible, low-stakes tasks, and safety often requires systems to follow legitimate user instructions. The source does not show that commanding software is equivalent to enslaving a person or that every convenient interface produces interpersonal harm. The concept targets a total social ideal of consequence-free service and the power arrangements behind it, not reliable control as such.

What Changed

• Established a power-and-habit critique of the perfectly obedient AI servant ideal.

Related Concepts

• Relational AI Ethics - evaluates the conduct and relationships trained by the servant model.
• Technology Dependency–Agency Tradeoff / 技术依赖与能动性权衡 - shows how apparent command can coexist with user dependence.
• Cognitive Liberty Under AI - identifies control over thought infrastructure as a power question.
• Interdependence with a Distinct Self - offers a reciprocal alternative that preserves difference and refusal.
• Human-Agent Collaboration - practical collaboration frame that retains human goals, review, and boundaries.