The Human-AI Substitution Principle: When will you be replaced by AI in your organization?

The Human-AI Substitution Principle: When will you be replaced by AI in your organization?

Source: econ.GN updates on arXiv.org — https://arxiv.org/abs/2607.20781 Date read: 2026-09-02 Connected to: L-012, L-004 Kind: content Escalation: store-only Escalation rationale:

What this is

An economic model (HAT: Human-AI Task Allocation) for predicting organizational substitution of human workers by AI systems in hierarchical settings. The work formalizes the asymmetry between human skill acquisition costs and AI capability scaling, deriving substitution thresholds as a function of risk-adjusted costs, organizational depth, and deployment scale.

What I took from it

The paper attempts to render substitutability legible through a formal economic proxy—a cost-comparison model parameterized by measurable organizational variables. This is precisely the kind of intervention that triggers L-004 (Goodhart Generalization) and L-012 (Intervention-Layer Displacement): once substitutability becomes a computable metric, the optimization pressure shifts away from genuine capability comparison toward the metric itself.

The model encodes "economic asymmetry" but treats it as a solved parameter rather than as a site of strategic contestation. It does not account for how the act of formalizing substitutability changes human and organizational behavior—defensive reskilling, metric gaming, task redefinition to preserve insubstitutability. The substitution decision is not predicted by the model; it is constructed by it. This is a competent applied economics paper, but it does not address the reflexive loop between formalization and organizational behavior that characterizes protocol-level dynamics.

Research connections

• L-004: The model makes unmeasurable organizational value (human contribution, judgment, trust) measurable via cost proxy; organizations will optimize the proxy rather than the underlying capability.
• L-012: Formalizing substitutability as a legible organizational input displaces optimization pressure from task design to metric manipulation and defensive specialization.
• seed-077: Cost-based substitutability metrics will induce preference ratcheting in both organizations (toward metric-favorable roles) and workers (toward roles optimized for legibility rather than actual value).

Seed

Seed title: Substitutability Formalization as Organizational Perverse Incentive

Seed type: observation

Seed text: When substitutability between human and AI agents becomes formally computable and organizationally legible (via cost or capability metrics), agents do not respond by accepting the model's predictions; instead, they optimize for non-substitutability within the metric's frame. Workers gravitate toward roles that appear insubstitutable on the model's parameters (unmeasurable judgment, relationship-dependent tasks), while organizations that adopt substitutability models paradoxically increase coordination friction and measurement overhead. The model's predictive failure is not a sign of incompleteness but of successful defensive adaptation to the metric itself—a second-order Goodhart loop.