AI-generated derivative output can contaminate the knowledge commons, creating a governance trap that preempts technical decline and narrows the policy window by about nine years; without prior investment in epistemic capacity, standard R&D subsidies are far less effective and inaction costs about 6.8% of consumption.

Main Finding

AI-generated derivative content can “contaminate” the knowledge commons and create two nested, self-reinforcing irreversibilities—a derivative trap and a governance trap. The governance trap (decline in institutional ability to identify frontier knowledge) typically arrives first and closes the policy window for effective intervention—about nine years before the derivative trap becomes binding in the baseline simulation—even while measured innovation remains positive. Competitive markets undervalue frontier knowledge and overvalue derivative output, producing a “double wedge.” Effective policy must follow a strict hierarchy: invest in epistemic capital (to restore/maintain governance capacity) before implementing broader governance measures, which in turn must come before expansionary R&D subsidies. Inaction incurs a large welfare cost (about 6.8% consumption-equivalent in the baseline).

Key Points

• Two nested irreversibilities:
  • Derivative trap: when recombinative AI output pervades the corpus beyond a threshold, frontier productivity degrades faster than firms/agents can reallocate talent or respond via R&D, causing persistent slowdown.
  • Governance trap: institutional capacity to distinguish frontier vs. derivative knowledge (epistemic capital) is itself a depletable stock; once degraded, the system loses the ability to govern knowledge quality and counter contamination.
• Timing: In the baseline, the governance trap occurs roughly nine years before the derivative trap, effectively preempting interventions that would otherwise prevent the derivative trap.
• Measurement risk: Standard innovation metrics can stay positive even after the governance trap begins, masking the impending collapse of effective frontier discovery.
• Market failure structure: a double wedge—frontier knowledge undervalued, derivative output overvalued—creates misaligned private incentives.
• Policy hierarchy: Epistemic investment → governance capacity → R&D subsidies. Epistemic investment is a prerequisite for effective use of other instruments.
• Welfare loss: The model’s baseline inaction generates about 6.8% loss in consumption-equivalent terms relative to the optimal policy path.

Data & Methods

• Model type: Dynamic growth model with endogenous innovation and a knowledge commons that accumulates outputs (human and AI-generated).
• Key mechanisms and stocks:
  • Knowledge corpus: accumulates both frontier and derivative outputs; presence of derivative content reduces signal quality of frontier discoveries.
  • Epistemic capital: institutional capacity to identify, verify, and curate frontier knowledge; modeled as a depletable and investable stock.
  • Talent allocation and R&D: agents allocate labor between production, R&D, and epistemic activities; R&D productivity depends on the quality of the corpus and on epistemic capital.
• Irreversibilities:
  • Derivative trap characterized by threshold crossing in the composition of the corpus; beyond threshold, externalities cause persistent productivity loss.
  • Governance trap arises because epistemic capital declines with contamination and is costly/time-consuming to rebuild.
• Calibration and simulation:
  • Parameters calibrated to match stylized features of innovation, R&D responsiveness, and plausible timescales for institutional change (details unspecified in brief).
  • Baseline numerical simulation traces evolution of corpus composition, epistemic capital, measured innovation, and welfare under competitive equilibrium vs. optimal policy.
  • Policy experiments evaluate timing and ordering of interventions (epistemic investment, governance measures, R&D subsidies) and compute consumption-equivalent welfare differences.
• Diagnostics reported:
  • Timing gap (~9 years) between governance and derivative traps.
  • Welfare cost of inaction (6.8% consumption equivalent).
  • Comparative statics showing the strict instrument hierarchy and the persistence of the double wedge in equilibrium.

Implications for AI Economics

• Prioritize building epistemic capacity: Investments in provenance, verification, curation, and institutions that can reliably identify frontier contributions should be treated as first-order economic policy tools for AI-era growth.
• Timing matters: There is a limited window for effective policy. Measured innovation may not reveal an impending governance collapse; hence early action is critical.
• R&D subsidies alone are insufficient and potentially wasteful if governance capacity is weak: Subsidizing R&D without restoring epistemic capital risks amplifying derivative content and exacerbating mispricing of frontier knowledge.
• Regulatory sequencing: Policy design should follow the instrument hierarchy—epistemic investment first, then governance reforms (standards, mandates, infrastructures for provenance), then standard innovation-support policies.
• Measurement and monitoring: Develop leading indicators of epistemic capital and corpus contamination (e.g., provenance coverage, retrievability of primary sources, verification rates) because headline innovation statistics can be misleading.
• Welfare stakes justify proactive public intervention: The modeled 6.8% consumption-equivalent loss implies substantial long-run costs from inaction, warranting public investment in epistemic institutions and governance infrastructure.
• Research agenda: Empirical work to measure epistemic capital, quantify contamination thresholds, and test calibration assumptions; policy experiments (pilot programs for provenance, curation markets, verification standards) to validate model prescriptions.