The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise

Synthesis: This conceptual paper (theory development in HRD) introduces the Cognitive Commons framework, integrating commons theory, HRD scholarship, and distributed cognition to explain how individually rational AI adoption decisions can collectively deplete the shared expertise pool professions require for renewal. It distinguishes Internalized Mastery (deep domain knowledge built through sustained practice) from Distributed Mastery (orchestrating human–AI systems) and develops the Validation Tether: effective AI oversight depends on the very expertise that AI adoption may undermine. Early labor-market and clinical evidence suggests disruption to expertise-regeneration pathways in highly AI-exposed sectors, five factors determine occupational vulnerability, and AI Governance may form at organizational, professional-association, and policy levels.

Key Findings

1. The Cognitive Commons as a collective-action problem. Generative AI creates a parallel to Hardin's tragedy of the commons: professions share a collective pool of deep expertise that no single organization owns, and rational individual decisions to eliminate entry-level developmental positions can deplete it.
2. Internalized vs. Distributed Mastery. AI-era work demands two forms of expertise — Internalized Mastery (deep domain knowledge concentrated in individual minds via deliberate practice) and Distributed Mastery (fluency in orchestrating intelligence distributed across human–AI systems). They serve complementary functions, but commons depletion jeopardizes both.
3. The Validation Tether. Effective oversight of AI outputs — recognizing domain errors, inappropriate recommendations, and subtle flaws — depends on the deep internalized expertise that AI adoption undermines. Distributed Mastery is not a substitute for Internalized Mastery; it presupposes it.
4. Three defining commons characteristics. The cognitive commons is collectively dependent (a public-goods problem), non-exclusive (free-rider incentives), and degradable through overexploitation of the regeneration mechanism — where depletion manifests through generational employment patterns, not immediate scarcity.
5. Early empirical signals. In the most AI-exposed US occupations, workers aged 22–25 saw a 16% relative employment decline (Oct 2022–Sep 2025) while workers 35–49 grew over 8% (Brynjolfsson et al. 2025), reinforced by Hampole et al.'s LinkedIn analysis. Two mechanisms drive this: direct position elimination, and augmentation-without-internalization (junior workers skip the cognitive struggle that builds mastery).
6. Five vulnerability factors. Task substitutability, regulatory intensity, safety criticality, professional-association strength, and work modularization determine an occupation's commons vulnerability — with software engineering, financial analysis, and legal research highest, and medicine/engineering slower due to regulatory and safety counterpressures.
7. Assisted performance ≠ independent capability. AI-assisted productivity does not transfer to unassisted performance (Wiles et al. 2024); heavy AI reliance reduces independent diagnostic accuracy (Budzyń et al. 2025; Natali et al. 2025); and validation habits are eroding — 40% of employees received flawed AI content, 60% do not routinely check AI accuracy.

The Cognitive Commons construct

The Cognitive Commons is the profession-level pool of deep human expertise — internalized domain knowledge, tacit understanding, robust mental models, and judgment capable of validating AI outputs and adapting to novel situations. It differs from adjacent constructs: human capital locates expertise in persons/firms; communities of practice describe learning processes; the commons describes the resulting stock and its regeneration. The construct directs attention to a profession-level stock-and-regeneration dynamic none of these terms make visible.

Commons dynamics apply when three conditions co-occur: AI performs tasks that historically served as developmental contexts for novices; those pathways are hard to replicate outside work; and the profession exhibits the three commons characteristics. The appropriate unit of analysis is the occupation, not the national economy — software engineering differs from plumbing, and even backend vs. frontend development can differ.

Why rational organizations deplete the commons

The overgrazing mechanism operates as a negative externality: each organization captures private efficiency gains while the cost of reduced expertise regeneration distributes across all who depend on collective expert availability. Organizations free-ride by planning to "hire experienced workers when needed" — but those experts exist only because other organizations made developmental investments 5–20 years earlier. The time delay obscures depletion: positions eliminated in 2023 won't manifest as experienced-worker shortages until 2030–2045.

The Human Reserve Paradox amplifies this: organizations need expertise held in reserve for validation, crisis, and novel situations, but lack incentive to maintain it when costs fall individually while benefits spread across the ecosystem. Traditional economic models cannot price the cognitive commons.

AI also transforms classical under-investment in general human capital into a regeneration challenge: prior equilibrium was not governance but accident — organizations maintained pipelines because they needed entry-level workers to do entry-level work. AI removes that hidden governance mechanism.

What this means for practice

• Policymakers. Treat entry-level professional roles as regeneration infrastructure rather than as legacy cost: start building AI Governance mechanisms — boundary definition, monitoring, graduated sanctions, collective choice — in the sectors where they are currently largely absent, before the projected 2030-2045 shortage window.
• Policymakers. Set priorities by occupation vulnerability — task substitutability, regulatory intensity, safety criticality, professional-association strength, and work modularization — beginning with software engineering, financial analysis, and legal research, where the authors see the highest exposure.
• Researchers. Measure the second mechanism directly: augmentation without internalization is invisible in the employment data (Brynjolfsson et al. 2025 captures position elimination, not skipped cognitive struggle), so pair labor-market series with qualitative studies of junior workers' developmental experiences.
• Researchers. Operationalize Internalized and Distributed Mastery and the Validation Tether so the framework's claims become falsifiable predictions rather than illustrations, and connect occupational Workplace Learning and Lifelong Learning outcomes to the individual-level Cognitive Offloading and skill-decay findings.
• Researchers. Reframe expertise development as collective stewardship in the theory-building work on AI and expertise: the framework is offered as a falsifiable account of where current incentives lead, not as a claim that depletion is inevitable — Ostrom's cases show shared resources can be sustained.

Limitations

• Conceptual paper with no primary data: the empirical case is a synthesis of other studies (Brynjolfsson et al. 2025's 16% relative employment decline for ages 22-25, Hampole et al., Budzyń et al., Wiles et al.), so the Cognitive Commons itself is never measured in any profession.
• The authors' own stated limits: the labor-market data span less than three years of widespread generative AI adoption and represent early-stage rather than mature dynamics; the patterns concentrate in the most AI-exposed occupations and do not characterize professional employment broadly; and the evidence does not demonstrate widespread validation failures or profession-wide expertise collapse.
• The framework assumes a particular cultural model of professional formation — expertise built through entry-level employment in hierarchical organizations and governed by formal credentialing — which the authors say "does not exhaust the ways human expertise has been or can be transmitted"; apprenticeship, guild, and community-based arrangements may show different commons dynamics.
• The second depletion mechanism (augmentation without internalization) is undetectable in the employment data the framework leans on, and the qualitative evidence that would test it does not yet exist.

Citation

Lovett, N. (2026). The tragedy of the cognitive commons: How AI could disrupt the regeneration of professional expertise. Human Resource Development Review, advance online publication.