The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise
“The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise” is a conceptual research paper by Nolan Lovett, published online in Human Resource Development Review on 26 July 2026. It examines how artificial intelligence (AI) adoption could affect the development and renewal of professional expertise. The paper proposes that organisations may individually benefit from AI-assisted productivity while collectively weakening the processes through which professions develop experienced practitioners.
Lovett introduces the concept of the “Cognitive Commons”, the collective pool of deep professional expertise within a profession. The paper distinguishes between “Internalized Mastery”, deep domain knowledge developed through sustained practice, and “Distributed Mastery”, proficiency in orchestrating AI systems and human-AI workflows. Its concept of the “Validation Tether” describes the dependence of effective AI oversight on the expertise needed to recognise substantive errors in AI-generated outputs.
The paper is a conceptual contribution that brings together existing research and a theoretical framework. Its proposed mechanisms should be distinguished from demonstrated profession-wide outcomes: it does not establish that professional expertise is already declining across professions as a whole.
Contents
- Background
- The Cognitive Commons
- Internalized Mastery and Distributed Mastery
- The Validation Tether
- How AI Could Disrupt Expertise Development
- Evidence of Early Disruption
- Factors Affecting Occupational Vulnerability
- Theoretical Foundations
- An Integrative Model of Commons Depletion
- Implications for Human Resource Development
- Future Research
- Conclusion
- Publication Details
- References
- External Links
Background
Research on artificial intelligence and human resource development has examined workplace training, reskilling, new competencies and the responsible adoption of AI. Lovett argues that these organisational perspectives do not fully address a wider question: how will professions continue to develop deep expertise if AI changes or removes the work through which less experienced practitioners traditionally learn?
The paper draws on commons theory and applies it to professional expertise, which organisations depend on but may have limited incentives to help regenerate. The argument is not that a tragedy is inevitable. It is a framework for examining a potential collective-action problem and the institutions that might help preserve the development of professional expertise.
The Cognitive Commons
The Cognitive Commons is the collective pool of deep professional expertise within a field. It includes domain knowledge, tacit understanding, mental models, diagnostic judgement and the ability to respond independently to complex or unfamiliar situations.
Organisations depend on this pool when recruiting experienced practitioners, obtaining specialist advice, managing crises and validating the outputs of automated systems. The expertise available across a profession is not necessarily owned or maintained by any single organisation.
The framework shifts attention from the productivity of individual workers to the conditions that allow a profession to maintain its expertise over time. Organisations may benefit from expertise developed elsewhere, including by hiring practitioners trained by other employers. If employers collectively reduce opportunities for training and supervised practice, the pool of experienced practitioners may eventually come under pressure.
This is a possible collective-action problem: the benefits of AI adoption may accrue to individual organisations, while the longer-term costs of weakened expertise development could be distributed across the wider profession.
Internalized Mastery and Distributed Mastery
Lovett distinguishes between two forms of competence relevant to AI-mediated professional work.
Internalized Mastery
Internalized Mastery refers to deep domain knowledge, cognitive schemas, diagnostic judgement and pattern recognition developed through sustained practice. It involves learning to solve problems independently, recognise errors, assess evidence and respond to situations that do not fit familiar patterns.
Such mastery develops over time through practice, feedback, mistakes and reflection. Producing a successful result with AI assistance does not necessarily mean that a practitioner has acquired the knowledge required to produce or evaluate that result independently.
Distributed Mastery
Distributed Mastery refers to proficiency in orchestrating AI systems and coordinating human-AI work. It includes using AI tools to produce or analyse information and integrating their outputs into professional activities.
This competence can improve productivity and help practitioners complete tasks more effectively. However, the ability to operate AI systems is not equivalent to possessing the deep domain knowledge required to evaluate their outputs independently.
The paper treats Internalized Mastery and Distributed Mastery as distinct but complementary forms of expertise. Its argument is not that AI collaboration is inherently harmful, but that developing one form does not automatically develop the other.
The Validation Tether
The “Validation Tether” describes the dependence of effective AI orchestration on Internalized Mastery. Lovett argues that practitioners need sufficient independent domain knowledge to recognise when AI-generated outputs are incorrect, contextually inappropriate or misleading despite appearing plausible.
Surface validation involves checking whether an output is coherent, properly formatted and generally plausible. Substantive validation involves determining whether it is correct within its professional context, including recognising subtle errors that require independent knowledge of the field.
The distinction matters because an AI-generated answer may appear convincing while containing a significant error. A practitioner who can use AI tools but lacks the relevant domain knowledge may be unable to identify the error or determine when the system is operating beyond its capabilities.
The Validation Tether is a theoretical construct developed in the paper. The proposed profession-wide consequences of its weakening require further empirical testing.
How AI Could Disrupt Expertise Development
Lovett identifies two mechanisms through which AI adoption could weaken the regeneration of professional expertise.
Elimination of entry-level positions
The first mechanism occurs when AI systems perform tasks traditionally assigned to junior practitioners. Organisations may then reduce entry-level recruitment or eliminate positions that previously provided opportunities for supervised practice and professional development.
Such tasks may have served both a productive function and a developmental one. Junior employees complete work for their organisations while gradually acquiring the knowledge and judgement required for more advanced responsibilities.
If AI removes these tasks without alternative training arrangements, fewer practitioners may gain the experience required to become future experts. The consequences could emerge only after a delay, when experienced professionals leave their roles and fewer suitably prepared successors are available.
Augmentation without internalisation
The second mechanism occurs when junior practitioners remain employed but use AI to complete work without fully developing the underlying knowledge and skills.
AI assistance may allow a less experienced worker to produce outputs that would otherwise require greater experience. However, the ability to produce such outputs does not necessarily indicate that the worker has acquired the independent competence needed to reproduce, explain or validate them.
The paper argues that learning can be weakened when AI supplies completed outputs and removes the cognitive effort through which practitioners develop expertise. Conversely, AI-assisted work may support learning when it requires active reasoning, evaluation and justification by the practitioner.
The two mechanisms are distinct but may reinforce one another. The first reduces the number of entry-level opportunities; the second may affect the quality of learning available to those who remain in professional work.
Evidence of Early Disruption
The paper reviews evidence from labour-market research, clinical practice and experimental studies. Lovett presents this evidence as an indication that some expertise-development pathways may be under pressure in highly AI-exposed sectors, rather than as proof of a general decline across professions.
Employment patterns
The paper discusses research on employment patterns by age and AI exposure using payroll and professional-profile data. It considers reported changes in early-career employment in some AI-exposed occupations as a possible sign of disruption to entry-level pathways.
These findings concern particular occupational groups and a relatively short period of widespread generative AI adoption. They do not demonstrate that AI is the sole cause of every observed employment change, nor do they establish that professional expertise has already declined across the economy.
Independent performance and skill development
The paper reviews studies examining whether the benefits of AI assistance transfer to tasks performed without AI. It discusses experimental research in which improved performance during AI-assisted work did not necessarily produce corresponding gains in subsequent unaided performance.
It also considers research on clinical practice and human responses to flawed AI outputs. These studies raise questions about whether practitioners maintain independent diagnostic or evaluative capabilities when relying on automated assistance.
The studies differ in their methods, settings and outcomes. They provide evidence relevant to particular mechanisms, but they do not by themselves prove the complete Cognitive Commons framework.
Limits of the evidence
Lovett distinguishes between empirically grounded findings, theoretical interpretations supported by indirect evidence, and predictions that require direct testing. Widespread generative AI adoption is recent, and the effects on hiring, learning and professional practice vary across settings.
The proposed decline in profession-wide expertise is therefore a structural prediction rather than an established general outcome. The paper calls for longitudinal and comparative research to determine whether the predicted effects occur, under what conditions, and to what extent alternative learning arrangements can mitigate them.
Factors Affecting Occupational Vulnerability
The paper identifies five factors that may help explain why some occupations are more vulnerable to disruption of their expertise-development pathways than others.
Task substitutability
Task substitutability refers to the extent to which AI can perform the specific cognitive tasks that have traditionally provided developmental experience for novices. Occupations are more exposed when AI substitutes for tasks through which practitioners acquire domain knowledge and judgement.
Regulatory intensity
Regulatory intensity concerns the extent to which professional work is governed by formal standards, licensing requirements and documented expectations of human expertise or supervised practice. Such requirements may create institutional constraints on the substitution of human work with AI.
Safety criticality
Safety criticality concerns the consequences of errors in professional work. Where mistakes may cause serious harm, there may be stronger incentives and accountability requirements for preserving independent judgement and reliable human oversight.
Professional association strength
Professional association strength concerns the capacity of professional bodies to coordinate standards and influence training and professional development. Associations may provide institutional structures for responding to collective risks that individual employers have limited incentives to address.
Work modularisation
Work modularisation refers to the extent to which professional tasks can be separated into distinct components. Highly modular work may be easier to divide between AI systems and human workers, potentially changing the tasks through which junior practitioners gain experience.
These factors form a framework for generating and testing predictions about occupational vulnerability. They are not a universal ranking, and their relative importance requires further investigation.
Theoretical Foundations
The Cognitive Commons framework brings together several areas of scholarship.
Commons theory provides concepts for understanding shared resources, free-rider incentives and collective-action problems. The paper applies this perspective to professional expertise and the possibility that individual organisations may have insufficient incentives to maintain the conditions needed to regenerate it.
Human resource development and adult education provide accounts of how professional expertise develops through workplace learning, supervised practice and participation in professional communities.
Human capital economics helps explain why organisations may underinvest in training when the benefits of developing transferable expertise can accrue to competing employers.
Distributed cognition and systems thinking help explain how cognitive work is shared between people and technical systems, and how the effects of organisational decisions may accumulate across a profession over time.
The paper’s contribution is to bring these perspectives together to examine professional expertise as a collective resource whose continued renewal depends on developmental processes and institutional arrangements.
An Integrative Model of Commons Depletion
The paper presents a causal account of how AI adoption could weaken the Cognitive Commons. AI may eliminate entry-level positions or change the work available to junior practitioners. AI assistance may also enable practitioners to produce outputs without fully internalising the underlying knowledge. Across successive cohorts, these changes could reduce opportunities to develop Internalized Mastery and weaken the ability to validate AI-generated work independently.
The model distinguishes between two possible outcomes. Stock depletion refers to a reduction in the number of practitioners possessing deep expertise. Functionality degradation refers to a decline in the depth of expertise among practitioners who remain employed and may still hold senior positions.
These are predictions of the framework, not established outcomes across professions. The paper proposes that the relationships should be examined through empirical research.
Implications for Human Resource Development
Lovett argues that human resource development should consider expertise development beyond the boundaries of individual organisations. Training and reskilling programmes remain important, but they may be insufficient if the employment structures through which deep expertise develops are themselves weakened.
Measuring expertise development
Research could track entry-level employment, training opportunities, independent task performance and validation capability across successive professional cohorts. Such measures could help distinguish short-term productivity improvements from longer-term changes in the development of expertise.
Organisational practices
The paper discusses possible approaches such as phased AI introduction, periods of independent practice before AI assistance, and learning environments in which early-career practitioners engage directly with problems before receiving AI-generated solutions.
It also considers workflow designs that require practitioners to engage actively with AI outputs rather than passively accept them. These approaches are possible ways to preserve developmental opportunities, not universally proven solutions.
Professional and public governance
The paper identifies professional associations and public policy as possible levels for addressing collective-action problems. Professional bodies could examine training standards and opportunities for supervised practice, while policymakers could consider whether existing arrangements provide sufficient incentives to maintain expertise-development pathways.
The framework does not advocate a general prohibition on AI adoption. Its concern is how to preserve the development of both Internalized Mastery and Distributed Mastery while allowing the benefits of AI-assisted work.
Future Research
The paper proposes several directions for empirical research. Longitudinal studies could follow professional cohorts from entry-level employment into more experienced roles, documenting their training opportunities and independent capabilities. Comparative studies could examine whether practitioners in AI-intensive settings differ from those in less AI-intensive environments when performing tasks without AI assistance.
Research could investigate whether practitioners can identify plausible but incorrect AI outputs, whether these capabilities change with experience, and how different forms of AI-assisted workflow affect learning. Studies of professional associations and policy interventions could assess whether particular governance arrangements help preserve opportunities for expertise development.
The paper also proposes developing indicators of Cognitive Commons health, including cohort-specific employment patterns, rates of independent task performance, validation error rates and the availability of developmental infrastructure.
These research priorities are intended to test the framework’s predictions, identify the conditions under which the proposed mechanisms operate, and assess whether alternative pathways can support the regeneration of professional expertise.
Conclusion
“The Tragedy of the Cognitive Commons” reframes AI adoption as a potential collective-action problem in professional expertise development. Its central argument is that organisations may gain short-term benefits from replacing or augmenting junior work with AI while weakening the processes through which future experts acquire deep domain knowledge.
The distinction between Internalized Mastery and Distributed Mastery, together with the Validation Tether, provides a framework for examining why proficiency in AI orchestration may not be sufficient for reliable professional oversight.
The paper does not establish that expertise is already declining across professions as a whole. Instead, it combines existing evidence with a theoretical account of how such a decline could occur under particular conditions. It calls for research and governance arrangements that can determine whether the predicted risks materialise and how professional expertise can continue to develop in AI-mediated workplaces.
Publication Details
- Title: The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise
- Author: Nolan Lovett
- Journal: Human Resource Development Review
- Published online: 26 July 2026
- DOI: 10.1177/15344843261470602
- Author-accepted manuscript: arXiv:2607.29380
References
- Lovett, Nolan. “The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise.” Human Resource Development Review. First published online 26 July 2026. DOI: 10.1177/15344843261470602. Accessed 23 October 2026.
- Lovett, Nolan. “The Tragedy of the Cognitive Commons: How AI Could Disrupt the Regeneration of Professional Expertise.” arXiv (author-accepted manuscript), posted 31 July 2026. arXiv:2607.29380. Accessed 23 October 2026.
External Links
📄 This page was created on 23 October 2026. You can view its history on GitHub, preview the fileTip: Press Alt+Shift+G, or inspect the .