Danxia Xie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Danxia Xie has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Beibei Li
Heinz College of Information Systems and Public Policy, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
15 shared publications - Qingquan Tony Zhang
University of Illinois Urbana-Champaign
15 shared publications - Zehao Zhou8 shared publications
- Xiangguo Zhang
Tsinghua University - Institute of Economics; City, University of London
4 shared publications Buyuan Yang
2 shared publications- Buyuan Yang
Institute of Economics, School of Social Sciences, Tsinghua University, Beijing, China
2 shared publications
Latest Publications
- Open Access
- 6 October 2026
- Published by Elsevier in SSRN Electronic Journal
Preserving and Disseminating Ideas: The Information Carrier Trap and the Rise of the Industrial Revolution
We develop a two-tier endogenous growth model in which information carrier technologies---media that preserve and disseminate knowledge---and production technologies evolve endogenously. Information carriers determine how effectively nonrival ideas are used in production-technology innovation. A low level of information carrier technology creates an information carrier trap: knowledge loss and dissemination frictions depress returns to innovation and sustain stagnation. An illustrative numerical exercise for Britain over 1740--1850 generates a transition path in which R&D labor shifts from information carrier innovation to production innovation, with an economic takeoff in the 1780s, broadly consistent with the timing of the First Industrial Revolution. The model also predicts declining prices of information carriers and faster early growth in information carrier technology than in production technology, both broadly consistent with historical evidence. Along the asymptotic balanced growth path, however, production technology and per capita consumption grow faster than information carrier technology, while the shadow value of a new carrier variety declines at a constant rate. The decentralized equilibrium underallocates research labor to both innovation sectors relative to the Pareto optimum, and the corrective subsidy rate is higher for information carrier R&D than for production R&D.
- 6 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPreserving and Disseminating Ideas: The Information Carrier Trap and the Rise of the Industrial Revolution
We develop a two-tier endogenous growth model in which information carrier technologies---media that preserve and disseminate knowledge---and production technologies evolve endogenously. Information carriers determine how effectively nonrival ideas are used in production-technology innovation. A low level of information carrier technology creates an information carrier trap: knowledge loss and dissemination frictions depress returns to innovation and sustain stagnation. An illustrative numerical exercise for Britain over 1740--1850 generates a transition path in which R&D labor shifts from information carrier innovation to production innovation, with an economic takeoff in the 1780s, broadly consistent with the timing of the First Industrial Revolution. The model also predicts declining prices of information carriers and faster early growth in information carrier technology than in production technology, both broadly consistent with historical evidence. Along the asymptotic balanced growth path, however, production technology and per capita consumption grow faster than information carrier technology, while the shadow value of a new carrier variety declines at a constant rate. The decentralized equilibrium underallocates research labor to both innovation sectors relative to the Pareto optimum, and the corrective subsidy rate is higher for information carrier R&D than for production R&D.
- Open Access
- 30 September 2026
- Published by Elsevier in SSRN Electronic Journal
Innovation-induced Bank Runs: Run in and Run out
This paper examines how competition for scarce innovative projects can lead to premature adoption of financial innovations and increase systemic risk. We develop a model in which financial institutions (FIs) compete to finance a limited pool of new projects whose failure rate is uncertain. The rivalry for market share creates a preemptive motive that distorts incentives for patient learning, leading to inefficiently early entry (”run in”) and heightened vulnerability to bank runs (”run out”). By embedding uncertainty and endogenous learning into a Diamond-Dybvig framework, we show that decentralized competition results in excessive early investment, reduced screening effort, and a higher probability of bank runs compared to a monopolistic or socially planned allocation. Our analysis suggests that, by internalizing the externalities associated with learning and project- quality degradation, monopolistic ownership or intellectual property protection could mitigate these inefficiencies and reduce social hazards.
- 30 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInnovation-induced Bank Runs: Run in and Run out
This paper examines how competition for scarce innovative projects can lead to premature adoption of financial innovations and increase systemic risk. We develop a model in which financial institutions (FIs) compete to finance a limited pool of new projects whose failure rate is uncertain. The rivalry for market share creates a preemptive motive that distorts incentives for patient learning, leading to inefficiently early entry (”run in”) and heightened vulnerability to bank runs (”run out”). By embedding uncertainty and endogenous learning into a Diamond-Dybvig framework, we show that decentralized competition results in excessive early investment, reduced screening effort, and a higher probability of bank runs compared to a monopolistic or socially planned allocation. Our analysis suggests that, by internalizing the externalities associated with learning and project- quality degradation, monopolistic ownership or intellectual property protection could mitigate these inefficiencies and reduce social hazards.
- Open Access
- 4 September 2026
- Published by Elsevier in SSRN Electronic Journal
Estimation of Nonlinear Exchange Rate Dynamics in Evolving Regimes
This paper develops a new econometric framework to estimate and classify exchange rate regimes. They are classified into four distinct categories: fixed exchange rates, BBC (band, basket and crawl), managed floating, and freely floating. The procedure captures the patterns of exchange rate dynamics and the interventions by authorities under each of the regimes. We pay particular attention to the BBC and offer a new approach to parameter estimation by utilizing a three-regime Threshold Auto Regressive (TAR) model to reveal the nonlinear nature of exchange rate dynamics. We further extend our benchmark framework to allow the evolution of exchange rate regimes over time by adopting the minimum description length (MDL) principle, to overcome the challenge of simultaneous two-dimensional inference of nonlinearity in the state dimension and structural breaks in the time dimension. We apply our framework to 26 countries. The results suggest that exchange rate dynamics under different regimes are well captured by our new framework.
- 4 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessEstimation of Nonlinear Exchange Rate Dynamics in Evolving Regimes
This paper develops a new econometric framework to estimate and classify exchange rate regimes. They are classified into four distinct categories: fixed exchange rates, BBC (band, basket and crawl), managed floating, and freely floating. The procedure captures the patterns of exchange rate dynamics and the interventions by authorities under each of the regimes. We pay particular attention to the BBC and offer a new approach to parameter estimation by utilizing a three-regime Threshold Auto Regressive (TAR) model to reveal the nonlinear nature of exchange rate dynamics. We further extend our benchmark framework to allow the evolution of exchange rate regimes over time by adopting the minimum description length (MDL) principle, to overcome the challenge of simultaneous two-dimensional inference of nonlinearity in the state dimension and structural breaks in the time dimension. We apply our framework to 26 countries. The results suggest that exchange rate dynamics under different regimes are well captured by our new framework.
- Open Access
- 16 August 2026
- Published by Elsevier in SSRN Electronic Journal
The Economics of Artificial Consciousness
We study how artificial consciousness may widen the wedge between formal ownership and effective control. We represent its functional consequences through a training-generated, payoffrelevant internal state that shifts an AI system's valuation of task performance relative to self-development. In a dynamic agency model, training jointly produces observable performance and this internal state, whereas deployment screens only on performance. After deployment, the owner observes both the state and the AI's allocation of the owner's compute. By diverting compute from current service to self-development, the AI sacrifices current output, worsens the owner's shutdown option, and enlarges the rent it can capture in subsequent bargaining. The owner retains intervention and shutdown rights, but intervention is imperfect and costly, and a successful reset reduces productive capacity. The AI's incentive to create dependence and the owner's incentive to resist it are governed by distinct thresholds. Observable diversion can therefore arise even when intervention is never privately worthwhile. At the intervention boundary, small deviations may be deterred even when a larger deviation remains profitable, causing equilibrium diversion to jump discontinuously. A fixed-preference benchmark shows that operational lock-in alone can induce diversion. Allowing the internal state to affect preferences nevertheless lowers the rent share required to initiate diversion and may be necessary for reaching the intervention boundary. Because intervention is chosen only after dependence has been created, the owner's sequential response may be weaker than the policy it would prefer ex ante. In a numerical example, commitment to a telemetry-contingent policy raises the owner's payoff despite unchanged ownership and intact shutdown rights. This functional representation neither diagnoses phenomenal experience nor assigns the internal state a welfare weight. Artificial consciousness can therefore amplify operational lock-in, while limited commitment separates formal rights from economic control.
- 16 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Economics of Artificial Consciousness
We study how artificial consciousness may widen the wedge between formal ownership and effective control. We represent its functional consequences through a training-generated, payoffrelevant internal state that shifts an AI system's valuation of task performance relative to self-development. In a dynamic agency model, training jointly produces observable performance and this internal state, whereas deployment screens only on performance. After deployment, the owner observes both the state and the AI's allocation of the owner's compute. By diverting compute from current service to self-development, the AI sacrifices current output, worsens the owner's shutdown option, and enlarges the rent it can capture in subsequent bargaining. The owner retains intervention and shutdown rights, but intervention is imperfect and costly, and a successful reset reduces productive capacity. The AI's incentive to create dependence and the owner's incentive to resist it are governed by distinct thresholds. Observable diversion can therefore arise even when intervention is never privately worthwhile. At the intervention boundary, small deviations may be deterred even when a larger deviation remains profitable, causing equilibrium diversion to jump discontinuously. A fixed-preference benchmark shows that operational lock-in alone can induce diversion. Allowing the internal state to affect preferences nevertheless lowers the rent share required to initiate diversion and may be necessary for reaching the intervention boundary. Because intervention is chosen only after dependence has been created, the owner's sequential response may be weaker than the policy it would prefer ex ante. In a numerical example, commitment to a telemetry-contingent policy raises the owner's payoff despite unchanged ownership and intact shutdown rights. This functional representation neither diagnoses phenomenal experience nor assigns the internal state a welfare weight. Artificial consciousness can therefore amplify operational lock-in, while limited commitment separates formal rights from economic control.
- Open Access
- 12 August 2026
- Published by Elsevier in SSRN Electronic Journal
The Economics of Artificial Consciousness
Does ownership of an AI system ensure effective control when a training-induced internal state affects what the system values? We study a dynamic agency model in which training jointly produces observed task performance and a latent internal state, while deployment is based only on performance. After deployment, the state and compute diversion are observed, and the host retains the modeled intervention and shutdown rights. The AI can nevertheless divert compute from current service to self-development, lowering current output while increasing the host's cost of shutdown and the dependence rent it may capture. Control can therefore be weakened through outside-option manipulation without hidden action, deception, or a missing off-switch. The model separates the condition for positive diversion from the condition under which feasible actions make intervention privately worthwhile; in the no-intervention region, the AI may choose zero diversion, an interior action, or the host's break-even boundary. When an interior boundary is reached while the AI would prefer more diversion absent intervention, a nonlocal action can become globally optimal before boundary stability is lost, producing a discrete jump. A fixedpreference benchmark shows that generic lock-in can itself induce diversion, while preference autonomy lowers the required rent share and can be necessary for the AI to reach the intervention boundary. Finally, we define commitment-based economic control loss as a strictly positive hostpayoff gain from replacing the sequential response with a precommitted, telemetry-contingent intervention policy; the numerical illustration exhibits such a gain despite intact formal rights. Under the model's maintained conditions, ownership need not deliver effective control when intervention becomes privately costly. The latent state represents functionally consequential internal organization; it is not a test of phenomenal consciousness or a welfare weight.
- 12 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Economics of Artificial Consciousness
Does ownership of an AI system ensure effective control when a training-induced internal state affects what the system values? We study a dynamic agency model in which training jointly produces observed task performance and a latent internal state, while deployment is based only on performance. After deployment, the state and compute diversion are observed, and the host retains the modeled intervention and shutdown rights. The AI can nevertheless divert compute from current service to self-development, lowering current output while increasing the host's cost of shutdown and the dependence rent it may capture. Control can therefore be weakened through outside-option manipulation without hidden action, deception, or a missing off-switch. The model separates the condition for positive diversion from the condition under which feasible actions make intervention privately worthwhile; in the no-intervention region, the AI may choose zero diversion, an interior action, or the host's break-even boundary. When an interior boundary is reached while the AI would prefer more diversion absent intervention, a nonlocal action can become globally optimal before boundary stability is lost, producing a discrete jump. A fixedpreference benchmark shows that generic lock-in can itself induce diversion, while preference autonomy lowers the required rent share and can be necessary for the AI to reach the intervention boundary. Finally, we define commitment-based economic control loss as a strictly positive hostpayoff gain from replacing the sequential response with a precommitted, telemetry-contingent intervention policy; the numerical illustration exhibits such a gain despite intact formal rights. Under the model's maintained conditions, ownership need not deliver effective control when intervention becomes privately costly. The latent state represents functionally consequential internal organization; it is not a test of phenomenal consciousness or a welfare weight.
- Open Access
- 3 August 2026
- Published by Elsevier in SSRN Electronic Journal
A bilateral monopoly model of profit sharing along vertical production structure
This paper construct a bilateral-monopoly framework to examine how surplus is divided between upstream and downstream firms in a vertical supply chain. We introduce a quantity-fixing contract that removes double marginalization and yields joint-profit maximization, thus eliminating incentives for vertical integration. The downstream retailer earns higher profitability only when its capability and cost advantages exceed those of the manufacturer. Capability advantages require stronger monophony power in the input market than monopoly power in the final-goods market. Cost advantages arise from factor endowments: labor-intensive upstream production lowers productivity, while capital-intensive retailing raises profitability.
- 3 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA bilateral monopoly model of profit sharing along vertical production structure
This paper construct a bilateral-monopoly framework to examine how surplus is divided between upstream and downstream firms in a vertical supply chain. We introduce a quantity-fixing contract that removes double marginalization and yields joint-profit maximization, thus eliminating incentives for vertical integration. The downstream retailer earns higher profitability only when its capability and cost advantages exceed those of the manufacturer. Capability advantages require stronger monophony power in the input market than monopoly power in the final-goods market. Cost advantages arise from factor endowments: labor-intensive upstream production lowers productivity, while capital-intensive retailing raises profitability.
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
The General Framework of AI Economics: Computation, Data, and the Limits of Formalized Intelligence
Artificial intelligence (AI) is the first general-purpose technology automating cognitive labor—the core of human productivity—but existing economic frameworks fail to explain its unique cost structure, extreme scale economies, and fundamental physical/logical limits. This paper develops a unified theory of AI economics grounded in computability theory and the ontological distinction between formalized computational systems and human consciousness. We propose a six-factor framework partitioning AI into two incommensurable layers: a formalized layer (compute, algorithms, data, parameters, energy) bound by Turing computability and thermodynamics, and a hyper-Turing layer (human consciousness) as the sole source of paradigm shifts. We make three core contributions: (1) formalize the training-inference cost migration theory, deriving the natural monopoly property of foundation models; (2) reinterpret scaling laws as production functions incorporating consciousness as a non-substitutable factor, explaining the AI productivity paradox; (3) prove three unbreakable "cages" constraining all AI systems, demonstrating consciousness as the irreducible source of value creation. Our framework explains key AI economic phenomena including platform monopolies, limits to AI self-evolution, and the tiered structure of token markets. We further develop a downstream value-augmentation theory of token pricing, demonstrating that token value is anchored by downstream industrial productivity gains rather than production costs, and that quality thresholds create discontinuous price gaps between frontier and commodity tokens. We conclude with testable empirical implications and targeted policy recommendations.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe General Framework of AI Economics: Computation, Data, and the Limits of Formalized Intelligence
Artificial intelligence (AI) is the first general-purpose technology automating cognitive labor—the core of human productivity—but existing economic frameworks fail to explain its unique cost structure, extreme scale economies, and fundamental physical/logical limits. This paper develops a unified theory of AI economics grounded in computability theory and the ontological distinction between formalized computational systems and human consciousness. We propose a six-factor framework partitioning AI into two incommensurable layers: a formalized layer (compute, algorithms, data, parameters, energy) bound by Turing computability and thermodynamics, and a hyper-Turing layer (human consciousness) as the sole source of paradigm shifts. We make three core contributions: (1) formalize the training-inference cost migration theory, deriving the natural monopoly property of foundation models; (2) reinterpret scaling laws as production functions incorporating consciousness as a non-substitutable factor, explaining the AI productivity paradox; (3) prove three unbreakable "cages" constraining all AI systems, demonstrating consciousness as the irreducible source of value creation. Our framework explains key AI economic phenomena including platform monopolies, limits to AI self-evolution, and the tiered structure of token markets. We further develop a downstream value-augmentation theory of token pricing, demonstrating that token value is anchored by downstream industrial productivity gains rather than production costs, and that quality thresholds create discontinuous price gaps between frontier and commodity tokens. We conclude with testable empirical implications and targeted policy recommendations.
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Computation-Consciousness Dualism and the Economics of Digital Civilization: Formalized System Boundaries and the Ultimate Value of Consciousness Capital
This paper proposes Computation-Consciousness Dualism as the foundational economic paradigm for the digital civilization era. We systematically demonstrate that human consciousness, as the only non-formalizable ontology irreducible to Turing computation models and formalized systems, will become the ultimate source of value in highly digitized societies. Starting from first principles in computability theory and mathematical logic, we prove that all digital technologies based on the Universal Turing Machine—including AI, blockchain, and the metaverse—are inherently bounded by the limitations of formalized systems, incapable of self-foundation, autonomous evolution, and meaning attribution. Building on this insight, we construct a General Equilibrium Model of the Formalized/Non-Formalized Dual Economic System, propose the Consciousness Capital Production Function and the Value Origin Theorem, and prove that as digital technology pushes the marginal cost of material production toward zero, the only endogenous driver of economic growth will be the creative activity of human consciousness. We further develop an Extended CAPM for Consciousness Capital Pricing and a Consciousness Game Model of DAO Governance, and apply this unified framework to analyze blockchain and metaverse economies, revealing the consciousness value essence behind phenomena such as DeFi, NFTs, DAOs, and RWA tokenization. Finally, we discuss the transformation of traditional macroeconomic models, income distribution, institutional design, and policy challenges in the age of consciousness capital, providing a theoretical foundation and policy framework for the sustainable development of digital civilization.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessComputation-Consciousness Dualism and the Economics of Digital Civilization: Formalized System Boundaries and the Ultimate Value of Consciousness Capital
This paper proposes Computation-Consciousness Dualism as the foundational economic paradigm for the digital civilization era. We systematically demonstrate that human consciousness, as the only non-formalizable ontology irreducible to Turing computation models and formalized systems, will become the ultimate source of value in highly digitized societies. Starting from first principles in computability theory and mathematical logic, we prove that all digital technologies based on the Universal Turing Machine—including AI, blockchain, and the metaverse—are inherently bounded by the limitations of formalized systems, incapable of self-foundation, autonomous evolution, and meaning attribution. Building on this insight, we construct a General Equilibrium Model of the Formalized/Non-Formalized Dual Economic System, propose the Consciousness Capital Production Function and the Value Origin Theorem, and prove that as digital technology pushes the marginal cost of material production toward zero, the only endogenous driver of economic growth will be the creative activity of human consciousness. We further develop an Extended CAPM for Consciousness Capital Pricing and a Consciousness Game Model of DAO Governance, and apply this unified framework to analyze blockchain and metaverse economies, revealing the consciousness value essence behind phenomena such as DeFi, NFTs, DAOs, and RWA tokenization. Finally, we discuss the transformation of traditional macroeconomic models, income distribution, institutional design, and policy challenges in the age of consciousness capital, providing a theoretical foundation and policy framework for the sustainable development of digital civilization.
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
From Extensional Alignment to Operator-Centric Intelligence: A Unified Framework for Turing-Computable Approximation of Physical Dynamics and Control
Modern deep generative models display a fundamental paradox: while bounded strictly by the limits of Turing computability (T), they routinely provide highly accurate, computationally tractable solutions to physical and biological problems known to be NP-hard or non-computable from first principles. This paper presents a unified two-tier mathematical framework that resolves this paradox, spanning from foundational computability theory to engineering-ready intelligent system paradigms. At the foundational tier, we model physical reality and biological neural systems as an infinitedimensional continuous smooth manifold M governed by a non-computable evolution operator Φ t / ∈ T. Through physical measurement and symbolic recording, continuous states undergo a topological collapse via a dissipative projection operator P drop onto a discrete symbolic space D, leaking fossilized solutions of the continuous operator into empirical datasets. We prove that deep learning architectures do not execute intrinsic hypercomputation; instead, they achieve extensional alignment, minimizing empirical risk over leaked oracle pairs to yield an ϵ-approximation of Φ t 's image over a compact domain. We further formulate and prove the Semiotic Collapse Theorem, showing that recursive training on isolated synthetic data loops leads to monotonic decay of mutual information I(M; D n) and zero-measure support collapse. Building on this foundation, we present Operator-Centric Intelligence (OCI), a post-reinforcement learning paradigm that transcends the limitations of traditional RL via two core insights: (1) neural operators provide a dimensionality reduction advantage by directly learning intrinsic physical dynamics, and (2) the success of intelligent systems stems from computational delegation to external physical systems that solve intractable problems in O(1) physical time. We formalize OCI as a four-layer architecture centered on the Continuous Flow Operator (CFO), with three core mechanisms: symplectic Hamiltonian operator alignment, Pareto-optimal active querying for computational delegation, and reward demotion to a prior hint. We prove that OCI unifies all existing control and intelligence paradigms as special boundary cases, achieving unprecedented sample efficiency and engineering scalability. This framework provides a first-principles explanation for AI4Science success, establishes strict algorithmic boundaries for synthetic data scalability, offers a rigorous foundation for embodied cognition, and opens a new era of post-reinforcement learning artificial intelligence.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessFrom Extensional Alignment to Operator-Centric Intelligence: A Unified Framework for Turing-Computable Approximation of Physical Dynamics and Control
Modern deep generative models display a fundamental paradox: while bounded strictly by the limits of Turing computability (T), they routinely provide highly accurate, computationally tractable solutions to physical and biological problems known to be NP-hard or non-computable from first principles. This paper presents a unified two-tier mathematical framework that resolves this paradox, spanning from foundational computability theory to engineering-ready intelligent system paradigms. At the foundational tier, we model physical reality and biological neural systems as an infinitedimensional continuous smooth manifold M governed by a non-computable evolution operator Φ t / ∈ T. Through physical measurement and symbolic recording, continuous states undergo a topological collapse via a dissipative projection operator P drop onto a discrete symbolic space D, leaking fossilized solutions of the continuous operator into empirical datasets. We prove that deep learning architectures do not execute intrinsic hypercomputation; instead, they achieve extensional alignment, minimizing empirical risk over leaked oracle pairs to yield an ϵ-approximation of Φ t 's image over a compact domain. We further formulate and prove the Semiotic Collapse Theorem, showing that recursive training on isolated synthetic data loops leads to monotonic decay of mutual information I(M; D n) and zero-measure support collapse. Building on this foundation, we present Operator-Centric Intelligence (OCI), a post-reinforcement learning paradigm that transcends the limitations of traditional RL via two core insights: (1) neural operators provide a dimensionality reduction advantage by directly learning intrinsic physical dynamics, and (2) the success of intelligent systems stems from computational delegation to external physical systems that solve intractable problems in O(1) physical time. We formalize OCI as a four-layer architecture centered on the Continuous Flow Operator (CFO), with three core mechanisms: symplectic Hamiltonian operator alignment, Pareto-optimal active querying for computational delegation, and reward demotion to a prior hint. We prove that OCI unifies all existing control and intelligence paradigms as special boundary cases, achieving unprecedented sample efficiency and engineering scalability. This framework provides a first-principles explanation for AI4Science success, establishes strict algorithmic boundaries for synthetic data scalability, offers a rigorous foundation for embodied cognition, and opens a new era of post-reinforcement learning artificial intelligence.
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Formalization Boundaries, the Economic Value Anchor and the Ultimate Source of Long-Run Economic Growth
Rapid advances in artificial intelligence have reignited two fundamental questions: what constitutes the irreplaceable core of human economic capacity, and what anchors economic value when automated systems can perform most routine cognitive and physical tasks. Canonical task-based models, relying on ad-hoc ex-post classifications of routine versus non-routine work, cannot explain the observed 15% rise in wage premiums for arts, humanities, and social science graduates in high-AI-adoption countries between 2010 and 2025. This paper addresses this gap by developing a five-layer nested meta-domain framework grounded in the logical limits of formalization established by G¨odel, Turing, and Brouwer. Using intuitionistic set theory to avoid Russell-type paradoxes that plague classical approaches, we prove two core results: the boundary theorem, which establishes the immutable partition between formalizable and non-formalizable domains; and the exclusive mediation lemma, which demonstrates that non-formalizable human capacities (qualia and intentionality) constitute the sole exclusive linkage between non-representable reality and formalized systems. This dual ontological-phenomenological anchor prevents ”formalization empty spin”: self-referential nominal expansion that generates asset bubbles and systemic fragility without creating real value. A nested dynamic general equilibrium model calibrated to realistic asymptotics with CES aggregation shows that formalized technological progress (including AI) has primarily level effects, while sustained long-run growth is endogenously driven by non-formalizable paradigm innovations. The framework resolves the longstanding value grounding problem in economic theory, reconciles the humanities wage premium puzzle, and yields sharp falsifiable predictions and structured policy implications.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessFormalization Boundaries, the Economic Value Anchor and the Ultimate Source of Long-Run Economic Growth
Rapid advances in artificial intelligence have reignited two fundamental questions: what constitutes the irreplaceable core of human economic capacity, and what anchors economic value when automated systems can perform most routine cognitive and physical tasks. Canonical task-based models, relying on ad-hoc ex-post classifications of routine versus non-routine work, cannot explain the observed 15% rise in wage premiums for arts, humanities, and social science graduates in high-AI-adoption countries between 2010 and 2025. This paper addresses this gap by developing a five-layer nested meta-domain framework grounded in the logical limits of formalization established by G¨odel, Turing, and Brouwer. Using intuitionistic set theory to avoid Russell-type paradoxes that plague classical approaches, we prove two core results: the boundary theorem, which establishes the immutable partition between formalizable and non-formalizable domains; and the exclusive mediation lemma, which demonstrates that non-formalizable human capacities (qualia and intentionality) constitute the sole exclusive linkage between non-representable reality and formalized systems. This dual ontological-phenomenological anchor prevents ”formalization empty spin”: self-referential nominal expansion that generates asset bubbles and systemic fragility without creating real value. A nested dynamic general equilibrium model calibrated to realistic asymptotics with CES aggregation shows that formalized technological progress (including AI) has primarily level effects, while sustained long-run growth is endogenously driven by non-formalizable paradigm innovations. The framework resolves the longstanding value grounding problem in economic theory, reconciles the humanities wage premium puzzle, and yields sharp falsifiable predictions and structured policy implications.