Victor Frimpong

Most economic analysis of artificial intelligence (AI), including generative AI, asks how much output and economic growth it will generate. This article asks instead whether AI changes the determinants of national wealth. Drawing on growth, institutional, trade, development and neo-Schumpeterian theory, it distinguishes productivity enhancement, in which AI raises the prosperity attainable from a given set of determinants, from determinant transformation, in which AI reorders which determinants matter most. The distinction is expressed as a testable restriction: transformation requires that AI integration persistently alter the coefficients linking labour, capital, human capital, natural resources and institutions to national prosperity, with out-of-sample gains over conventional models. Four mechanisms could produce such change: factor substitution, productive complementarity, economic revaluation and institutional reconfiguration. They are integrated in the AI-Mediated National Wealth Formation Framework, which separates value creation from value capture and measures prosperity as output, resident income and comprehensive wealth. Of five refutable propositions, three predict that labour-cost advantage in AI-exposed services will erode, that the gap between output and resident income will widen where AI assets are foreign-owned, and that countries with similar AI integration will diverge for reasons that pre-existing advantage does not explain, with implications for catch-up and convergence. Cases from advanced, emerging, low-income, resource-rich and small economies illustrate the framework, and export controls and supplier concentration make access to compute partly a matter of state policy and supplier power. The result is a middle-range theory of AI and economic development: whether AI changes what nations must possess to become wealthy.

Generative AI has intensified scholarly publishing's need to distinguish formal accountability for a manuscript from the epistemic capacity to answer for it. Existing authorship guidance assigns responsibility to human authors regardless of AI involvement, but assigning responsibility does not establish that it can be meaningfully exercised. This paper addresses that gap by distinguishing nominal from substantive authorship responsibility and introducing the Duty to Defend (DtD): the obligation, internal to the status of scholarly authorship, to retain sufficient epistemic command of one's work to provide a reasonable account of its central claims, methods, evidence, and interpretations when legitimately challenged. Although the expectation that authors can defend their reasoning is already emerging in the literature, the normative question of why authorship itself carries such an obligation remains open. The paper derives this obligation from the reciprocal structure of scholarly credit rather than treating it as an additional compliance requirement and distinguishes it from an AI-detection standard: the relevant question is never whether AI was used, but whether the named author remains capable of standing behind the contribution attributed to them. A conceptual model then identifies three configurations — integrated, attenuated, and displaced authorship — describing how technological assistance can preserve or weaken the connection between credit and human intellectual agency. The paper translates this framework into a graded, trigger-based approach to editorial assurance, in which scrutiny escalates only as substantive scholarly concerns remain unresolved, and examines its own limits, including evidentiary difficulty, unequal capacity to demonstrate understanding, and the risk of converting a safeguard for authorship into a mechanism of suspicion.

As artificial intelligence (AI) regulation moves from principles towards enforceable obligations, an important institutional question concerns who should supervise compliance and under what conditions. This study examines AI supervision as a problem of supervisory allocation, distinguishing the design of regulatory obligations from the institutional arrangements that monitor and enforce them. Using a qualitative, desk-based comparative case-study design, it analyses five established supervisory regimes: Swiss anti-money-laundering supervision, U.S. securities regulation, EU data protection, EU conformity assessment and market surveillance, and Indian securities-market self-regulation. The cases are compared across institutional dimensions including expertise, information access, independence, enforcement authority, adaptability, accountability, supervisory capacity, and exposure to capture or conflicts of interest. The analysis finds that no single supervisory architecture suits all functions. Delegation suits bounded and technically specialised tasks, while functions requiring coercive authority and heightened public accountability favour direct public involvement. Where technical expertise and statutory authority must operate together, hybrid arrangements become particularly relevant. Delegation also creates a secondary governance requirement: meta-supervision, whereby public authorities oversee delegated supervisors' competence, independence, and performance. Building on these findings, the study develops the AI Supervisory Allocation Framework (ASAF) to assess the institutional fit of direct, delegated, and hybrid arrangements for specific supervisory functions. The study advances a functional approach to AI supervision by focusing on how supervisory responsibilities can be allocated based on a function's requirements and the capabilities and constraints of the institutions involved.

Generative artificial intelligence (GenAI) is transforming how scholarly references are discovered, synthesised, and incorporated into research. Yet a reference can exist, be bibliographically accurate, and appear relevant while remaining inadequately examined as evidence for the claim it supports. Existing approaches address reference verification, contextual support, AI disclosure, provenance, and human accountability, but give less explicit attention to whether the evidentiary pathway connecting an individual citation to a scholarly claim remains reconstructable. This conceptual article defines citation auditability as the degree to which the provenance, verification, evidentiary use, and human responsibility associated with a scholarly citation can be reconstructed and independently assessed. The framework comprises four interrelated dimensions: citation provenance, verification, evidentiary traceability, and human accountability. A 2 × 2 matrix further distinguishes citation auditability from citation correctness, demonstrating that a citation may be correct yet poorly auditable, or incorrect yet auditable. The article argues that citation integrity in AI-assisted scholarship should extend beyond verifying bibliographic objects toward ensuring that the evidentiary pathway from source to evidence to claim remains reconstructable and attributable to responsible human judgment. The article also considers the limits of this framework, including the risk that auditability claims may themselves be difficult to verify independently.

This is a useful and well-structured narrative review with a particularly valuable strength: it considers not only the benefits of service-learning but also the potential harm to students and community members when programmes are poorly implemented....

As financial institutions become increasingly reliant on similar foundational AI models, cloud infrastructure, data providers, and AI middleware platforms, diverse firms may begin to interpret and respond to market signals in increasingly similar ways. This paper examines model monoculture risk as a potential source of system-level exposure in banking and financial markets. Drawing on a structured, purposive conceptual review of the academic and regulatory literature, it develops the M³ Framework—Model, Market, and Middleware—to integrate three interacting mechanisms: Model Similarity, Market Synchronisation, and Middleware Concentration. Rather than treating these as a fixed causal sequence, the framework identifies their independent and cross-layer effects and distinguishes AI-mediated synchronisation from correlation arising from common exposures, conventional herding, and shared macroeconomic shocks. The paper further proposes the Model Monoculture Risk Index (MMRI) as a prototype index architecture and conceptual supervisory screening framework for characterising configurations of AI-related convergence, rather than as a calibrated quantitative measure of realised systemic risk. An illustrative application demonstrates how exposure classifications vary across alternative M³ configurations and attribution assumptions. The paper argues that governance should prioritise cross-institutional monitoring, dependency mapping, meaningful substitutability, and targeted stress testing rather than diversification by default. The framework provides a structured basis for future empirical research on AI-driven convergence and financial stability.

AI governance has primarily asked whether artificial intelligence systems are adequately governed. A less examined but increasingly consequential question is whether the expectations, demands, and authority that humans and organisations place on these systems are themselves adequately governed. This paper develops Demand-Side AI Governance as a conceptual framework for addressing this overlooked dimension. It distinguishes supply-side governance, concerned with the capabilities, design, deployment, behaviour, and control of AI systems, from demandside governance, concerned with the expectations placed on those capabilities, the decisions delegated to them, and the authority granted to their outputs. The framework identifies three interconnected objects of demand-side governance: expectation, delegation, and authority and theorises how available AI capability can be translated into progressively consequential organisational roles. A supply-demand governance matrix further identifies four configurations: Governance Alignment, Authority Inflation, Capability Vulnerability, and Compound Governance Failure. Of particular theoretical significance is Authority Inflation, in which an AI system may be technically and procedurally well governed while the organisational authority assigned to it exceeds what those controls justify. To translate the framework into organisational practice, the paper proposes three complementary mechanisms: Expectation Review, Delegation Control, and Authority Assurance. The paper argues that effective AI governance requires dual alignment between responsible provision of AI capabilities and the proportional expectations, delegation, and authority established around them. AI governance is incomplete when it governs the supply of algorithmic capability without also governing the demand for algorithmic authority.

What happens when the ethical and interpretive demands placed on artificial intelligence (AI) exceed what AI systems can legitimately support? This paper develops Expectation Inflation as a conceptual framework for explaining how AI-related ethical risk can arise not only from technical failures but also from escalating human and organisational expectations and the over-delegation of moral and interpretive authority. Adopting a conceptual and normative-analytical approach, the study synthesises interdisciplinary literature and constructs a theoretical model rather than seeking empirical validation. Drawing on Science and Technology Studies (STS), anthropomorphism, automation reliance, algorithmic authority, human–AI collaboration, and AI ethics, the paper conceptualises Expectation Inflation as an escalation process in which functional confidence can develop into epistemic and ultimately moral expectations. The Expectation Inflation Curve captures this progression through three stages: Rational Delegation, Normative Drift, and Moral Substitution. The paper further introduces Moral Design Capacity (MDC) as a context-dependent normative boundary beyond which the moral and interpretive authority attributed to an AI system becomes disproportionate to its demonstrated capabilities, decision context, and surrounding governance safeguards. To address this misalignment, the paper proposes Expectation Governance, comprising Delegation Boundaries, Expectation Auditing, and Moral Accountability Indexing. The framework complements existing approaches to responsible AI by shifting analytical attention from system-side properties alone to the demand side of AI governance: how organisations construct, amplify, and institutionalise expectations about what AI should know, interpret, and morally decide. The study thereby provides a theoretical basis for examining expectation–capacity misalignment in organisational and institutional settings and establishes propositions and governance mechanisms for subsequent empirical investigation.

Artificial intelligence (AI) can disadvantage individuals who do not use it, even when their own capabilities and absolute performance remain unchanged. This paper conceptualises this overlooked consequence as the social cost of AI non-use and introduces comparative technological disadvantage (CTD), grounded in positional-goods theory and Red Queen competitive dynamics, to explain how AI adoption by some actors can alter the relative position of others. CTD arises when AI-enabled gains in speed, scale, quality, or costefficiency reshape the standards against which non-augmented performance is evaluated. The central mechanism is AI-driven benchmark inflation, through which initially exceptional AI-assisted performance becomes normalised as the expected standard, generating economic, professional, educational, organisational, and social costs for non-users. The paper distinguishes constrained, voluntary, and principled forms of AI non-use, and shows how formally optional AI adoption can become effectively compulsory once non-use carries escalating comparative penalties. It also identifies boundary conditions under which non-use remains neutral or even advantageous, particularly where authenticity, confidentiality, independent judgment, expertise preservation, resilience, or human interaction matters. By showing that AI can redistribute advantage through the comparative environments its use creates, the paper extends technological inequality beyond questions of access and adoption. It argues that inclusive AI governance must address not only the conditions for responsible AI use, but also the conditions under which meaningful AI non-use remains viable.

This paper introduces the Governance Inversion Hypothesis (GIH) to explain a growing paradox in artificial intelligence (AI) governance: under conditions of increasing regulatory expansion and technological complexity, organisations may become more formally governed while simultaneously experiencing a decline in operational control over AI systems. Existing AI governance frameworks generally assume that stronger regulation improves accountability, oversight, and organisational control. This paper challenges that assumption by arguing that governance formalisation contributes to the erosion of control in AI-intensive environments. Drawing on institutional theory, organisational governance research, accountability scholarship, and emerging AI governance literature, the paper develops a conceptual framework that explains how regulatory expansion weakens operational authority through four interconnected mechanisms: authority fragmentation, symbolic governance expansion, externalisation of control, and authority paralysis. As governance systems become increasingly layered and procedurally dense, organisations may struggle to maintain coherent authority, technical visibility, escalation capability, and meaningful intervention power over opaque and externally mediated AI infrastructures. The paper extends institutional decoupling theory by introducing governance inversion as a structural condition in which governance expansion actively undermines operational coherence rather than strengthening it. It concludes that the central risk in AI governance may not be the absence of governance structures but the emergence of institutions that appear increasingly governed while progressively losing the capacity to govern effectively.