Authorship in the age of generative AI hinges on whether human reasons continuously and effectively guide a work, not on how many manual steps a person performs; the paper proposes 'dynamic-reflexive tracking'—diachronic reason formation, reflective uptake, trajectory efficacy, and contestability—to define meaningful human control over creation.

Main Finding

Human authorship in generative-AI–assisted creation should be judged by whether a human’s reasons continuously, reflectively, and effectively shape the development trajectory of a work—not by counts of manual steps, origin of initial ideas, or final-selection acts. To capture this, the paper introduces Meaningful Human Control (MHC) adapted to creative contexts and proposes Dynamic-Reflexive Tracking (DRT): a four-part, diachronic normative framework that specifies when human judgment genuinely exerts authorial control in human–AI co-creation.

Key Points

• Distinction: creativity of the artefact (novelty/value) is separable from creative agency (the human subject that forms reasons and guides development). GenAI makes these two dimensions diverge.
• Levels of impact:
  • Individual: GenAI expands productive capacity (more outputs, lower technical thresholds) but can reduce agential engagement via cognitive offloading and predictive personalization.
  • Collective: Model capacity is built on social-cultural resources but is sedimented in platforms; user interactions may not translate into accumulable, transferable collective skills or diversity.
  • Industrial: Firms can capture technical capacity, altering labour markets, bargaining power, and the institutional power to define authorship and distribute value.
• Four senses of “empowerment” to evaluate GenAI effects: productive empowerment, agential empowerment, process autonomy, distributive justice—these need not move together.
• Legal/normative problem: authorship attribution cannot rest on physical acts or volume of prompts; it requires assessing which human activities instantiate creative judgment.
• Classical MHC’s tracking condition is too static for creative processes because human reasons often form, change, or are abandoned through ongoing interaction with AI.
• Dynamic-Reflexive Tracking (DRT) — proposed normative replacement for static tracking — requires that a creator’s reasons:
• Diachronic reason formation: can form and evolve over time in relation to the work’s evolving trajectory.
• Reflective uptake: the creator reflectively engages with AI outputs (not merely automatic/habitual responses).
• Trajectory efficacy: those reasons have genuine causal efficacy in shaping subsequent development.
• Contestability and redirection: the creator can reject system defaults and redirect the process.
  • Plus a minimal tracing requirement: sufficient traceability to link reasons and interventions to outcomes.
• Practical output: the paper translates DRT into reflective self-interrogation tools for creators (prompts/criteria to assess whether their interaction satisfies DRT).
• Conclusion: DRT is a necessary normative condition for sustaining human creative agency and authorial attribution under GenAI, though not by itself sufficient to settle full legal authorship.

Data & Methods

• Methodological approach: conceptual, normative, and interdisciplinary analysis (philosophy of action, legal theory, human–computer interaction, and socio-technical critique).
• Evidence and grounding:
  • Literature synthesis across aesthetics, agency theory, human–AI interaction, copyright law, and labor/industrial studies.
  • Analytical critique of existing MHC/tracking conditions and their limitations for creative, diachronic processes.
  • Proposal of an explicit normative framework (DRT) with four operational conditions and a minimal tracing requirement.
  • Translation of the framework into actionable reflective checklists for practitioners.
• No primary empirical dataset or econometric analysis; the paper situates and motivates its normative claims with referenced empirical and theoretical work (studies on detectability of AI outputs, personalization effects, industry cases, market trends).

Implications for AI Economics

• Attribution and value allocation:
  • Authorship standards grounded in DRT would affect rights, licensing, and revenue splits—potentially altering revenue claims between individual creators, platforms, and model owners.
  • Clearer normative criteria for human authorship could change incentives for platforms to design interfaces that enable reflective uptake and redirection (to preserve creator claims).
• Labor markets and bargaining power:
  • If authorship is tied to diachronic, reflective control, employers and platforms may need to preserve roles that allow creators genuine redirection and evaluative labor—affecting job design, upskilling needs, and wage premia for higher-agency tasks.
  • Conversely, if market practice treats mere invocation/selection as sufficient, downward pressure on professional creative wages and skill transmission may continue.
• Product and platform design:
  • Economic value of tools will increasingly depend on their ability to facilitate DRT (features: provenance/tracing logs, interactive affordances that support reflective engagement, options for contesting defaults).
  • Platforms could monetize “authorship-enabling” features (audit trails, provenance badges, workflow controls) or face regulatory pressure to provide them.
• Market structure and competition:
  • Model/ platform concentration matters: when generative capacity is concentrated, access and control over accumulable creative capacities are restricted, reinforcing incumbent market power and affecting distribution of rents.
  • Policies that require traceability or richer provenance could raise costs for centralized providers but create markets for interoperable, auditable tools.
• Innovation and diversity:
  • Risk of homogenization: optimization for platform metrics and default model styles can compress diversity and long-run creative complementarities—negative externalities for innovation and consumer surplus.
  • Policies or business models that preserve process autonomy and agential empowerment (e.g., open-models, community-curated datasets, compensation for training-data originators) could sustain long-run creative capital.
• Regulation and measurement:
  • Economists and policymakers need operational metrics for agential empowerment and DRT compliance (e.g., measurable indicators of reflective interventions, trace-backed causal links between human inputs and trajectory changes).
  • These metrics would inform copyright policy, procurement rules, subsidy programs for creative labor, and antitrust/regulatory assessments of platform power.
• Empirical research agenda:
  • Measure how different UI/UX designs and business models affect creators’ ability to satisfy DRT conditions and capture value.
  • Quantify effects of attribution rules on creator earnings, innovation rates, and market concentration.
  • Study dynamic complementarities between human evaluative labour and model generation across creative industries to guide optimal labor and training investments.

Overall, the paper reframes authorship as a normative, processual property tied to the formation and efficacy of human reasons over time. For AI economists, that shift directs attention from simple productivity gains to the institutional, contractual, and design changes needed to preserve (or commodify) human agency, authorship rights, and the equitable distribution of economic value in AI-mediated creative markets.