When Fan Clips Help or Hurt: A New Math Model Rethinks Video Copyright Strategy

Every hit film and television series now lives a second life on short-video platforms. Commentary videos, highlight edits, mashups, character analyses, and condensed plot summaries circulate through TikTok, Douyin, and their peers, reaching audiences that official distribution alone may never touch. For copyright holders, this flood of user-produced derivative content has long posed a deceptively simple question: is all this secondary creation free advertising, or is it theft? A new study published in Frontiers in Communication argues that the question itself is badly framed. Using a sophisticated mathematical framework borrowed from game theory, researchers Yuan Meng and Qingqing Shi show that copyright governance in the platformized video industry is not a matter of suppressing infringement, but of steering an entire ecosystem of creators whose behavior can shift in ways that defy intuition.

The study, published on 14 October 2026, tackles a phenomenon that has transformed how audiovisual works circulate. Secondary creators occupy a peculiar dual position: they are audiences who consume and interpret original works, yet they also act as platform-dependent media producers who reorganize source material, frame evaluations, and shape public understanding of a film or series. Their output can generate attention spillovers, help viewers discover and sample new content, and redirect audiences toward the full-length original. Evidence from digital markets supports this complementary view: research on sampling shows that partial access to content can reduce uncertainty about quality and stimulate legal consumption, and studies of Chinese streaming markets have found that removing user-generated condensed clips actually reduced demand for the corresponding long-form works in the settings examined.

But complementarity is not guaranteed. Content that reveals central plot developments, condenses a substantial share of the narrative, or leans heavily on protected footage can divert attention or partially substitute for full-length viewing. The authors therefore distinguish two behavioral dimensions that existing research often conflates. The first is evaluative and narrative orientation, ranging from recommendation and appreciation to criticism, satire, and unfavorable interpretation. The second is creation and source-reliance intensity, which combines how active a creator is with how much the resulting content depends on materials from the original work. Crucially, these dimensions need not move together: a creator can produce highly active but interpretively original criticism, or glowing praise built on extensive unauthorized clips. A high level of activity does not constitute infringement by itself, just as positive content is not necessarily authorized.

To capture how these choices aggregate across a vast, anonymous creator population, the researchers built a finite-horizon leader-follower mean-field model. In this framework, the copyright holder acts as the leader, committing at the outset to a governance path composed of three instruments: dynamic content-quality governance that keeps the original work attractive and discussable, licensing openness that provides materials and cooperative opportunities for selected forms of derivative production, and copyright enforcement that raises the expected cost of intensive or unauthorized source reliance. The followers are a continuum of heterogeneous creators, each differing in traffic capital, the accumulated visibility and audience reach that determines how easily new content converts into views and monetization. Each creator chooses an orientation and an intensity level to maximize intertemporal returns, while individual decisions interact through collective reinforcement and aggregate into ecosystem-level outcomes that feed back into the copyright holder’s evaluation of its own governance path.

The mathematics is demanding. Individual creator optimization is captured by a Hamilton-Jacobi-Bellman equation, while the evolution of the population’s traffic-capital distribution follows a Fokker-Planck-Kolmogorov equation, with consistency conditions ensuring that the distribution-weighted strategies reproduce the collective environment each creator anticipated. The copyright holder then evaluates every candidate governance path together with the complete creator response it induces, solving an open-loop leader-follower problem in which the population’s evolution is internal to the leader’s objective. Numerical analysis under normalized benchmark parameters, with a log-normal initial distribution representing an ecosystem rich in small and medium accounts and sparse in high-traffic creators, produced a set of sensitivity experiments in which one mechanism at a time was varied across twenty values, all passing the authors’ convergence and robustness checks.

The first major finding is a warning against measuring an ecosystem by its volume alone. Creator activity, collective content orientation, and cross-creator dispersion turned out to be distinct dimensions that can move in different directions. In one experiment, raising the cost of enforcement increased average creation and source-reliance intensity from 0.973 to 1.074 while leaving average orientation essentially unchanged. In another, stronger licensing incentives raised both average orientation and activity while sharply reducing orientation variance, indicating convergence toward a strongly positive collective stance rather than polarization. A highly active secondary-creation ecosystem, the model shows, is not necessarily favorable to the original work; activity alone is an incomplete and potentially misleading description of ecosystem conditions.

The second finding concerns the gap between direct effects and equilibrium outcomes. When the licensing-incentive coefficient rose from 0.10 to 0.70, creators became more positively oriented and more active, exactly as the direct incentive would suggest. Yet the copyright holder’s complete objective declined monotonically from 2.138 to 1.987, because the optimized governance portfolio simultaneously rebalanced: licensing openness fell, quality governance declined, and enforcement rose. A mechanism that looks beneficial in isolation can be partly offset by the endogenous adjustments it triggers across the whole governance portfolio and the creator population. Similarly, stronger quality inducement raised average orientation and lowered orientation dispersion, yet the complete copyright-holder objective still declined over the examined range, and licensing openness responded non-monotonically. Evaluating any single instrument in isolation, the authors conclude, answers a different question than the one copyright holders actually face.

The third finding may be the most practically consequential. When the researchers steepened the penalty schedule, making expected consequences rise more sharply with source-reliance intensity, the results diverged from what broad enforcement achieves. Average creation and source-reliance intensity fell monotonically from 1.341 to 0.905, its variance dropped from 0.02559 to 0.00843, and average enforcement itself declined from 0.751 to 0.627, while quality governance and licensing openness both rose. This was the only experiment in which the complete copyright-holder objective increased over the full range, climbing from 1.738 to 2.264. In other words, intervention that targets intensive source reliance specifically can restrain substitutive behavior while reducing the need for sweeping takedowns and monitoring. Across-the-board enforcement, by contrast, risks suppressing transformative commentary, fan editing, and culturally productive reuse alongside genuinely problematic content.

The authors are careful about the limits of these conclusions. The platform is modeled as an algorithmic and institutional environment rather than a strategic actor, audience behavior enters only indirectly through traffic returns and substitution terms, and creator heterogeneity is summarized by traffic capital alone. The parameter values are normalized benchmarks, not structural estimates for Douyin, TikTok, or any specific work, so the results provide mechanism-based evidence rather than empirical causal claims. The authors also stress that a negative orientation is not equivalent to infringement and that high source reliance does not by itself establish legal liability; the penalty term is a reduced-form expected cost, not a jurisdiction-specific legal test. Criticism, satire, and unfavorable interpretation remain analytically distinct from source-material dependence, and the model offers no legal standard for either.

Even with those caveats, the study reframes a debate that has mostly been fought in legal and cultural terms. Copyright governance in short-video environments, the authors argue, is better understood as creator-ecosystem governance: the coordination of content-quality management, licensing conditions, and differentiated intervention to shape the conditions under which different forms of participation coexist with the preservation of an original work’s value. Practical implications follow. Authorized material libraries, clear reuse boundaries, creator-collaboration programs, and attribution mechanisms can lower coordination costs for participation that contributes to discovery and interpretation, while closer review can be reserved for source-intensive formats that reproduce substantial narrative content. The goal is not to minimize secondary creation or maximize short-term control, but to sustain a media ecosystem in which circulation, reinterpretation, and original incentives reinforce rather than undermine one another. As long-form and short-form video continue to intertwine, the future of copyright may depend less on enforcement muscle than on the design of the participation conditions themselves.

Subject of Research: A mean-field game-theoretic analysis of how copyright holders' quality governance, licensing, and enforcement strategies shape secondary-creation ecosystems on short-video platforms.

Article Title: Copyright governance and the dynamics of secondary-creation ecosystems in the platformized video industry

Article References: Meng, Y., & Shi, Q. (2026). Copyright governance and the dynamics of secondary-creation ecosystems in the platformized video industry. Frontiers in Communication, 11, Article 1962569. https://doi.org/10.3389/fcomm.2026.1962569

Image Credits: AI Generated

DOI: 10.3389/fcomm.2026.1962569

Keywords: copyright governance, secondary creation, short-video platforms, mean-field games, creator ecosystem, platformization, licensing openness, copyright enforcement, long-form video, participatory media, leader-follower model, digital copyright