arXiv Science⌕ Search

arXiv · 2610.30153

The Universal Law of Academic Capital: A Parameter-Unified Theory of Academic Integritology

Abstract

Academic capital, comprising research funding ($\mathcal{F}$), accumulated titles and honours ($\mathcal{L}$), and administrative power ($\mathcal{A}$), has undergone a silent global decoupling from verifiable epistemic output since the 1990s. Here we present a parameter-unified theory of Academic Integritology that integrates 64 major parameters into a coherent chassis of twelve fundamental equations (Eqs. 0-11b). The theory introduces two computable metrics: $Ψ$ (net epistemic indispensability) and $P_{\mathrm{safe}}$ (the Markov-blanket safety boundary of systemic integrity). The $D_{zj}$ fraudulence index, defined as $κ[\mathrm{Cap}]/E[P]$, quantifies this decoupling. Calibrated against retraction data from 40 jurisdictions and the careers of 12 Nobel-caliber scientists, the theory derives a universal law: when $E[P]/(w_1[\mathrm{Cap}])$ falls below $θ_c \approx 0.25$, honest research ceases to be an evolutionarily stable strategy. Three jointly necessary levers, raising $E[P]/(w_1[\mathrm{Cap}])$, auditing $R_{\mathrm{pol}}$, and constraining $ω$, restore $P_{\mathrm{safe}}$. Three levers, pulled simultaneously, suffice. The theory transforms academic integrity from a moral aspiration into an engineerable system property.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Yiguo Zhang, Ruiguang Wang, Xi Chen, Zhengwen Zhang. 2026-08-14. The Universal Law of Academic Capital: A Parameter-Unified Theory of Academic Integritology. https://arxiv.org/abs/2610.30153

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Opinion dynamics on switching networks

We study opinion dynamics over a two-layer multiplex network with weighted, directed influence and reciprocal links. Within each layer, an agent weighs its neighbors equally, so the influence of a neighbor is inversely proportional to the agent's number of neighbors in that layer. Agents update their opinions over time to coordinate with their neighbors. However, the frequency of agents' interactions with neighbors in different network layers differs: one layer is always active, whereas the other one is active only once every $k$ time steps. Consequently, the multiplex network's adjacency matrices are time-varying. We characterize how the activation frequency of the second layer impacts the convergence of the opinion dynamics process. Without requiring agents to place positive weight on their own opinions, we prove that, for every activation period $k$, the opinions reach a consensus whose value depends on $k$, and that the asymptotic per-step convergence factor equals the $k$th root of the second-largest eigenvalue modulus of the one-period transition matrix. As $k$ grows, this factor generically converges to that of the always-active layer when this layer is connected; when it is disconnected, the time to consensus generically grows linearly in $k$. Numerical experiments on 1,000 random multiplex networks confirm these results.

physics.soc-ph↗

Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits

The scientific claims drawn from LLM social simulations should be no stronger than the robustness audits that support them. Generative agents bring new expressive power to agent-based modeling, enabling simulations of collective social processes like cooperation, polarization, and norm formation. Yet they also introduce complexity through additional architectural choices, such as agent specification, memory representation, interaction protocols, and environment design. Small perturbations that appear minor to researchers can cascade into macro-level outcomes through repeated interaction, creating a "butterfly effect." Consequently, scientific claims drawn from LLM social simulations may reflect implementation artifacts rather than the social mechanisms being modeled. We support this position with two case studies: a repeated Prisoner's Dilemma and a social media echo chamber simulation. Across multiple models, minor perturbations in persona format and game-instruction framing shift cooperation rates by up to 76 percentage points, while network homophily and hub assignment produce significant and consistent shifts in polarization metrics. We also find that sensitivity is unevenly distributed across both architectural choices and model families: the same perturbation that produces the 76 pp shift in one frontier model only shifts another by 1 pp. Robustness is therefore a property that should be measured per claim and per model, not assumed. To address this validation gap, we introduce TRAILS (Taxonomy for Robustness Audits In LLM Simulations), a robustness-audit taxonomy spanning three levels of simulation design: agent (micro-level), interaction (meso-level), and system (macro-level). We call for robustness to become a first-order validation requirement before LLM social simulations are used to explain mechanisms, evaluate interventions, or inform decisions.

physics.soc-ph↗

Fractional-order SIS-UAU subdiffusion model for rumor propagation

Understanding information and rumor diffusion in complex social systems is essential for risk management. Classical SIS-UAU (Susceptible-Infected-Susceptible / Unaware-Aware-Unaware) models rely on Markovian assumptions and thus fail to capture the memory effects and subdiffusive behaviors commonly observed in real social interactions. To address this limitation, we propose a fractional-order SIS-UAU model based on Caputo derivatives, embedding non-Markovian memory into a duplex awareness-rumor spreading process. Methodologically, we derive the corresponding mean-field equations and establish existence, uniqueness, positivity, boundedness, and a Matignon-type stability criterion. The basic reproduction number is obtained, revealing how awareness dynamics reshape the effective rumor threshold. Numerically, we develop a block-adaptive FFT (Fast Fourier Transform) solver to efficiently handle long-horizon fractional dynamics, enabling large-scale simulations. Our results show that fractional memory significantly slows the transient evolution of rumors without altering the steady state. Furthermore, infection-layer parameters dominate long-term rumor prevalence, awareness transmission substantially suppresses rumor levels, and multilayer network topology-especially degree heterogeneity-critically influences propagation outcomes. Collectively, these findings underscore the necessity of integrating memory effects, awareness dynamics, and network structure for realistic rumor modeling

physics.soc-ph↗