摘要：

目的 为从根本上将学术不端阻遏在形成之前，基于数智技术和主体交联构建前置化、系统化、智能化的学术不端风险分级预警体系。 方法 利用访谈法收集学术研究者、出版单位、监管部门等对学术不端治理的认知与要求，通过网络检索、电话调查等了解学术不端预警实践进展，运用德尔菲法收集专家意见并识别学术不端预警关键因素，采取逻辑分析法构建学术不端风险分级预警体系并分析其实现机制。 结果 创新性设计学术不端风险分级预警体系，构造学术不端风险预警指数（RAM）度量学术不端可能性。基于数据采集、分析研判、即时预警3大子系统，通过跨平台嵌入式模块进行流程预警、负面预警、互动预警，面向学术研究各环节、学术出版全流程、学术生态多领域协同避免萌芽态学术不端行为。 结论 学术不端风险分级预警体系是将学术不端防范关口前移的开拓性研究与前瞻性探索，促进学术不端治理从事后惩戒向事前预防的根本转变，从而有效维护学术诚信，确保学术研究健康发展。

关键词: 学术诚信, 学术不端, 风险预警, 分级预警, 学术征信, 数据挖掘

Abstract:

Purposes In order to fundamentally prevent academic misconduct before it forms， a pre-positioned， systematic and intelligent graded risk early warning system of academic misconduct is constructed based on digital intelligence technology and subject cross-linking. Methods The cognitions and requirements of academic researchers， publishing institutions， regulatory authorities， etc. on academic misconduct governance are collected by applying interview method； the practical progress of early warning of academic misconduct is understood through online retrievals and telephone surveys， etc.； the key factors of early warning of academic misconduct are identified by applying Delphi method to collect expert opinions； and the graded risk early warning system of academic misconduct is constructed and the implementation mechanism is analyzed by applying logical analysis method. Findings The graded risk early warning system of academic misconduct is designed innovatively， and the early warning index of risk of academic misconduct （RAM） is constructed to measure the likelihood of academic misconduct. Based on three main subsystems of data collection， analysis and judgment， and real-time warning， the process warning， negative warning， and interactive warning can be achieved through cross-platform adaptive embedded modules. Hence， the budding academic misconduct activities can be avoided collaboratedly facing the various aspects of academic research， the entire process of academic publishing， and the multi field of academic ecosystem. Conclusions The graded risk early warning system of academic misconduct is a pioneering research and prospective exploration that shifts the prevention threshold forward of academic misconduct， which can promote the fundamental transformation of academic misconduct governance from post punishment to pre prevention， so as to effectively maintain academic integrity， and ensure the healthy development of academic research.

Key words: Academic integrity, Academic misconduct, Risk early warning, Graded early warning, Academic credit, Data mining