Hindawi解决大规模出版操纵问题白皮书中译

2023年12月12日，Hindawi发表了一份白皮书，介绍解决大规模出版操纵问题的历程和对学术出版的启示（微信公众号 https://mp.weixin.qq.com/s/3Jzuv6jBAemd532vj92jxw ，Hindawi Blog https://www.hindawi.com/post/publication-fraud-research-integrity-whitepaper/ ），原文：https://www.wiley.com/en-us/network/publishing/research-publishing/open-access/hindawi-publication-manipulation-whitepaper

Tackling publication manipulation at scale: Hindawi’s journey and lessons for academic publishing

大规模解决出版操纵问题：Hindawi的历程及对学术出版的启示

Section 1: Introduction

第1节：简介

The importance of open and timely communication by trusted scholarly publishers has arguably never been greater. The COVID-19 pandemic illustrated the value of rapid and open dissemination of research findings (Tavernier, 2020), and the rise in open access publishing means that restrictions to accessing research are increasingly being removed. At the same time, new technologies such as artificial intelligence (AI) hold the promise to fuel a rapid increase in scientific discovery (Economist, 2023), remove language barriers to writing papers, and accelerate the use of the scholarly record to develop new hypotheses (Extance, 2018).

可以说，值得信赖的学术出版商进行公开、及时的交流从未像现在这样重要。COVID-19 大流行说明了快速、公开传播研究成果的价值(Tavernier, 2020)，而开放获取出版的兴起意味着获取研究成果的限制正日益消除。与此同时，人工智能（AI）等新技术有望推动科学发现的快速增长(Economist, 2023)，消除撰写论文的语言障碍，并加快利用学术记录提出新的假设(Extance, 2018)。

Yet alongside this promise, the scholarly publishing industry faces a deepening crisis – one where some of the drivers of progress also pose an increasing threat (Liverpool, 2023). We are in a time when researchers face increasing external pressures to publish for career advancement while fabricating a paper using generative AI has never been easier. These factors make journals a target for the continued and systematic manipulation of the publication process via paper mills and other forms of problematic activity. The last few years have seen increasing examples of publishers facing these challenges at an unprecedented scale (Else and van Noorden, 2021, van Noorden, 2023).

然而，在实现这一承诺的同时，学术出版业也面临着日益加深的危机——其中一些推动进步的因素也构成了日益严重的威胁(Liverpool, 2023)。我们正处于这样一个时代：研究人员面临着越来越大的外部压力，他们需要通过发表论文来获得职业发展，而使用生成式人工智能来编造论文则变得前所未有的容易。这些因素使得期刊成为通过论文加工厂和其他形式的问题活动持续、系统地操纵出版流程的目标。过去几年中，出版商以前所未有的规模面临这些挑战的例子越来越多(Else and van Noorden, 2021, van Noorden, 2023)。

Although far from being the only publisher affected, the last 18 months have seen Hindawi at the eye of this storm of publication manipulation at scale, and we are on a continued journey through it. Hindawi and Wiley have together learned many lessons about how bad actors infiltrate publishing systems, and we have taken new approaches to clean up the scholarly record at scale and protect against further attacks. We believe that sharing our experience with others across the industry plays an important part in developing a collaborative approach to tackling systematic manipulation at scale.

尽管Hindawi不是唯一一家受影响的出版商，但在过去的18个月里，Hindawi一直处于这场大规模出版操纵风暴的中心，而我们也在继续经历这场风暴。Hindawi和Wiley共同汲取了许多关于不良分子如何渗透出版系统的教训，我们采取了新的方法来大规模清理学术记录并防止进一步的攻击。我们相信，与整个行业的其他人分享我们的经验，对于开发一种协作方法来应对大规模的系统性操纵具有重要作用。

In this article, we describe what paper mills are and how we believe they operate, as well as other – often related – unethical activities that are distorting the scholarly landscape and endangering public trust in research. These activities include unethical authorship practices, the use of AI in manuscript fabrication and image manipulation, and peer review manipulation. We explore the significant impacts and disruptions resulting from these factors and the steps Hindawi has taken– and is continuing to take – to investigate and clean up the scholarly record through “designing a new retraction process that will help us, and potentially others, accelerate and deal with this new era of mass retractions fairly” (Flynn 2023).

在本文中，我们将介绍什么是论文工厂，我们认为论文工厂是如何运作的，以及其他往往与之相关的不道德活动，这些活动正在扭曲学术环境，危害公众对研究的信任。这些活动包括不道德的作者署名行为、利用人工智能编造手稿和篡改图像，以及操纵同行评议。我们探讨了这些因素造成的重大影响和破坏，以及Hindawi已经采取并将继续采取的措施，即通过“设计一种新的撤稿流程，帮助我们并可能帮助其他人，加快并公平地应对这个大规模撤稿的新时代”(Flynn 2023)来调查和清理学术记录。

We conclude with a call to all stakeholders within the scholarly publishing ecosystem, with recommendations to work together and introduce a collaborative approach to journal security and research integrity. By sharing our approach to the challenges that have led to such widescale and deliberate undermining of trust in research, the hope is that others can build on and refine the process, helping to further ensure publishing systems are less vulnerable to manipulation in future.

最后，我们向学术出版生态系统中的所有利益相关者发出呼吁，建议他们携手合作，采用合作的方法来解决期刊安全和研究诚信问题。通过分享我们应对挑战的方法，我们希望其他人能借鉴并完善这一过程，从而帮助进一步确保出版系统在未来不那么容易受到操纵。

Section 2: An industry-wide problem

第2节：一个全行业的问题

There is a “continuation of unethical activity” in the publishing world that represents a “major and ongoing challenge for the entire scholarly publishing industry and all who rely on the integrity of the scholarly record” (Ferguson, 2022). Such activity could range from inappropriate research and publishing practices due to lack of training through “misconduct, fraud, or even criminality” (Kolstoe, 2023; UKRIO, 2023). Whether it is due to researchers feeling the pressure to publish (Oransky and Marcus, 2023) or through “fake paper factories” operating at scale (Else and van Noorden, 2021, van Noorden, 2023), publishing processes have been compromised and publication misconduct is widespread.

出版界的“不道德行为仍在继续”，这对“整个学术出版业和所有依赖学术记录完整性的人来说都是一个重大且持续的挑战”(Ferguson, 2022)。此类活动可能包括因缺乏培训而导致的不当研究和出版行为，也可能包括“不当行为、欺诈甚至犯罪”(Kolstoe, 2023; UKRIO, 2023)。无论是由于研究人员感受到了发表论文的压力(Oransky and Marcus, 2023)，还是由于“假论文工厂”的规模化运作(Else and van Noorden, 2021, van Noorden, 2023)，发表论文的过程已经受到损害，发表论文的不当行为普遍存在。

Unethical authorship practices

作者的不道德行为

The academic culture of ‘publish or perish’ has incentivized unethical behaviour as researchers feel intense pressure to produce and publish high volumes of articles to advance in their careers (Camacho, 2021). In recent years, there has been a concerning surge in authorship misconduct, a trend that poses a serious threat to the integrity of scholarly work. Unethical practices that fall into this category include ghost-writing, guest authorship, and authorship for sale; while not new, they are becoming more prevalent.

“出版或毁灭”的学术文化助长了不道德行为，因为研究人员感到了巨大的压力，他们必须撰写和发表大量文章，才能在职业生涯中取得进步(Camacho, 2021)。近年来，作者不端行为的激增令人担忧，这一趋势对学术工作的完整性构成了严重威胁。这类不道德行为包括代写、荣誉作者和出售作者身份；虽然这些行为并不新鲜，但却越来越普遍。

Some researchers are publishing an astonishing number of papers, with certain authors putting their name to hundreds of articles within a single year. This includes papers that seemingly are not within their usual field of expertise and alongside unrelated co-authors (Ansede, 2023). On the flip side, authorship positions are also being offered for sale. Although it is difficult to identify the specific companies concerned, advertisements selling co-authorship have been found (e.g., Abalkina, 2021, Kincaid, 2023).

一些研究人员发表的论文数量惊人，某些作者在一年内发表了数百篇文章，包括似乎不属于他们的专业领域的论文，以及与不相关的合著者一起发表的论文(Ansede, 2023)。另一方面，作者位置也被出售。虽然很难确定具体的相关公司，但还是发现了出售合著者的广告(e.g., Abalkina, 2021, Kincaid, 2023)。

These deceptive tactics not only compromise the credibility of academic publications but also undermine the principles of transparency and accountability that make scholarly research trustworthy and which are crucial to innovation and advancing knowledge.

这些欺骗手段不仅损害了学术出版物的可信度，还破坏了学术研究值得信赖的透明和问责原则，而这些原则对于创新和知识进步至关重要。

The use of AI in manuscript fabrication and image manipulation

人工智能在手稿制作和图像处理中的应用

Artificial intelligence will be a transformative tool for good in scholarly research that will help fuel innovation, but it is also a powerful instrument for unethical activity in the wrong hands, giving rise to the creation of deceptive ‘original’ research. AI tools can be exploited to generate fabricated data, manipulate images, and even devise entire studies with seemingly legitimate results (Naddaf, 2023).

人工智能将成为学术研究中的变革性利器，有助于推动创新，但它也是不道德活动的有力工具，一旦落入坏人之手，就会产生欺骗性的“原创”研究。人工智能工具可被用来生成捏造的数据、篡改图像，甚至设计出具有看似合法结果的整个研究(Naddaf, 2023)。

This poses a significant threat to the credibility of scientific inquiry, as researchers and publishers may struggle to discern between authentic and manipulated content.

这对科学研究的可信度构成了重大威胁，因为研究人员和出版商可能难以分辨真实内容和被操纵的内容。

AI “brings challenges to research integrity and accuracy, including the potential increase in plagiarism, image manipulation, and paper mills” (Zhou, 2023), through its ability to quickly fabricate research and evade detection by standard checks. Detecting synthetic output from large language models is a society-wide challenge, and as of today there are no reliable methods to do so. With increasingly sophisticated language models such as ChatGPT, plagiarism is likely to become undetectable and AI-human writing will become a normal part of the creative process (Eaton, 2023). Humans, however, will remain responsible for the trustworthiness of scholarly outputs. The challenge in such a ‘post-plagiarism world’ is how an ethical framework can be created and taught that ensures appropriate attribution and holds researchers, publishers, funders, institutions, and others to account for the integrity of scholarly work (Eaton, 2023).

人工智能“给研究的完整性和准确性带来了挑战，包括剽窃、图像篡改和论文工厂的潜在增加”(Zhou, 2023)，因为人工智能有能力快速编造研究成果并逃避标准检查的检测。检测大型语言模型的合成输出是一项全社会的挑战，目前还没有可靠的方法来做到这一点。随着 ChatGPT等语言模型的日益成熟，剽窃行为很可能变得无法检测，人工智能与人类共同写作将成为创作过程的正常部分(Eaton, 2023)。然而，人类仍将对学术成果的可信度负责。在这样一个“后剽窃世界”中，面临的挑战是如何创建和传授道德框架，以确保适当的归属，并使研究人员、出版商、资助者、机构和其他人对学术成果的完整性负责(Eaton, 2023)。

Paper mills

论文工厂

Paper mills have been most recently defined by the Committee on Publication Ethics (COPE) as: “An individual, group of individuals, or organisation that aims to manipulate the publication process to achieve the publication of articles for the purposes of financial gain” (COPE Council, 2023a).

出版伦理委员会（COPE）最近将论文工厂定义为：旨在操纵出版过程以实现以谋取经济利益而发表文章的个人、团体或组织(COPE Council, 2023a)。

Awareness of firms selling papers to researchers or arranging to bypass peer review goes back to at least 2010 (Else and Van Noorden, 2021). In the past two years, however, there has been a dramatic increase in systematic manipulation and paper mills are now a serious “sort of industrial scale fraud” (Nunes and Bishop, 2023).

人们对公司向研究人员出售论文或安排绕过同行评议的认识至少可以追溯到2010年(Else and Van Noorden, 2021)。然而，在过去两年中，系统性的操纵行为急剧增加，论文工厂现在已经成为一种严重的“工业规模的欺诈行为”(Nunes and Bishop, 2023)。

Paper mills contaminate the scholarly literature with fabricated or nonsensical results. If these papers are discovered by the publisher or the academic community, they require retraction by the publisher in order to correct the publication record. The discovery that a papermill has infiltrated a journal can have severe consequences, including the loss of credibility and trust among researchers and removal from scholarly indexes.

论文工厂用捏造的或无意义的结果污染学术文献。如果这些论文被出版商或学术界发现，出版商就必须撤回这些论文，以纠正出版记录。期刊被发现被论文工厂渗透会造成严重后果，包括失去研究人员的信誉和信任，以及被从学术索引中删除。

Journals soliciting proposals for Special Issues (SIs) can present particularly attractive targets for paper mills, where ‘Guest Editors’ are recruited to take a temporary editorial role in a journal. Guest Editors can submit proposals for SIs, help solicit the papers themselves, and oversee and contribute to peer review. Strategies of paper mills include manipulating the identities of Guest Editors, authors, and even reviewers (to appear as genuine researchers) and/or fabricating content (to appear as legitimate papers).

征集特刊（SIs）提案的期刊是论文加工厂特别青睐的目标。客座编辑可以提交特刊提案，帮助征集论文，监督并参与同行评议。论文加工厂的策略包括操纵客座编辑、作者甚至审稿人的身份（以真正研究人员的身份出现）和/或编造内容（以合法论文的身份出现）。

The infiltration of an SI within a journal by a papermill often involves some or all of the following steps:

论文工厂渗入期刊内的特刊通常涉及以下部分或全部步骤：

1. Create, manipulate, or hijack the identity of a researcher who will act as the SI Guest Editor, or pay a genuine researcher who is open to participation. The latter can include researchers who do not have expertise in the field of the SI, relying on their valid credentials and academic affiliation to mask the misrepresentation.

1. 创建、操纵或劫持将担任特刊客座编辑的研究人员的身份，或向愿意参与的真正研究人员支付报酬。后者可以包括在特刊领域不具备专业知识的研究人员，依靠他们的有效证书和学术关系来掩盖虚假陈述。

2. Create a proposal for an SI that is related to the scope of the target journal. Proposals that are broad in scope are often put forward, which enables paper mill manuscripts across wideranging topics to be submitted.

2. 提出与目标期刊范围相关的特刊建议。通常会提出范围广泛的建议，这样就能提交主题范围更广的纸质稿件。

3. Create papers for the SI that will look legitimate to non-experts.

3. 为特刊撰写在非专业人士看来合理的论文。

4. Plant inappropriate or unnecessary citations in the text and reference list to boost selfcitations or citations to other researchers.

4. 在正文和参考文献列表中加入不恰当或不必要的引文，以增加自引或对其他研究人员的引用。

5. Create, manipulate, or hijack the identities of authors and peer reviewers; or influence genuine researchers to participate.

5. 创建、操纵或劫持作者和审稿人的身份；或影响真正的研究人员参与。

6. Submit the fabricated papers authored by the fake authors to the SI. Or solicit genuine papers from authors who are willing to pay for an ‘expedited’ route to publication.

6. 将伪造作者撰写的虚假论文提交给特刊。或向愿意花钱购买“快速”发表途径的作者征集真正的论文。

7. Use the fake or compromised peer reviewers to provide comments on the papers that appear legitimate but, like many of the papers, are revealed to be duplicates, unrelated, or otherwise unreliable.

7. 利用伪造的或受到损害的审稿人对论文提出意见，这些论文看似合法，但与许多论文一样，被发现是重复的、不相关的或不可靠的。

8. While this is going on, potentially also sell co-authorships to customers.

8. 与此同时，还可能向客户出售合著者身份。

9. Accept the papers via the fake or compromised Guest Editor and collect payment from the co-authors.

9. 通过假冒的或被篡改的客座编辑接受论文，并向合著者收取报酬。

As indicated above, we note that the deliberate inclusion of inappropriate citations is a frequent feature of paper mills (e.g. Tomentella, 2023). In relation to this, the phenomenon of ‘citation rings’ – where groups of researchers act together to boost the citation of each other’s papers (Bik, 2022) – is another form of publication manipulation at scale, although not discussed further here.

如上所述，我们注意到，故意加入不恰当的引用是论文加工厂的一个常见特征(e.g. Tomentella, 2023)。与此相关，“引文圈”现象——即研究人员群体共同提高彼此论文的引用率(Bik, 2022)是另一种大规模操纵论文发表的形式，但在此不做进一步讨论。

Peer review manipulation

操纵同行评议

Peer review, a cornerstone of scholarly publishing, is designed to ensure the quality and credibility of academic research. However, “the peer review process is showing signs of strain. A growing shortage of skilled reviewers, organised fraud in the form of peer review rings, fake papers and manipulated results, and new challenges brought on by new AI tools and large language models, all pose threats to the integrity of the process” (COPE Forum, 2023).

同行评议是学术出版的基石，旨在确保学术研究的质量和可信度。然而，“同行评议过程正显示出紧张的迹象。熟练审稿人的日益短缺，以同行评议圈、伪造论文和篡改结果为形式的有组织欺诈，以及新的人工智能工具和大型语言模型带来的新挑战，都对同行评议过程的完整性构成了威胁”(COPE Forum, 2023)。

Peer review manipulation and the formation of peer review rings is on the rise. In instances of manipulation, dishonest researchers may engage in various deceptive practices, such as submitting fake contact information for suggested reviewers, providing false email addresses, or even reviewing their own submissions under different identities (Lechner and Evans, 2020; COPE Council, 2021).

操纵同行评议和同行评议圈的形成呈上升趋势。在操纵的情况下，不诚实的研究人员可能会采取各种欺骗行为，如提交虚假的推荐审稿人联系信息、提供虚假的电子邮件地址，甚至以不同的身份审阅自己提交的手稿(Lechner and Evans, 2020; COPE Council, 2021)。

The formation of peer review rings exacerbates the challenge of upholding unbiased peer review by creating networks of individuals willing to provide favorable reviews for each other’s submissions. These collusive efforts undermine the objectivity and impartiality that peer review is intended to uphold. Peer review rings often operate covertly, with members strategically submitting and reviewing each other’s work to ensure acceptance without genuine scrutiny (Ferguson et al, 2014).

同行评议圈的形成加剧了坚持无偏见同行评议的挑战，因为它创建了愿意为彼此的投稿提供好评的个人网络。这些相互勾结的行为破坏了同行评议本应坚持的客观性和公正性。同行评议圈往往暗中运作，其成员有策略地提交和评审彼此的作品，以确保在没有真正审查的情况下被接收(Ferguson et al, 2014)。

Many publishers use multiple third-party platforms to help provide reviewer recommendations and to verify reviewer (and editor) identity . As part of our investigations at Hindawi, we have uncovered instances where those involved in peer review rings and paper mills were deeply embedded in various industry databases and the infrastructure used for selection and verification of peer reviewers.

许多出版商使用多个第三方平台帮助提供审稿人推荐并验证审稿人（和编辑）身份。在对Hindawi的调查中，我们发现了一些情况，即参与同行评议圈和论文加工厂的人深深扎根于各种行业数据库以及用于遴选和验证审稿人的基础设施中。

Section 3: What happened at Hindawi

第3节：在Hindawi发生的事情

SIs can benefit journals by allowing them to promote new emerging fields or cross-disciplinary topics. They also provide researchers from many different disciplines and regions around the world the opportunity to publish the research that is important to them.

特刊可使期刊促进新兴领域或跨学科课题的发展，从而使期刊受益。它们还为来自世界各地不同学科和地区的研究人员提供了发表对其重要的研究成果的机会。

Until October 2022, Hindawi’s SIs program largely involved attracting proposals from researchers volunteering to act as Guest Editors for SIs. Our program utilized a range of checks to assess the expertise of potential Guest Editors and to help verify their identity. These included both manual checks and, like most large academic publishers, the use of third-party tools and systems. Although several downstream checks were in place at multiple steps along the workflow to publication, this process relied to some extent on the inherent trust that has been built up over decades in the publishing industry, which assumes that academics act ethically, rigorously, and impartially as editors and reviewers, and in line with Hindawi’s guidelines for researchers taking on these roles.

在2022年10月之前，Hindawi的特刊计划主要是吸引自愿担任客座编辑的研究人员发起特刊提议。我们的计划利用一系列检查来评估潜在客座编辑的专业知识，并帮助核实他们的身份。这些检查包括人工检查，以及像大多数大型学术出版商一样使用第三方工具和系统。尽管在出版工作流程的多个步骤中都设置了若干下游检查，但这一流程在一定程度上依赖于出版业数十年来建立起来的固有信任，即假定学术界人士作为编辑和审稿人，其行为符合Hindawi对担任这些角色的研究人员的指导原则，是道德、严谨和公正的。

Despite the checks that were in place, suspicious patterns were identified by Hindawi research integrity staff. An initial investigation, prompted by an increasing number of escalations to that team coupled with new analytic capabilities afforded by the in-house peer review system, provided evidence that suspicious activity was identifiable across multiple Hindawi SIs and the findings were escalated to the company’s senior management. Around this time, the scientific community, especially independent researchers with an interest in research integrity, also began noticing indicators of large-scale systematic manipulation.

尽管进行了检查，Hindawi 的研究诚信人员还是发现了可疑的模式。由于向该团队提交的报告数量不断增加，再加上内部同行评议系统提供的新分析能力，初步调查提供了证据，证明在 Hindawi 的多个特刊中可以发现可疑活动，调查结果被上报给公司高级管理层。大约在这个时候，科学界，尤其是对研究诚信感兴趣的独立研究人员，也开始注意到大规模系统操纵的迹象。

Hindawi research integrity staff undertook a more detailed investigation in September 2022, with an initial retraction of 511 articles announced in October 2022 (Kincaid, 2022). Prior to this, Hindawi and Wiley raised the alarm to the publishing industry that we had detected paper mill activity at scale, including alerting other publishers and relevant industry stakeholders of the presence of bad actors in their portfolios and systems.

2022年9月，Hindawi公司的研究诚信工作人员进行了更详细的调查，并于2022年10月宣布撤回511篇文章(Kincaid, 2022)。在此之前，Hindawi和Wiley向出版业发出警报，称我们发现了大规模的论文工厂活动，包括提醒其他出版商和相关的行业利益相关者，在他们的产品组合和系统中存在不良行为者。

These initial investigations coupled with the findings of independent researchers pointed to the infiltration of Hindawi SIs at an even greater scale than first anticipated, making it clear that thousands of manuscripts would need to be investigated. As a result, October 14 2022 saw Hindawi pause the publication of all SI content. We knew we had to both secure our existing workflows and clean up the papers we had already published that had been compromised. We therefore made two key decisions.

这些初步调查和独立研究人员的发现表明，Hindawi的特刊被渗透的规模比最初预计的还要大，因此显然需要对数以千计的手稿进行调查。因此，Hindawi于2022年10月14日暂停了所有特刊内容的发布。我们知道，我们既要确保现有工作流程的安全，又要清理我们已经发表的被泄露的论文。因此，我们做出了两个关键决定。

The first decision was that every SI manuscript yet to be published would be reassessed to prevent the publication of further compromised papers. Suitably qualified editorial staff, and suitably qualified vendors overseen by our own staff, subsequently assessed all SI manuscripts in their entirety, including the use of a comprehensive checklist of specific hallmarks of papermill papers.

第一项决定是对所有尚未发表的特刊稿件进行重新评估，以防止发表更多有问题的论文。随后，具有适当资质的编辑人员和由我们自己的员工监督的具有适当资质的供应商对所有特刊稿件进行了全面评估，包括使用一份关于论文工厂论文具体特征的综合核对表。

At the same time, we carried out a thorough review and strengthening of our existing checks on editors, authors and reviewers.

与此同时，我们还对现有的编辑、作者和审稿人检查进行了全面审查和加强。

Pausing the SIs program was essential to stop the publication of additional compromised content and to ensure that we could have confidence in the quality of future output. This decision had a significant financial impact on the business, which is a matter of public record (Butcher, 2023).

暂停特刊计划对于停止发表更多有问题的内容和确保我们对未来产出的质量有信心至关重要。这一决定对公司的财务状况产生了重大影响，这是有案可查的(Butcher, 2023)。

The second major decision was to investigate the thousands of papers published in Hindawi SIs. The scale of the task necessitated the development of a new protocol aimed at detecting manipulation patterns and retracting papers rapidly and at scale. We go on to describe this approach, which required a significant investment in the resources required to support such a large-scale investigation process, in more detail below.

第二个重大决定是调查 Hindawi的特刊发表的数千篇论文。由于这项任务规模庞大，因此有必要制定一项新的协议，旨在检测操纵模式，并快速、大规模地撤回论文。我们将在下文中详细介绍这种方法，它需要投入大量资源来支持如此大规模的调查过程。

The issues identified within Hindawi SIs during our investigations encompass multiple signs of paper mill activity, including problems and patterns relating to the papers themselves, the citations within them, and the individuals involved with the peer review process.

在我们的调查过程中，在Hindawi 的特刊中发现的问题包含了论文加工厂活动的多种迹象，包括与论文本身、论文中的引文以及参与同行评议过程的个人有关的问题和模式。

The consequences

后果

The consequences of papermill papers infiltrating Hindawi’s systems and reaching publication were severe. At the highest level is the issue of contamination of the scholarly literature and the undermining of trust in journals, publishers, and research itself. Also of crucial concern is the damaging effect on innocent authors who had published genuine papers in the journals affected – journals whose reputations had now been tarnished.

论文工厂的论文渗入Hindawi的系统并得以发表，后果十分严重。最严重的后果是污染学术文献，破坏对期刊、出版商和研究本身的信任。同样令人担忧的是，在受影响的期刊上发表过真实论文的无辜作者所受到的损害——这些期刊的声誉现已受损。

We also recognized that many of those who had perpetrated the abuse within our portfolio had established publication records and profiles elsewhere before engaging with our program.

我们还认识到，在我们的项目范围内，许多滥用权力者在参与我们的项目之前，已经在其他地方建立了发表论文的记录和档案。

Consistent with the ethos that led us to share our early findings with other publishers, we felt a deep responsibility for ensuring that the abuses that had proliferated within our portfolio could be tracked, identified, and made useful to others.

本着与其他出版商分享我们早期发现的精神，我们深感自己有责任确保对在我们项目中泛滥的侵权行为进行跟踪、识别，并使其对他人有用。

As a result, we undertook the complex, resource-intensive process of investigating systematic manipulation at scale, working through the incredibly detailed and involved process of retracting thousands of papers, while balancing fairness to authors with the urgent need to clean up the scholarly record in a timely manner.

因此，我们开展了复杂、资源密集型的工作，大规模调查系统性操纵行为，通过令人难以置信的详细而繁琐的流程撤回数千篇论文，同时兼顾对作者的公平性和及时清理学术记录的紧迫性。

Retractions at scale: a new approach

大规模撤回：新方法

As outlined above, having identified that the Hindawi SI program had been affected at scale, Hindawi implemented a new approach to investigation and retraction necessitated by the magnitude of the problem and the need to correct the scientific record as quickly as possible.

如上所述，在发现 Hindawi的特刊项目受到大规模影响后，Hindawi 采取了新的调查和撤稿方法，这是因为问题的严重性和尽快纠正科学记录的必要性。

Typically, when there is suspected misconduct connected with a single paper or group of papers, the editors of the journal, in consultation with a research integrity team and publisher, will take joint responsibility to investigate and retract a paper. In dealing with manipulation at scale, however, patterns of manipulation must be analyzed across different SIs and journals (Bishop, 2023). The publisher alone has the resource and access to relevant data to enable an investigation at scale across multiple articles and journals simultaneously, and to act on what we discover across the portfolio.

通常情况下，当一篇论文或一组论文涉嫌存在不当行为时，期刊编辑会与研究诚信团队和出版商协商，共同负责调查和撤稿。然而，在处理大规模操纵行为时，必须分析不同的特刊和期刊的操纵模式(Bishop, 2023)。只有出版商才有资源和途径获取相关数据，以便同时对多篇文章和期刊进行大规模调查，并根据我们在整个组合中的发现采取行动。

From January to March of this year, therefore, we developed a new protocol for research integrity investigations at scale, aiming to detect patterns of manipulation across multiple articles simultaneously. Key to the approach was bringing together a dedicated, multifunctional team including experts in research integrity, publishing operations, and data analysis.

因此，从今年1月到3月，我们为大规模研究诚信调查制定了一项新方案，旨在同时发现多篇文章的操纵模式。这种方法的关键在于组建一个专门的多功能团队，其中包括研究诚信、出版运营和数据分析方面的专家。

It was also crucial to take an evidence-informed approach and to adhere as closely as possible to the principles that COPE uses to guide investigations of authors and potential retractions. Hindawi has been open with COPE about the approach, and COPE themselves have recently released specific guidelines to deal with paper mills (COPE Council, 2023b).

同样至关重要的是，要采取基于证据的方法，并尽可能严格遵守 COPE 用于指导作者调查和潜在撤稿的原则。Hindawi一直与 COPE保持开放的态度，而COPE本身最近也发布了处理论文工厂问题的具体指导原则(COPE Council, 2023b)。

An evidence-informed approach

基于证据的方法

The evidence we have been using to retract a paper in this new at-scale approach includes one or more of the following indicators of manipulation:

在这种新的大规模方法中，我们用来撤稿的证据包括以下一个或多个操纵指标：

1. Discrepancies in scope.

1. 论文的主题与期刊或特刊的主题范围不一致。

2. Discrepancies in the description of the research reported.

2. 所报告的研究前后描述不一致。

3. Discrepancies between the availability of data and the research described.

3. 数据可用性声明与所描述的研究之间不一致。

4. Inappropriate citations.

4. 不当引用。

5. Incoherent, meaningless, and/or irrelevant content included in the article.

5. 文章内容不连贯、无意义和/或不相关。

6. Compromised or manipulated peer-review.

6. 同行评议被损害或操纵。

The decision to retract, given this evidence, is based on the rationale that the publication process has been undermined and we can no longer vouch for the integrity of the article. We have intentionally limited the specific details of what is under investigation in the retraction notice, in part because it is essential that the intelligence shared with bad actors is restricted. Moreover, to proceed expeditiously and communicate with thousands of authors, standardized wording of retraction notices was essential.

鉴于这些证据，我们决定撤稿的理由是：文章的发表过程受到了破坏，我们无法再为文章的完整性提供担保。我们在撤稿声明中有意限制了调查的具体细节，部分原因是必须限制与不良分子共享的情报。此外，为了尽快与成千上万的作者进行沟通，撤稿声明的标准化措辞是必不可少的。

The investigation workflow combined checks performed by suitably qualified and trained individuals alongside an analysis of data from internal systems and data shared publicly by independent research integrity experts. To proceed at a reasonable pace, we allocated work to PhD-qualified staff at vendors. Each paper was assessed by a minimum of two people, and assessed a third time if arbitration was needed. Internal staff with expertise trained the vendors, and checked for consistency and accuracy, doing additional investigation where needed.

调查工作流程结合了由具备适当资格并经过培训的人员进行的检查，以及对来自内部系统的数据和独立研究诚信专家公开共享的数据进行的分析。为了以合理的速度开展工作，我们将工作分配给供应商中具有博士学位的员工。每篇论文至少由两人进行评估，如果需要仲裁，则进行第三次评估。具有专业知识的内部员工对供应商进行培训，检查一致性和准确性，并在必要时进行补充调查。

Responses were collected via a software application we developed specifically for this purpose. In parallel, computational tools were used to provide additional supporting evidence about fabricated content, plagiarized peer-review reports, and suspect peer review turn-around times. We also built on the valuable work done by independent research integrity sleuths, by collecting, evaluating, and categorizing comments provided on PubPeer as part of Smut Clyde’s list.

我们通过专门为此开发的软件应用程序收集答复。与此同时，我们还使用计算工具来提供更多的辅助证据，以证明编造的内容、剽窃的同行评议报告和可疑的同行评议周转时间。我们还以独立研究诚信调查员所做的宝贵工作为基础，收集、评估和分类了PubPeer上提供的评论，作为Smut Clyde的名单的一部分（译者注：即6666条记录的电子表格）。

The bespoke software application guided investigators through the questionnaire and enabled the same paper to be reviewed by different teams to help ensure consistency. All the variables contributed to an overall score for the paper, which was used by Hindawi to make the final decision about whether the paper should be retracted. Thresholds were assigned to the scores and any article that reached these thresholds were retracted.

这款定制的软件应用程序可指导调查人员完成问卷，并使同一篇文章能够由不同的团队进行审核，以帮助确保一致性。所有变量都会对论文产生一个总分，Hindawi根据该总分最终决定是否撤回论文。得分设有阈值，达到阈值的文章将被撤回。

As we gathered the evidence, we have also been able to systematically pinpoint particular bad actors - Guest Editors who have been responsible for handling multiple retracted papers. We have since sent sanction letters to several hundred Guest Editors notifying them that they can no longer take on an editorial role in our journals or publish with us in future.

在收集证据的过程中，我们还系统地找出了某些不良行为者——负责处理多篇被撤论文的客座编辑。此后，我们向数百名客座编辑发出了制裁信，通知他们今后不得再在我们的期刊上担任编辑或发表文章。

What are we doing now?

我们现在在做什么？

To date, thousands of papers have been retracted from Hindawi SIs as a result of our large-scale investigation and we continue to investigate and issue retractions where appropriate. In some cases, we suspect that additional papers in particular SIs are likely to be problematic. Where these papers have not yet been investigated – or where we do not have specific indications of problems with individual papers – we plan to issue Expressions of Concern at the level of whole SIs to alert readers that they should take additional care interpreting all papers in the SI – whether retracted or not.

迄今为止，经过我们的大规模调查，已有数千篇论文从 Hindawi 的特刊上撤下，我们将继续调查并在适当的情况下发布撤稿声明。在某些情况下，我们怀疑某些特刊中的其他论文可能存在问题。在尚未对这些论文进行调查的情况下，或者在我们没有具体迹象表明个别论文存在问题的情况下，我们计划在整个特刊的层面上发布“关注声明”，以提醒读者在阅读该特刊的所有论文时应格外小心——无论这些论文是否被撤回。

Peer review processes – both at Hindawi and more widely across Wiley – are being continually strengthened through increased controls and expert staffing, new workflows, and AI-based tools that are being developed and refined to significantly improve manuscript screening. Much more stringent checks on all SI proposals and Guest Editors have been introduced to provide a clearer understanding of Editor identity, research experience, publication history, and indicators of previous unethical publishing activity before allowing those researchers to lead SIs going forwards. Informed by the learnings of our investigations, new checks have been introduced at multiple steps in the publication process, including much greater scrutiny of peer review.

Hindawi和整个Wiley的同行评议流程都在不断加强，包括增加控制和专家人员配备、新的工作流程，以及正在开发和完善的人工智能工具，以显著提高稿件筛选能力。我们对所有特刊提案和客座编辑进行了更为严格的检查，以便更清楚地了解编辑的身份、研究经验、出版历史以及以往不道德出版活动的迹象，然后再允许这些研究人员继续领导特刊。根据我们的调查结果，在出版流程的多个步骤中都引入了新的检查措施，包括对同行评议进行更严格的审查。

Section 4: Conclusion and call to action

第4节：结论和行动呼吁

The need for a collaborative approach

合作的必要性

Painful lessons have been learned over the past year, by Hindawi and several other publishers. We are committed to acting in the interests of all in countering the increasing threat and sophistication of those who undermine the integrity of published research for their own gain.

在过去的一年里，Hindawi 和其他几家出版商吸取了惨痛的教训。我们致力于从所有人的利益出发，打击那些为一己私利而破坏已发表研究成果完整性的人所带来的日益严重的威胁和他们的狡猾手段。

It is vital that publishers and other stakeholders (including funders, institutions, and indexers) continue to come together to share intelligence on the activities of papermills and other bad actors, update each other on their approaches to investigation and retraction, and share new processes to combat publication misconduct at scale. For example, the STM Integrity Hub is enabling this collaboration by acting as a “knowledge exchange: where publishers may share experiences and learnings regarding how best to safeguard research integrity in science; a think tank for policy and legal frameworks; and a living library of infrastructure and tools” (STM, 2023).

至关重要的是，出版商和其他利益相关者（包括资助者、机构和索引编制者）应继续携手合作，共享有关论文工厂和其他不良行为者活动的情报，相互更新调查和撤稿的方法，并分享新的流程，以大规模打击出版不当行为。例如，STM Integrity Hub 通过充当“知识交流中心：出版商可在此分享有关如何最好地保障科学研究诚信的经验和教训；政策和法律框架智库；以及基础设施和工具的活图书馆”(STM, 2023)来促进这种合作。

The infrastructure that they have built, and the knowledge-sharing communities established to support prioritization, are essential components of translating that collaboration into practice.

他们建立的基础设施，以及为支持确定优先事项而建立的知识共享社区，是将合作转化为实践的重要组成部分。

There is important work underway led by the United 2 Act initiative, initially conceived through a forum set up by COPE and the STM Association, bringing stakeholders together to establish areas of focus and prioritization. We will be actively working with this diverse group and developing a shared strategy to mitigate the impact of systematic manipulation. It is critically important that we continue to speak out and share what we have learned as a community, whether in our roles as journal editors, publishers, or researchers with an interest in research integrity. COPE provided an excellent forum for this community participation during its Publication Integrity Week held in October of this year.

由 “United 2 Act initiative”倡议牵头的重要工作正在进行之中，该倡议最初是通过促进平等国 际委员会和科技管理协会设立的论坛构想出来的，它将利益相关方聚集在一起，以确定 重点领域和优先次序。我们将积极与这一多元化团体合作，制定共同战略，以减轻系统性操纵的影响。至关重要的是，无论是作为期刊编辑、出版商，还是对研究诚信感兴趣的研究人员，我们都应继续大声疾呼，分享我们作为一个群体所学到的知识。中国出版商协会在今年 10 月举办的“出版诚信周”上为这一群体的参与提供了一个绝佳的论坛。

Standardization across research publishing is also essential to establishing a cohesive approach to upholding research integrity. The work that the National Information Standards Organization (NISO) is leading to establish best practices around post-publication amendments, their publication, and consistent metadata has been available for public consultation since October (NISO, 2023).

整个研究出版领域的标准化对于建立维护研究诚信的统一方法也至关重要。美国国家信息标准协会 (NISO) 正在牵头开展工作，围绕出版后修订、出版和一致的元数据建立最佳实践，这项工作从10月份开始接受公众咨询(NISO, 2023)。

One consequence of what we have learned from the experience of Hindawi is a growing awareness of the vulnerability to manipulation not just of our own workflows but also that of the third-party services we rely on and the downstream platforms that provide services and metrics for scholarly communication more widely. For example, we, like many other publishers, rely on trusted third-party providers to help find recommendations for reviewers. Such platforms often rank reviewer performance and reliability based on the number of reviews an individual has done.

我们从Hindawi的经验中学到的一个结果是，我们越来越意识到，不仅我们自己的工作流程容易被操纵，我们所依赖的第三方服务以及为学术交流提供服务和衡量标准的下游平台也容易被操纵。例如，我们和许多其他出版商一样，都依赖于可信的第三方供应商来帮助寻找推荐审稿人。这些平台通常会根据审稿人的审稿数量对审稿人的表现和可靠性进行排名。

If these platforms are infiltrated with bad actors using false or multiple identities, they can in turn infiltrate the reviewer databases of many publishers simultaneously. In addition, downstream services that provide information about author identity or that rely on citations and mining the literature to evaluate researchers will also be impacted. To date, there is little published evidence of the impact of wide scale manipulation on the scholarly infrastructure, although a recent analysis of ‘sneaked references’ (Lonni, 2023; Chawla, 2023) and the impact of ‘hijacked’ journals on bibliographic databases points to the insidious damage this can cause (Abalkina, 2023). Given the scale of manipulation that is emerging across multiple publishers (van Noorden, 2023), the knockon effects are therefore likely to impact both researcher evaluation and knowledge discovery much more widely. There is a pressing need to take collective responsibility and collective action.

如果这些平台被使用虚假或多重身份的坏人渗透，他们就会同时渗透到许多出版商的审稿人数据库中。此外，提供作者身份信息或依靠引用和文献挖掘来评估研究人员的下游服务也会受到影响。尽管最近对“偷偷引用”(Lonni, 2023; Chawla, 2023)和“被劫持的”期刊对书目数据库的影响进行了分析，但迄今为止，关于大范围操纵对学术基础设施影响的公开证据并不多(Abalkina, 2023)。鉴于多个出版商都在大规模操纵期刊(van Noorden, 2023)，其连锁反应可能会对研究人员的评价和知识发现产生更广泛的影响。我们迫切需要承担集体责任，采取集体行动。

In summing up, we hope that our message is clear: publishers need to work together – and with all stakeholders in the scholarly publishing ecosystem – and devote substantial resources to ensure the integrity of our journals and the content we publish . From the experience of Hindawi, we have learned a great deal and offer below our recommendations for other publishers and stakeholders :

总之，我们希望我们传达的信息是明确的：出版商需要与学术出版生态系统中的所有利益相关者共同努力，投入大量资源，确保我们的期刊和出版内容的完整性。从 Hindawi 的经历中，我们学到了很多，下面是我们对其他出版商和利益相关者的建议：

• We need to support the spirit of a trust-based research publishing ecosystem in the form of a ‘trust-but-verify’ system, in which participants have independently verified and secure identities that link accounts to real individuals.

- 我们需要以“信任但验证”系统的形式支持基于信任的研究出版生态系统的精神，在该系统中，参与者拥有独立验证的安全身份，并将账户与真实的个人联系起来。

• Publishers should continue – and accelerate – the responsible sharing of data and techniques, analogous to the successful work done within the financial industry to combat fraud. Such sharing – both between publishers and with other appropriate stakeholders – is crucial to prevent the momentum of bad actors and to preserve the quality of the scholarly literature.

- 出版商应继续并加快、负责任地共享数据和技术，类似于金融业为打击欺诈所做的成功工作。这种共享，包括出版商之间的共享以及与其他相关利益方的共享，对于防止不良行为者的势头和维护学术文献的质量至关重要。

• Publishers should continually review their vetting processes for editorial board members and any Guest Editors, alongside careful consideration of the responsibilities given to Guest Editors.

- 出版商应不断审查其对编委会成员和任何客座编辑的审查程序，同时认真考虑赋予客座编辑的责任。

• Publishers should ensure that retraction notices are linked to the original article by updating the metadata of the article and made publicly discoverable as soon as possible to readers and machines. Ideally this should be done through centralized community-run metadata services such as Crossref so that downstream service providers can also ensure their platforms, services, and records are up to date (see also NISO, 2023).

- 出版商应通过更新文章的元数据，确保撤稿声明与原文链接，并尽快向读者和机器公开。理想情况下，这应该通过集中式社区元数据服务（例如Crossref）来完成，这样下游服务提供商也能确保其平台、服务和记录是最新的(see also NISO, 2023)。

• We need to establish best practices around when to add a notice or flag to a set of papers that carry hallmarks of a potential paper mill while an investigation is underway. This gives balance to our responsibility to investigate while also notifying the community that valid concerns have been raised. In the same spirit, we need to continue to commit resource to the efficient investigations and their outcomes.

- 我们需要建立最佳实践，以确定在调查进行期间，何时对具有潜在论文工厂特征的论文集添加通知或标记。这样既能平衡我们的调查责任，又能通知社会已经提出了合理的担忧。本着同样的精神，我们需要继续为高效的调查及其结果投入资源。

• Publishers, and third parties on whom publishers rely for data and identity verification, need to recognize the challenges inherent in this environment – including the rapid developments with artificial intelligence – and upgrade their capabilities accordingly.

- 出版商以及出版商在数据和身份验证方面所依赖的第三方需要认识到这一环境中固有的挑战——包括人工智能的快速发展——并相应地提升自己的能力。

• All stakeholders need to incentivize open research practices such that the focus is not only on the final published article; this would both bring increased transparency to the research process and act as a means of mitigating research integrity concerns.

- 所有利益相关者都需要鼓励开放式研究实践，使人们不仅关注最终发表的文章；这既能提高研究过程的透明度，也是减轻研究诚信问题的一种手段。

While Hindawi and Wiley have taken many concrete actions in both the short- and long-term, there is much to do. The issues covered in this white paper are multi-faceted and stretch across the entire scholarly publishing ecosystem. The hope is that in sharing these insights, a useful perspective and suggestions of action have been provided, regarding an issue threatening to impact the future health of our industry and, critically, society’s trust in scholarly research.

尽管Hindawi和Wiley已经采取了许多短期和长期的具体行动，但仍有许多工作要做。本白皮书所涉及的问题是多方面的，横跨整个学术出版生态系统。我们希望通过分享这些见解，为我们提供一个有用的视角和行动建议，因为这个问题可能会影响我们行业未来的健康发展，关键是会影响社会对学术研究的信任。

Authorship and acknowledgments

作者和致谢

Louisa Flintoft, Senior Publishing Director at Wiley

Catriona J. MacCallum, Director of Future Scholarly Research at Wiley

Mike Streeter, Director of Research Integrity Strategy and Policy at Wiley

Dave Flanagan, Senior Director of Data Science at Wiley

Liz Ferguson, Senior Vice President, Research Publishing at Wiley

With special thanks

特别鸣谢

Naomi Blumsom

Charlotte Crouch

Ben Dickinson

Emily Horne

Joyce Griffith

Adam Price-Evans

As well as the support from our peers, we appreciate the commitment of the academic community, especially the independent research integrity experts who devote their time and expertise to protecting the scholarly record. Some have pseudonyms such as Smut Clyde, Parashorea Tomentella, Rhipidura Albiventris and Huanzi Zhang, and others work under their own names such as Anna Abalkina, Elisabeth Bik, Dorothy Bishop, and Nick Wise. They have worked tirelessly to raise the alarm about the scale of the systematic manipulation at Hindawi and across the industry. That commitment is a critical line of defence in ensuring the authenticity and integrity of the world’s research. We would also like to acknowledge the advice and expertise of two independent research integrity consultants, Drs Jigisha Patel and Margaret Rahmati, who helped us in the development phase of our new protocol for retractions at scale.

除了来自同行的支持，我们还感谢学术界的承诺，尤其是独立研究诚信专家，他们为保护学术记录奉献了自己的时间和专业知识。他们有的化名为Smut Clyde、Parashorea Tomentella、Rhipidura Albiventris和Huanzi Zhang，还有使用真名的Anna Abalkina、Elisabeth Bik、Dorothy Bishop和Nick Wise。他们孜孜不倦地工作，为Hindawi和整个行业的系统性操纵行为敲响了警钟。这一承诺是确保世界研究真实性和完整性的重要防线。我们还要感谢两位独立研究诚信顾问Jigisha Patel博士和Margaret Rahmati博士的建议和专业知识，他们在我们制定新的大规模撤稿协议阶段为我们提供了帮助。