香港六合彩官网

Author: Chris Graf

Research Integrity Director

The vast majority of researchers undertake and submit for publication research that is conceived and done well and honestly, contributing mostly incrementally and sometimes substantively to the advances that societies and communities around the world benefit from. However, unethical content does sometimes reach our door and, as the volume and complexity of scholarly content continues to grow, so too do the challenges we face as a sector, as well as the opportunities for maintaining trust in the scholarly record. That鈥檚 why we鈥檙e continuing to invest in innovative technologies and ways of working that help us to detect and prevent threats to research integrity before they reach publication. 

Over the past year, we鈥檝e made real strides in , many of which use artificial intelligence (AI), to support our editors, reviewers, and authors. These tools are designed not to replace human judgment, but to enhance it鈥攂y identifying signals of potential manipulation or integrity breaches early in the submission process, and by freeing-up valuable time for deeper editorial assessment. 

A new tool to detect non-standard phrases 

Our latest addition is a in submitted manuscripts. These are phrases that appear awkward or overly convoluted, such as 鈥渃ounterfeit consciousness鈥� instead of 鈥渁rtificial intelligence.鈥� These unusual constructions, known as 鈥渢ortured phrases鈥�, can be a sign that paraphrasing tools have been used to disguise plagiarism or manipulate content. 

We use our non-standard phrases tool to identify if a submission contains a significant number of these non-standard phrases. If it does the manuscript will be withdrawn, protecting the time and expertise of our editors and peer reviewers and preventing problematic papers with this signal from entering the scholarly record. 

The tool draws on the publicly available catalogue of tortured phrases developed by Guillaume Cabanac, Cyril Labb茅, and Alexander Magazinov as part of their Problematic Paper Screener (PPS), an online platform that identifies a number of indicators of potentially problematic research articles. Their work has been valuable for the research and research publishing sector in identifying papers that may have been generated or manipulated using questionable methods. We鈥檝e built on their efforts by integrating this approach into our editorial workflows. 

Complementing a growing suite of integrity tools 

The non-standard phrases detector joins a growing family of 香港六合彩官网-developed tools that we use to support research integrity in our publishing: 

• This tool, previously known as 鈥淕eppetto鈥� internally, identifies manuscripts that contain AI-generated nonsense text.  
• Our image integrity tool, SnappShot, uses AI to detect duplicated or manipulated images in gels and blots, with ongoing work to expand into microscopy.  It also checks for image plagiarism against a growing library of known problematic content.  
• This tool flags citations that appear unrelated to the place in the manuscript鈥檚 content where they appear 鈥攁nother potential indicator of manipulation. If a number of references are identified as irrelevant, the submission is flagged for manual review by a member of our expert team. 

Each of these tools has been developed in-house by our research integrity and technology teams, and each has undergone rigorous testing and validation. Together, they form a robust, multi-layered, ongoing approach to safeguarding the quality and credibility of the research we publish. 

Growing potential from third-party tools  

Innovators and software developers with research integrity solutions that we, and other publishers, can plug-in and use are a welcome and relatively recent arrival (although of course we have been using the plagiarism detection software iThenticate from Turnitin via the CrossRef Similarity Check service for many years). With a thoughtful approach to using these tools, we can respond to opportunities and close technology gaps鈥攚ithout compromising quality or agility.  

We prefer to build tools where direct integration with our editorial and peer review platform (Snapp) offers performance advantages and long-term control. But partnering offers options when building is too risky or slow, and when third-party tools offer unique capabilities, including pre-existing integrations with other editorial and peer review platforms we use. 

For example, we continue to develop SnappShot for integration with Snapp and broad use, and in the meantime we use Proofig, including to screen images in life sciences manuscripts in Nature and all of the Nature Research journals. 

Collaboration  

The structured approach we take to integrity technology ensures that every investment鈥攚hether built in-house or delivered through partnerships鈥攃ontributes meaningfully to the overarching goal: protecting the integrity of research at scale. But to really do that, we believe that collaboration is essential to combat challenges that are industry-wide. That鈥檚 why we鈥檙e active contributors to the STM Integrity Hub, a cross-publisher initiative that facilitates data sharing and joint development of integrity tools. By working together, we can respond more effectively to emerging threats and share best practices across the industry.  As part of this commitment, , where it can be used across the sector to help detect problematic submissions. 

We also continue to invest in our expert research integrity team, which has grown significantly in recent years. Importantly, the members of this specialist team collaborate with the wider research community鈥攊ncluding our editors, peer reviewers, and others鈥攖o ensure that our tools are used responsibly and that decisions are made with care and in the right context. 

Looking ahead 

As the landscape of scholarly publishing evolves, so too must our approaches to research integrity. The rise of generative AI, the increasing sophistication of paper mills, and the pressures faced by researchers all contribute to a complex environment. We鈥檙e confident we鈥檙e meeting these challenges with the right people, partnerships and tools鈥攁nd we are also always open to fresh ideas and innovation.   

Our goal remains the same: to ensure that the research we publish is trustworthy and meaningful and, as of this year, we鈥檙e publishing our annual submission, publication and retraction statistics alongside further information about research integrity to help illustrate that work. By combining technological innovation with human expertise, we believe we are meeting that goal鈥攁nd we鈥檙e committed to continuing this work into the future, with new tools and approaches on the horizon, all backed up by highly skilled teams of fantastic people.