bims: rfi pathways to ai enabled research

expertise sharing systems

Biomed News is an AI-driven platform designed to manage the overwhelming influx of biomedical research abstracts published weekly. Traditional retrieval methods like prompts and targeted searches often take up significant time and often yield broad and irrelevant results, making it difficult for researchers, clinicians, policymakers, and the public to access and stay abreast of pertinent information. Biomed News addresses this important issue by providing a ranked list of abstracts weekly, curated by global experts across a wide range of career stages. These experts (“selectors”) select relevant abstracts, which are then used to train a supervised machine learning algorithm to identify and rank pertinent papers for subsequent issues. The tool builds on a historical context of AI in bibliographic retrieval, tracing back to initiatives like the RePEc digital library and NEP: New Economics Papers, which emerged in the 1990s. RePEc’s bespoke software, ernad, developed in 2003, was the first AI-driven bibliographic tool and serves as the foundation for Biomed News. Launched in 2017 Biomed News leverages this technology to simplify the discovery of new biomedical knowledge, making it readily accessible and shareable. With proper marketing and increased support, Biomed News can become a vital resource for the dissemination of new biomedical research.

Biomed News enhances the scientific process by facilitating efficient literature management through AI-driven curation. Biomed News leverages AI to sift through the massive influx of 30,000 weekly biomedical research abstracts on PubMed. This technology identifies and ranks relevant papers, providing a streamlined and efficient way for users to stay updated, reducing the noise from irrelevant data. Moreover, the platform’s free open access system disseminates weekly reports, making scientific updates accessible to experts and non-experts alike, thereby broadening the impact of scientific knowledge. Inspired by the NEP and RePEc models in economics, Biomed News aims to create a sustainable open science practice in biomedicine. The future potential includes developing machine-processable literature review and knowledge translation objects, which could be compiled and used in various innovative ways. These objects would consist of accepted and rejected documents, capable of being compiled by Boolean operators, setting a new standard for sustainable open science practices. Thus, Biomed News helps facilitate the scientific process by accelerating knowledge dissemination and translation, enhancing the opportunity for collaboration, as well as expediting and informing patient care.

Efficient Literature Management: With 30,000 new weekly publications on PubMed, keeping abreast of new developments is a daunting task. Biomed News makes it easier for researchers, clinicians, policymakers, and the public to access the most pertinent and up-to-date information. This efficiency surpasses traditional search-based tools, ensuring that users stay current with minimal effort.
Public Record and Recognition: Biomed News allows experts to share their curated lists, making a public record of their contributions. This not only acknowledges their efforts but also demonstrates their expertise in their field thus promoting collaboration and trust within the scientific community.
Bespoke Mailing List System: The tool's mailing list system disseminates weekly reports to subscribers, facilitating regular and structured updates on the latest research, thus enhancing the accessibility of scientific information, ensuring that even those outside the core expert community can benefit from curated content.
Commitment to Open Science: Despite financial challenges, the creators of Biomed News, remain committed to the open science ethos providing free access to the tool.

Biomed News mitigates potential risks associated with its AI technology through several key strategies. It avoids the use of neural networks, which are prone to generating misleading information, ensuring the integrity of the curated content. Human editors, experts in their fields, play a crucial role in manually selecting relevant abstracts, providing oversight and ensuring that new and innovative ideas are highlighted. The machine learning algorithms used incorporate safeguards against overfitting, enhancing the system's ability to accurately detect and rank relevant documents. This synergy between AI and human judgment ensures that groundbreaking research is not missed. Furthermore, the platform maintains transparency by making its processes and results openly available, allowing for community scrutiny and fostering trust. By balancing AI capabilities with human expertise, Biomed News effectively reduces the risk of errors and enhances the discovery of novel biomedical research, supporting the principles of open science and expanding access to the scientific process.

Since its inception Biomed News has made significant strides in facilitating access to scientific literature. The platform is built on a robust technology stack comprising two main components: ernad and nitpo. Ernad, implemented since 2018, separates procedural code (Python, Perl, and JavaScript) from descriptive code (XSLT), offering a flexible and efficient architecture. Biomed News utilizes ernad for front-end services to editors and readers and back-end AI operations. These services integrate with external software to compile new additions from PubMed. Nitpo, funded by a 2021 grant from NlNet, manages email distribution of report issues to subscribers, customizing emails to filter duplicates and avoid redundancy, enhancing user experience. Overall, this platform has successfully implemented advanced AI and human curation to streamline the discovery and dissemination of research, demonstrating considerable potential for future growth and impact.