Future of AI

How AI can help the scientific process

Philipp Koellinger, CEO and Co-Founder of DeSci Labs

AI is seen by many political and economic leaders as key to revolutionising economies. New advances with AI models such as DeepSeek caught the headlines, and wider use of ChatGPT has many organisations and business looking at how AI can change or assist their work. For example, scientific publishing has emerged as an area that could benefit from integrating AI.

For scientists globally, getting work published is a long, tiresome process. Researchers, who are often peer reviewers themselves, spend a lot of time developing their work to suit specific high-impact journals only to enter what feels like publication purgatory because of the lengthy and opaque nature of the review process.

Peer review problems

The peer review process relies on volunteers who donate their time, either out of a sense of charity or because of a perceived duty of reciprocity. Researchers often feel that, in exchange for their own work being reviewed and published, they owe it to their colleagues to review their manuscript submissions.

As a result of the exponential increase in the number of published scientific articles, the demand for peer review has grown considerably in recent years. A study from Publons[1] suggests that peer review completion rates have been reducing year-on-year, with editors increasing the number of invitations they are required to send.

Scientists donate their time to journals to conduct peer review for them for free. In 2020, it was estimated that over 100 million hours were spent on peer review, with an approximate monetary value of reviewers’ time exceeding $2.5 billion between the US, UK and China.[2]  This is a substantial cost to the scientific community and each individual referee, which often results in referees to de-prioritize their work on reviews. In turn, this leads to long waiting times for the authors, who typically have to wait three months or more to hear back from a journal about their submission.

AI offers a lot of possibilities to improve the situation, for example, by helping editors to find suitable referees and checking certain properties of a contribution automatically, aiding both editors and referees in their work. In addition, AI tools can provide instant feedback to the researcher that can help them to improve their work and to find a suitable publication outlet.

Novel works

A common reason for rejecting submissions at many journals is a perceived lack of novelty. However, as it stands, assessing novelty is a subjective and time-consuming process that often results in different referees offering different opinions. Instead of relying entirely on the subjective opinions of a few people, novelty can also be assessed with a mathematical measure of a paper’s novelty, which measures how surprising the combinations of topics and citations are compared to all previously published work.

Furthermore, AI enables users to find closely related to work to a particular article quickly. This can help authors and readers find relevant papers they may have missed so far, offering a quick tool to better understand the current state-of-the-art on a particular research topic. Furthermore, this can help authors refine their publication strategy by getting information about which journals recently published related work.

The process

There are lots of tasks in the scientific process that can take scientists a long time to perform. Data wrangling and formatting can easily be facilitated with several mainstream tools, while AI is continuously improving its capability of searching for relevant references and summarising information.

In addition, as an assistive technology, AI can teach scientific researchers how to ask better questions. AI can scan many papers quickly and find gaps in the literature that can identify aspects of a subject area that requires further investigation.

By providing language models with more examples of questions, ideas and hypotheses, they can train and fine-tune these, learning from them very quickly. Language models can be trained to assist you in providing novel insights, sharpen ideas to make them more actionable and impactful, and help decide experiments that can effectively put those ideas to the test. Models such as ChatGPT can often provide relatively good feedback at a much higher rate than your colleagues could.

DeSci Labs is developing numerous tools to help accelerate the scientific process, particularly regarding publishing work. Its mathematical novelty scores metric, the first publicly available novelty score calculator, can instantly the content and context novelty of works when uploaded to its DeSci Publish platform. Having high novelty scores can help authors to quickly get the attention of the editors at their target journal, increasing publication chances.

With more AI tools in the pipeline, including a related article finder and future impact predictions, DeSci Labs is helping scientists at all stages of the research and publication process to accelerate scientific progress.

To find out how quickly you can receive measures to help you accelerate your work to publication, sign up to DeSci Publish.

Author

Related Articles

Back to top button