Scientific publishers will increasingly have to lean on artificial intelligence tools to identify research misconduct, a conference has heard.

Speaking at the Times Higher Education Asia Universities Summit in Macao, Anders Karlsson, vice-president of global strategic networks at Elsevier, said, as global spending on research and development has increased, so has the number of articles being submitted to publishers.

The number of articles, reviews and conference papers has grown from 2.6 million per year in 2014 to 3.7 million in 2023, he said.

As of 2022, more than 50 per cent of scientific publications have authors in low-income or middle-income economies ¨C which includes India and China ¨C compared?with 1995 when this figure was 13 per cent.

ÍøÆØÃÅ

Karlsson said Elsevier reached nearly 3.5 million submissions each year, of which the organisation publishes about 15 per cent following peer review.

Although it is a ¡°good thing¡± that interest in science is ¡°increasing¡± and becoming ¡°more inclusive¡± across the world, he said, more articles being published overall - irrespective of where they are from - puts?¡°the system under pressure¡± and increases the?need to make sure research integrity is maintained.

ÍøÆØÃÅ

The publisher has hired more employees to investigate research misconduct and is also investing in AI tools to support this.

¡°We need to be prepared to handle quantity,¡± agreed Stephan Kuster, director of external affairs at Frontiers. ¡°The input is going up, and that is a good thing ¨C we need more science, not less.

¡°So the quantity and the volume is not going away, but we need to assess and reward quality and not reward the quantity per se,¡± he continued. ¡°There¡¯s definitely a role for publishers as gatekeepers of the quality of what ends up being in the permanent scientific record.¡±

However, he added, ¡°a lot of the misconduct¡± encountered is ¡°not detectable by human eyes¡±.

ÍøÆØÃÅ

¡°It¡¯s not going be detected in peer review or by research integrity experts in your editorial office. You need technology to detect this, and that means AI.¡±

Kuster said one role for AI is ¨C somewhat ironically ¨C to help detect cases of misuse of AI in research.

Additionally, ¡°AI can very quickly read thousands of manuscripts [and] detect patterns that will point towards paper mills, for example,¡± he continued, adding that the technology can also be used to identify plagiarism cases and conflicts of interest.

¡°That kind of thing that usually¡­takes days or weeks. This can be done now in a matter of minutes,¡± he said.

ÍøÆØÃÅ

Speaking at a separate session at the conference, Dhruv Galgotia, chief executive of Galgotias University in India, said the ¡°race for volume¡± in research is ¡°plaguing¡± India, adding that it needs to ¡°shift to a race to quality¡±.

Praveena Nair Sivasankaran, director of the clean technology impact lab at Taylor¡¯s University in Malaysia, said when it comes to keeping up with ¡°research superpowers¡±, the Global South should ¡°not follow the mould by the Global North¡±.

ÍøÆØÃÅ

¡°We can create our own framework,¡± she continued. ¡°It is time we take charge of this, to have localised solutions to create more impact-based research and not just keep up. It¡¯s time for us to lead.¡±

helen.packer@timeshighereducation.com