Scientists are increasingly using generative AI (GenAI) to generate new ideas, draft research proposals and support scientific writing. As a result, the time and effort required to produce funding applications and journal submissions is decreasing.<\/p>\n\n\n\n
This growing volume of scientific output is placing additional pressure on the peer-review system, which was already facing significant challenges. In response, reviewers are increasingly turning to AI tools to assist in evaluating this expanding body of AI-supported research.<\/p>\n\n\n\n
Existing studies have shown that AI can underperform when assessing highly novel scientific ideas and proposals. Using real-world proposal data, our research examines what happens when GenAI is introduced on the reviewer side of the evaluation process.<\/p>\n\n\n\n
Our findings suggest that, without appropriate safeguards, GenAI may favour more conventional ideas over highly novel ones. Importantly, we also test and identify an approach that can help mitigate this bias.<\/p>\n","protected":false},"template":"","solution":[147],"sector":[106,132,261],"class_list":["post-67807","report","type-report","status-publish","hentry","solution-evaluation-sv","sector-digitalisation-sv","sector-research-innovation-sv","sector-science-technology-and-innovation-sti-sv"],"acf":[],"yoast_head":"\n