MetaWriter: Exploring the Potential and Perils of AI Writing Support in Scientific Peer Review

DOI

Recent advances in Large Language models (LLMs) show the potential to significantly augment or even replace complex human writing activities. However, in a complex writing task where people need to make decisions as well as a justification, it is not clear whether the LLMs can make the writing tasks faster or hinder users’ agency. In this paper, we explored the trade-offs of providing intelligent support for writing meta-review as part of the academic peer review process. We prototype a system “MetaWriter” trained on 5 years of open peer review data to support meta-reviewing: The system highlights common topics in the original peer reviews, extracts key points by each reviewer, and on request provides a preliminary draft of a meta-review that can be further edited. To understand whether and to what extent meta-reviewers both experienced and less experienced see value in writing support for this task, we conducted a within-subject study with 32 participants. Each participant wrote meta-reviewers for two papers: one with and one without MetaWriter. We found that MetaWriter significantly expedited the meta-review authoring process and improved the coverage, as rated by experts, compared to the baseline In a within-subjects experiment, 32 participants wrote meta-reviews using MetaWriter and a baseline environment with no machine support. We found that MetaWriter significantly expedited the meta-review authoring process and improved the coverage, as rated by experts, compared to the baseline. While participants recognized the efficiency benefits, they raised issues and concerns around trust, over-reliance, and agency. We also interviewed six paper authors to understand their opinions of using machine intelligence to support the peer review process and reported critical reflections. We discuss implications for future interactive AI writing tools to support complex synthesis work.