CritiQ-Bot – Automated Scientific Writing Review
University of Pittsburgh researchers have developed a novel multi-agent software solution, CritiQ-Bot to automate the screening and initial review process of scientific writing. Trained on an array of large language models (LLMs) that specialize in scientific writing, CritiQ-Bot can streamline the writing of documents along with the screening and review process for deliverables such as research articles, grant applications, and review papers. CritiQ-Bot can operate locally on the user’s instrument offering enhanced security and privacy. It provides feedback using an easy-to-use graphic user interface (GUI) enabling authors to improve the quality of their scientific writing prior to submission and reduce the number of post-submission revisions required.
Description
Scientific writing of original research articles, grant and research proposals, and review papers is a key requirement for academics. For non-native English speakers the challenges of scientific writing are amplified and can act as a barrier to career development. Additionally, journal editors and funding agencies are inundated with submissions, all requiring expert review. CritiQ-Bot is designed to assist authors, editors and funding agencies. Authors would benefit from initial reviews of scientific writing to accelerate manuscript preparation and submission. Editors and funding agencies could initially screen or triage submissions to ensure they meet selection criteria for journals or grants. CritiQ-Bot could revolutionize scholarly communication by streamlining the scientific writing process, reducing bottlenecks, and improving access and equity for non-native English speakers.Applications
- Scientific writing support- Screening and initial review process by scientific journal editors or grant agencies
Advantages
Current AI-tools designed to assist writers primarily focus on plagiarism, citation verification, statistical approaches and grammar. The LLMs currently available are rarely tailored to scientific writing and often require users to upload documents to cloud-based servers raising privacy and security concerns and could constitute a public disclosure of intellectual property.CritiQ-Bot is designed as a multi-agent software to assist the entire process from drafting to review post-submission. A triage agent initially screens the document to check for suitability for submission and style. A builder agent then performs at least two independent expert reviews of the work, providing targeted feedback that highlights strengths and weaknesses of the work, and suggesting suitable journals to aid submission choices. The “persona” of the reviewers can be tailored (i.e., methodology expert or copyeditor) to meet the needs of the user to tailor submissions and reduce the number of post-submission revisions. Crucially, CritiQ-Bot can work locally on the user’s computer, removing the privacy and security concerns related to cloud-based servers.